The campaniform sensilla of the wings are necessary for the regulation of wing twisting in locusts. Control of forewing twisting during periods of constant lift depends upon the hind-wing sensilla being intact, whereas the forewing sensilla are essential for stability about the three body axes.
The campaniform sensilla are located on the ventral surface of the wings. Two groups are present on the subcosta of the forewing and one on the subcosta of the hindwing. A few single sensilla occur on the costa. The cuticular parts of sensilla from distal and proximal forewing groups differ with respect to the lengths of their ovally shaped cuticular parts. The sensilla are orientated with the cupolas parallel to the long axis of the wing, except for those of the proximal forewing group, which are arranged in a fan.
During steady-state flights activity from campaniform units was high during the downstroke and low during the first part of the upstroke. Significant changes in the response were found to occur when the body angle was changed.
The effect of a 15° change in body angle on the motor output to the basalar and subalar muscles is described. At the start of a flight these motor patterns are rather unstable, units falling in and out. Stability is gained within 10 sec.
A distinct part of the response from the campaniform units in the hindwings can be abolished by the application of an anodic block. The duration of the blocking pulse equalled one-sixth of the flight period. The effect on regulation of forewing twisting varied according to the part of the response which is removed. Regulation could be abolished almost completely when the anodic block was introduced during the first part of the hindwing downstroke. When the anodic block is removed, twist regulation builds up again and is completed within 100-150 wing-beats.
Free flights including both intact and deafferentated animals made possible an evaluation of the importance of the different groups in stability reactions. It was shown that control of angular movement is accomplished by the forewing groups only, especially the proximal ones.
The integrative processes within the pterothoracic ganglia are discussed. During the constant-lift reaction, the slow, intersegmental reflex for control of forewing twisting seems to depend on central processing and memorizing of measurements of total lift. The sensory input from the hindwings is phasic and patterned, but it is still undecided whether it is the phase or the pattern which is the essential parameter.
In flying insects control of power by alteration of the setting of the wing relative to the air was suggested for Drosophila by Chadwick (1951), and in Schistocerca Weis-Fogh (1956b) discovered that lift was kept constant during extended flights and calculated that this was achieved by active regulation of wing twisting. Wilson & Weis-Fogh (1962) were able to photograph a change in forewing twisting with body angle resulting in an almost constant setting of the forewing relative to the wind, but it was not until the work of Gettrup & Wilson (1964) that experimental evidence was presented for the participation within this reflex of the campