Head Movements in Flies (Calliphora) Produced by Deflexion of the Halteres

The hind wings of the dipteran flies are reduced and take the form of small clubshaped appendages called halteres, which have at their bases rows of campaniform sensilla, arranged in several fields (Pflugstaedt, 1912; Pringle, 1948; Thurm, Stedtler & Foelix, 1975; Smith, 1969).

During flight the halteres oscillate up and down at the same frequency as the wing beat but in antiphase to the wings. In many, but not all flies, the halteres also oscillate during walking and in the absence of wing vibration (Schneider, 1953; Miller, 1977). The movement of the halteres during walking is certainly brought about by the action of the haltere muscles of which there are two for each haltere, a levator and a depressor (Schneider, 1953). In flight it is likely that the mechanical distortions of the thorax are important at least in maintaining the phase relationships between the wings and the halteres. It is not known what function the haltere muscles play during flight, nor what function the oscillating halteres perform during walking.

Muscid flies are not able to maintain stable flight if the halteres are removed. Haltereless flies are able to fly once more however if a thread is attached to the abdomen (Pringle, 1948). This experiment does not demonstrate the function of the halteres so much as to show perhaps that a fly, prevented from making rapid turns by the trailing thread, no longer needs the control of the haltere system.

The haltere system is presumed to operate like a gyroscope. Stable about its plane of oscillation, any yaw movements will be resisted by the physical inertia of the beating halteres. This results in a tendency for the track of the halteres to alter, a torsion at the base of the haltere, and a consequent activation of the campaniform sensilla situated there (Pringle, 1948).

The haltere nerve is purely sensory, is reported to contain about 380 nerve fibres and runs into the posterior part of the fused thoracic ganglion (Pringle, 1948; Power, 1948). The central projections of the axons have not been described^* nor is the central mechanism known by which the halteres achieve stability in flight. Imposing yaw on a blinded flying fly results in compensatory actions in the wings (Schneider, 1953) such that the thrust on the one side is increased. The net effect compensates for the imposed torque.

In this study we describe the results of anatomical, physiological and behavioural investigations on the haltere system. We show the extensive central projections of the haltere nerve, the presence of rapid reflexive pathways to wing and neck muscles and the action of the neck muscles which bring about the head movements. We also show that removing the halteres has a small but significant effect on the orientation of blind, walking flies. A model for the stablization of the fly during flight is proposed which incorporates the halteres, optomotor system and neck-proprioceptor organs.

All experiments were performed in the laboratory of H. Markl in the University of Konstanz.

The larvae of Calliphora erythrocephala were bought from a local supplier. They were allowed to pupate and hatch. The adult flies were given water and sugar and used for experiments from 2 to 8 days after hatching.

To measure the head movements produced by haltere deflexion the animals were first waxed by the thorax to a support and the wings and the membrane flaps covering the halteres were removed. The flies were given a small polystyrene ball to hold in their feet. Fine wire probes, attached to piezo-electric movement transducers (VALVO Pkl multimorph 4322 020 04930) were positioned behind the haltere on each side. The movement transducers were connected to a function generator. Head movements were measured either with a capacitative sensing device (Sandeman, 1968) which allowed unrestrained head movement, or with a strain guage (Akers 801) which allowed only small head deflexions. For blinding, the eyes of the fly were covered with white paint (Pelikan Plaka).

Cobalt chloride (6% in water) was introduced into the haltere nerve by cutting the haltere off at the base and placing a drop of cobalt solution over the cut end. The drop was then sealed over with petroleum jelly, and the preparation left for 1-3 h. In the most successful preparations (invariably those in which the fly remained alive) the cobalt had permeated as far as the cerebral ganglion after 1 h. The thoracic ganglion was removed from the animal, soaked for 10 min in a weak solution of ringer and ammonium sulphide, washed in insect ringer (Yamaoka, Masahiro & Hirao, 1971) and fixed in alcoholic Bouin’s or Carnoy’s fixative. After fixing, the ganglion was dehydrated and cleared in methyl salicylate. Successful preparations were drawn (camera lucida).

Cobalt chloride was introduced into the neck nerves of isolated ganglia in the following way. A ring of petroleum jelly was squeezed from a hypodermic syringe onto a Sylgard base. This well was divided in half with a thin petroleum jelly wall and the ganglion arranged so that the neck nerve lay across the dividing wall. The neck nerve was then sealed into the centre wall by adding another layer of petroleum jelly across it. Cobalt chloride solution was pipetted into one half of the well, and ringer into the other. Preparations were left for 1-3 h and then treated as described above for the haltere nerve.

Transverse sections of the neck nerves were made from material fixed first in Kamovski’s fluid (1 h) and then in osmium tetroxide (30 min). The ganglia were embedded in plastic (Spurs) and thin (1 µm) sections stained either with methyl-blue for the light microscope or with lead citrate and uranyl acetate for the electron microscope.

Electrical recordings from the cut ends of the nerves to the neck, wings and legs were made with glass suction electrodes, and conventional A.C. coupled amplifiers. The fly was pinned on its back and part of the sternum, anterior to the prothoracic legs, removed. This affords access to the anterior wing nerves, the prothoracic leg nerves, the neck nerves and the oesophageal connectives. A drop of saline covered the nerves in the prothorax during the experiment. Haltere movements were produced as described above.

Flies were fixed, with the tip of the abdomen removed, in alcoholic Bouin for several days. Dissections were made after hemisecting the thoraces with a sharp razor. Tissue and unwanted muscle blocks were removed with fine forceps. Drawings of the final preparations were made with the aid of a grid in the eyepiece of the microscope.

The participation of haltere sensilla in the perception of and orientation with respect to linear (gravity) and angular acceleration was tested in blinded, wing-clipped, free-running flies on a tilt-board or on a turntable, the slope e of which could be set from 0° (horizontal) to 90° (vertical). The flies’ tracks were marked on the paper-cover of the running-board by following the animals with a soft pencil ; angular relations between tracks before and after change of stimulus conditions were measured, and length and direction of the mean vector were calculated and statistically analysed according to normal procedures of circular statistics (Markl, 1964; Batschelet, 1965, 1978).

A fly which is suspended from the pronotum and holds a light-weight polystyrene ball in its feet, will alternately keep still, clean itself with the pro- and metathoracic legs or walk the ball around.

If, in a resting fly, the haltere is pushed sharply forwards with a fine probe, the head of the fly makes a small quick motion away from the side on which the haltere is stimulated. This reflex motion is produced only when the haltere is pressed forward and not when it is moved up and down or backwards. Repeated movement of the haltere at 15 Hz produces, in flies blinded but with their heads free, a relatively rapid head movement (about 50°/s) through an angle of about 4°. Movements are sometimes larger, sometimes smaller. Continued stimulation can cause a slow head deflexion of not more than 2° before a plateau is reached. In some cases, where the measurements were not interrupted by the fly beginning to walk, the head flicked rapidly back to the original position. Continued haltere stimulation caused the head to turn again away from the stimulated side. Thus, a series of head movements resembling somewhat the fast and slow phases of optokinetic nystagmus can be produced (Fig. 1 A, B).

When walking, the fly makes vigorous head movements, both up and down and side to side (Fig. 1C). As these movements are generally the same size as, or greater than the haltere induced head movements, all the measurements of haltere induced head movements were made in resting flies.

Fixing the head with the torque transducer and measuring the turning tendency of the head showed that for each deflexion of the haltere, the head twitches once to the side away from the stimulated haltere. These twitches are superimposed on a more gradual torque which is maintained during haltere stimulation (Fig. 1 A, B). Corrective saccades were not observed in fixed-head preparations.

In the above measurements, the probe was positioned so that it just rested against the haltere shaft before a stimulus was applied. It is clear from the records that in all cases an amplitude threshold exists, initial haltere deflexions being too small to produce the reflex. To test if this was related to the position of the haltere relative to the body, the stimulus amplitude was set to be subthreshold and the probe was then advanced against the shaft of the haltere. The result (Fig. 2C) shows that the neck twitches increase incrementally (up to 6 steps) in size as the haltere is forced slowly forwards, although the stimulus amplitude remains unchanged. This positional sensitivity is critical in an interpretation of the action of the halteres in flight (see Discussion).

A consideration of the head movements in flies requires a certain knowledge of the anatomy and action of the neck muscles. There are no previous accounts which attempt an analysis of muscle function and so it is necessary here to redescribe the muscles of Calliphora in spite of there being in existence careful anatomical accounts of the thoracic musculature of several kinds of dipterans (Luks, 1883; Lowne, 1890; Nussbaum, 1960; Mickoleit, 1962; Vater, 1962).

The head of Calliphora is supported by two relatively large neck sclerites. These are called ‘condyles’ by Lowne and they lie in the arthrodial membrane which forms the anterior part of the prothorax (Fig. 3). Medially, two smaller sclerities articulate with the condyles in such a way that movement of the condyles forces the two small medial sclerites down onto the hairs of the prosternal organs (Fig. 3). Laterally, the condyles articulate with the prothorax through two short, bar-like sclerites.

Dissection shows that the neck muscles all bring about some particular movement of the condyles. The muscles can be separated into four functional groups : (i) condyle retractors which pull the condyles back into the prothorax, (ii) condyle protractors which force the condyles out of the prothorax, (iii) condyle rotators which roll the condyles inwards and (iv) condyle-head muscles which hold the head firmly onto the condyles.

The action of the condyles can be appreciated from Figs. 4-6. The critical factor is that the condyles extend deep within the prothorax and that muscles from their posterior rims extend forwards to attach to the pronotum. Contraction of these muscles thus forces the condyle forwards and the head is turned out and away. The schematic diagram (Fig. 6) shows how angular displacements of the head are achieved by retracting and rotating the condyle on one side while extending that on the other side.

The motoneurones supplying the neck muscles emerge from the thoracic ganglion in two nerves, the frontal prothoracic (Nf) and the dorsal prothoracic nerve (Nd) (terminology anglicized from Vater, 1962). Where it emerges from the ganglion we have found that Nd contains only a single axon. It fuses however with the cervical nerve (Nc) and extends dorsally to innervate a set of muscles which rotate the condyle inwards.

Nf contains eight large axons ranging from 8 to16µm in diameter. In addition there are 28-30 axons of 3-4 µm in diameter and about 150 axons of between 1 and 2µm (Fig. 7). Serial sections show that the small axons are from the prosternai organs. The precise destinations of the various axons in Nf are not yet known. The lower branch of Nf extends out to the condyle retractor muscles, the upper branch projects to the condyle protractor muscles (Fig. 4).

Cross-sections of the haltere nerve reveal totals of between 330 and 390 axons, the largest of which are 8 µm and the smallest about 1 µm. Cobalt fills of the haltere nerve show that after the haltere nerve enters the ganglion, a large tuft of short branches extends medially, ending along the anterior margin of the ipsilateral metathoracic neuropile. No branches of the haltere nerve were seen to penetrate the contralateral metathoracic neuropile. More anteriorly, two bundles of axons project from the main stem across to the neuropiles of the contralateral pro- and mesothoracic neuropiles and end there. A tuft of very short fibres projects laterally from the main stem into the ipsilateral pro- and mesothoracic neuropiles. A bundle of sensory axons continues on through the oesophageal connective to end in the brain (Fig. 8 A).

Cobalt fills of Nf show that the motoneurones terminate in the ipsilateral prothoracic neuropile and have their cell bodies in the angle between Nf and the single oesophageal connective. The largest axons in Nf, some haltere nerve sensory axons, and the axons from the prosternal organs all end in the same area in this neuropile. Some smaller motoneurones in Nf ramify more laterally in the prothoracic neuropile. No contralateral projections were observed (Fig. 8B).

The anterior wing nerve (Nal, Nma) contains three large axons of which two probably supply the basalar muscles (Heide, 1968, 1971a, b and personal communication). Recordings from the whole anterior wing nerve during forward deflexion of the ipsilateral haltere reveal the presence of two large units which are activated. No activity is produced by contralateral haltere deflexion. The larger of the two units is more labile than the smaller and the shortest latencies to be recorded were 2 ms. Slow movements of the haltere are ineffective in activating the moto- neurones, but with a rapid haltere deflexion the motoneurones will follow, one to one, haltere movements of up to 200 Hz (Fig. 9 A).

(b) Neck nerves

Recording from the whole frontal prothoracic nerve (Nf) reveals a compound action potential made up of several large axon spikes. These large discharges occur only with ipsilateral haltere stimulation and only when the haltere is moved sharply forwards. Stimulation of the contralateral haltere evokes only small spikes (Fig. 9B).

The axon in dorsal prothoracic nerve (Nd) is excited by both ipsi- and contralateral haltere stimulation. All motoneurone responses whether evoked by forward deflexion of the ipsi- or contralateral haltere have a latency of 2-3 ms and are capable of following haltere vibration up to 200 Hz.

Electrically stimulating Nf by placing two fine wire electrodes on either side of the nerve and applying a train of pulses to it results in a movement of the head away from the stimulated side. This result supports the idea that activity of each neck nerve is predominantly concerned with turning the head away from that side by the activation of the condyle protractors. Haltere activation is cleaner than electrical stimulation because then not only are the protractors on one side activated alone, but also the retractors on the other side (see Fig. 10).

The action of the condyle rotators is not as clear. The axons which supply them (Nd) are excited by deflexion of both halteres.

Sometimes during recording the flies moved their leg bases as though walking. This was always accompanied by a massive discharge in the neck nerves and it was not possible to determine the effect of haltere deflexion during this time. There is evidence from studies on walking flies that the muscles systems in the neck are relatively inactive in resting flies, and come into play only during walking and, presumably, flying (Horn & Lang, 1978).

The only tonic responses which could be recorded from neck motoneurones were the result of prosternal organ stimulation. Only a single tonic unit could be found in each neck nerve and the destination of this motoneurone was not determined. Significantly the units responded only to ipsilateral prosternal organ stimulation, which fits with the observed unilateral central projection of the motoneurones and the prosternal- organ sensory axons.

The observation that the halteres oscillate when Calliphora walks appears to have been made initially by Schneider (1953). Miller (1977) has confirmed this observation for Calliphora and reported haltere oscillation in walking hippoboscid flies which are parasitic on birds and do not fly. It has also been reported that in the winter generations of some flies, the wings are reduced in size whereas the halteres are not. Whether these flies oscillate their halteres or not when walking is not known.

We collected flies of several species in the fields around Konstanz and found that haltere oscillation during walking occurred invariably in those dipterans with membranous covers over the halteres. Other species (including Drosophila)do not necessarily oscillate the halteres during walking in spite of having haltere levator and depressor muscles.

Calliphora spends a considerable amount of its time walking. Typically it will run rapidly and then pause, before running again. Quick turns are often made and thequestion was posed as to whether the oscillating halteres had any influence on walking and orientation with respect to linear (gravity) or angular acceleration.

The proprioceptive gravity receptor system (see Markl, 1974) of flies consists predominantly of pedal and antennal receptors (Horn, 1969; Horn & Kessler, 1975 ; Horn & Lang, 1978). Involvement of haltere receptors has been denied (Hom, 1969). This was tested in eight blinded and wing-clipped flies by letting them run on a vertically positioned tilt-board, where they readily ran either straight up or downward. When a fly had run ca. 10 cm straight up or down, the board was tilted rapidly end over end, and it was recorded whether the fly turned around and continued to run within ± 30° of its original direction with respect to gravity. Intact flies returned correctly in 145 out of 250 ‘tilt-runs’ which is highly significantly different from chance (χ²-test). After both halteres had been removed with their bases, return to original direction with respect to gravity after tilting was still highly above chance (116 out of 300 runs), but the difference in the proportion of positive responses to intact flies is also significantly different at P < 0 01. Therefore, although gravity orientation is perfectly possible without haltere proprioceptors, their removal clearly affects gravity orientation in flies under these experimental conditions.

A blinded, wing-clipped fly was released in the centre of a circular turntable that was set either at a slope of ε = 0° (horizontal) or ε = 60°. If the fly proceeded in straight direction from its release point, the angle with respect to 0° ( = upward for ε = 60°) at which it crossed a circle of radius 5 cm drawn around the starting point was recorded with 5° accuracy. Immediately upon this crossing, the turntable was rapidly (ca. 360°/S) turned by 90° to the left or to the right, and it was again recorded, where the fly-if it continued running - crossed a second concentric circle of radius 15 cm. From these two angular measurements, the difference in angle (Δα) between the fly’s walking direction before and after turning was determined (+ Δα, deviation to right ; – Δα, deviation to left). From these values of Δα, length α and direction a_T of the mean vector was calculated (α _T = 0° and a = 1 indicating a complete return to former walking direction after passive deflexion away from it). The results for intact and unilaterally or bilaterally haltere-less flies are given in Table 1. Since all data distributions were unimodal the significance of differences of a from chance (uniform) distribution was analysed by the Rayleigh test and the significance of deviation of from 0° according to Stephens. Differences between the mean vectors of different samples were tested according to procedures detailed in Markl (1964) and Batschelet (1965). The results can be summarized as follows:

1. Intact and uni- or bilaterally haltereless blinded walking flies reorient into the pre-turned direction after passive turning on a horizontal or on a sloping (ε = 60°) turntable (all a > o at P ≤ 0.01). The direction after reorientation deviates marginally in some samples (P = 0.05) from the original direction, with no clearcut tendency to larger deviations after loss of halteres. There is, however, a slight tendency to deviate to the right after right-turning and to the left after left-turning (under-compensation of passive turn). Reorientation is slightly (P = 0.05) better at the inclined than at the horizontal surface, but only in intact flies, not after haltere ablation.

2. There is a significant (P = 0.01) decrease in mean vector length, i.e. increase in orientational variance after passive turning, after loss of both halteres as compared to intact flies, while the difference after loss of only one haltere is not significant. This means that haltere mechanoreceptors are involved in perception of and orientation with respect to angular acceleration, although they are neither the only nor the most important sensory receptors involved.

3. In order to test whether loss of the halteres specifically affects orientation with respect to angular acceleration, or generally influences the co-ordination of walking, intact and bilaterally haltereless unblinded flies were compared on a horizontal plane with respect to their ability directly to approach a light-source after having been shaken into ‘escape-mood’ : both samples (three flies, 30 runs each a_r = 0°, 0° ; a = 0.94, 0.87) are statistically identical.

Therefore the loss of haltere mechanoreceptors directly affects the perception of angular acceleration although only to a slight degree. This agrees with the finding that electrical recordings from the leg nerves during haltere deflexion reveal no activity from fast motor systems such as are present in the wing and neck nerves.

Our most significant result is the demonstration of a rapid, possibly monosynaptic link between the haltere sense organs and the neck muscles, and the associated reflexive head turn which is brought about by haltere deflexion.

Before attempting a description of the head turning in flies during flight and a reinterpretation of the way in which the halteres operate, it is worthwhile mentioning several incidental issues which emerge from this study.

The wide distribution of the central projections of the. haltere nerve indicates a complex system of which the wing and neck reflex systems are only a part. The projections of the haltere nerve to the brain are as yet undescribed and the possible interaction of the haltere input with visual or antennal input is unexplored. The fairly massive projection to the metathoracic neuropiles is interesting, because there is some evidence of the halteres having their own intrinsic system of oscillators and sensory feedbacks which are unrelated to flight activity. It was frequently observed, for example, during leg movements, and in the absence of activity in the wing muscles, that the halteres oscillated. When strobed they were seen to be at times in phase, at times in antiphase, and sometimes only one haltere oscillated, the other remaining at rest. Touching the shaft of the haltere in a resting fly results in the haltere being drawn up and away from the stimulus. Flies which clean the abdomen with their hind legs, lift the halteres out of the way when doing so.

Given the presence of an intrinsic haltere system, what is the function of the different fields of campaniform sensilla at the haltere base? Could it be that only certain fields are used for the detection of torques relevant to angular acceleration detection, and that other fields are part of the system which monitors and maintains haltere oscillation when the animal is not flying?

The neck motoneurones certainly respond only to rapid forward motion of the haltere and are insensitive to any other haltere motion. This selectivity is remarkable and suggests that the organization of the sensilla at the haltere base is extremely refined. It is perhaps another example of an angular acceleration detection system where the sensory apparatus is very finely tuned as to the direction and magnitude of the angular acceleration, and provides an appropriate signal directly to the motoneurones. Exactly the same principle is found in the circular statocyst canals of the swimming crabs, where precise information about the direction and magnitude of angular accelerations is relayed directly to the eye muscles. There, intracellular recordings from the eye-muscle motoneurones have shown how the sensory information causes subthreshold fluctuations of motoneurone resting potentials which can be boosted to suprathreshold by the addition of non-specific mechanoreceptor inputs which arouse the animal (Silvey & Sandeman, 1976). A parallel is found in fly head movements : flies orientate their heads according to gravity only when walking (Horn & Lang, 1978). When resting the heads are more or less aligned with the bodies. If the head is weighted it hangs forward in resting flies but is held erect and in position in walking flies. Our observations of the massive discharges to the neck muscles which accompany leg movements fit well with Horn & Lang’s findings. We assume that normally the haltere-influenced neck movements are superimposed on this background.

The picture that emerges is that in active flies the heads are suspended by the neck muscles and their position is the result of the continuous activity in all the motoneurones. Ciné films of Calliphora also show the head to be slightly extended during flight and that saccades are made during turning (Land, 1975). This is apparently not true in Syritta, a hover fly, where the head is held clamped to the thorax (Collett & Land, 1975). In Calliphora the head is therefore unlikely to act as an accelerometer and proposed for dragon flies (Mittelstaedt, 1950).

The dominant input to the neck muscles from the halteres is fast and phasic. During angular acceleration in flight the haltere is beating at the wing frequency (150-180 Hz). If the haltere is regarded to be a stiff pendulum rather than a gyroscope, some estimates can be made about the resultant change in the path traced out by its tip. On the application of torque, the tip of a pendulum is deflected from its linear oscillation plane and describes an arc instead. The greatest deflexion from the original track occurs at the moment when the pendulum has reached its peak velocity, in other words, half way through the cycle.

During yawing the fly halteres would therefore be expected to experience an anterior/posterior force, twice in every cycle.

Combined with the fast acting phasic nature of the output to the neck motoneurones this would ensure a continuous train of action potentials in the appropriate neck motoneurones for as long as the angular acceleration lasted. With increasing angular acceleration the anterior/posterior force on the haltere will be greater and consequently also the angle through which the head turns, given that the same recruitment of units occurs in the dynamic situation as we have shown to be the case in the static condition.

Both halteres will be affected by such torques however, and here, perhaps lies the significance of the positional sensitivity of the halteres. We showed that the haltere has to move forward to a particular position relative to the body before the rapid forward displacements produced a head reflex. If the haltere system is working the way we propose, then such a positional threshold is necessary to ensure the operation of the input from only one haltere, given that both experience rapid anterior/posterior force during yawing. The forces operating on a haltere during yawing may be thus broken down into (1) a rapid pulsed anterior/posterior force, and (2) an overall positional change in relation to the body.

The repeated activation of a phasic system to convey a continuous input is used by the fly in another system. During flight the antennae sense the wind on the head by the deflexion of their joints (Gewecke, 1967). The receptor cells there are phasic but as the antennae oscillate during flight there is a continuous input which monitors, without adaptation, the force of the wind on the head.

In some of our experiments, we allowed the animal to turn its head while we repeatedly moved the haltere back and forth and recorded corrective saccades which returned the head to its original position. These saccades are similar to those which occur in the locust (Kien & Land, 1978) during optokinetically driven head movements, and, as in the locust, the return saccade is probably mediated in the fly by the neck proprioceptors. Our evidence for this is the lack of corrective saccades in insects where head torque was measured and head displacements were minimized. In relation to this it is interesting to note that the only tonic responses of neck motoneurones were produced by deflexion of the prosternal organ hairs. The corrective saccade is unlikely to have been caused by these motoneurones which, depending upon the muscles they innervate, would be more likely to produce a slow turn of the head. Little more can be said until the destinations of the motoneurones are better known.

Three separate systems are now known to affect the wing beat in a way which will alter the direction of flight of the fly. These are the visual system, the prosternal organs (neck proprioceptors) and the halteres.

The visual system will detect the slip of an image past the retina and will alter the input to the wing muscles in a way which will counteract the movement of the visual field. The visual system therefore tends to lock the fly onto some goal. A complete discussion of this system can be found in Reichardt & Poggio (1976) and Poggio & Reichardt (1976).

The prosternal organs detect the misalignment between the head and the thorax and it has been clearly shown that the wings will increase the thrust on the one side of the body so as to bring the body again into line with the head (Liske, 1977).

The prosternal proprioceptor system is therefore apparently unable to ‘unlock’ the visual system from the target (unless by a fast saccade) and normally satisfies its own feedback loop by aligning the body. The dominance of the visual system over the neck proprioceptor system is also shown in non-flying flies which will follow the movement of a moving visual field by turning their heads and introducing a misalignment between head and body. Again the neck proprioceptors can either align the body in freely moving flies or break the visual lock with a fast saccade. The visual system however has clear linkages with the neck muscles, which act to drive the head out of alignment, as well as those with the wings, which apparently act to restore the alignment, at least during turning in flight.

The halteres have long been known to increase the thrust on one side if the animal is turned about its vertical axis (Schneider, 1953) and the effect is to compensate for the imposed displacement.

Thus in flies the general goal seems to be to keep the head pointing at the target and then to align the body behind the head. In considering the three systems together we can take two situations which are common in flies : the maintenance of straight and level ‘goal oriented’ flight in the face of imposed disturbances, and active turns.

Imagine a fly in free flight which is subjected to sudden yaw by wind currents. The whole fly with head and body in line is suddenly pointing in the wrong direction. The angular acceleration about the vertical axis affects the halteres which, because of their gyroscope properties, tend to maintain the original flight direction. Thus, if the animal is yawed to the right, the right haltere experiences a pulsed forward displacement which in not more than 3-4 ms drives the head left, i.e. along the original flight direction. That the initial yaw and the corrective head movements are rapid is critical. Slower drifts could be corrected by the optokinetic system. A parallel can be found in the crab eye movement system where the optokinetic and statocyst inputs controlling eye movements have different dynamic ranges (Sandeman, 1977), the optokinetic system coping with slow and the statocyst with fast compensation.

When in the fly the head is realigned with the original flight direction the optokinetic system will ensure that the ead is anchored there and the prosternal organ will align the body with the head.

Rapid, active turns are characteristics of flying Calliphora and other muscids which are dependent upon their halteres for stable fllight. The turns in Calliphora are always accompanied by a sharp turn of the head and a misalignment of the head with the body. There is some doubt as to whether the head turns before or at the same time as the body (Land, 1973; Geiger & Poggio, 1977) but this may be irrelevant here. The significant point is that all authors agree on there being a period during which the head is pointing in a direction different to the body. During such a turn the animal experiences a considerable angular acceleration about its vertical axis. After such turns, they continue to fly straight ahead and there is no tendency to spin (see traces of Fanma\Land & Collett, 1974). The effect of a sudden left turn, for example, on the halteres would be for them to mediate a rapid turn of the head to the right, but only after the left turn has been made. Therefore, a few milliseconds after its active turn, the fly is actually prevented from turning further and going into a spin. Also the tighter the initial turn, the greater the effect on the halteres and the more powerful the correction. The analogy here is not with the well known auto-pilot used in aircraft, but with a system which compensates only after an aerodynamically dangerous turn has been made. Thus the fly does not have to sacrifice its manoeuvrability in order to remain stable. The interpretation now of the classical thread experiment is that the trailing thread prevents the animal from making a dangerous turn and that haltereless threadless flies are able to fly only until they attempt the first rapid turn, when they spin in.

The behavioural measurement on orientational performance of walking flies reinforces the conclusion that the haltere mechanoreceptor system is predominantly involved in flight control, notwithstanding the fact that significant, though slight, influences on perception of and orientation by gravity and angular acceleration could be demonstrated, which make haltere mechanoreceptors another part of the complex proprioceptive system that controls orientation in space in insects (Markl, 1974).

We are indebted to the Alexander von Humboldt Stiftung for support. We thank Monika Winter and Christine Dittrich for their assistance with the light and electron microscopy, and Bernd Haack for help with the statistical computations.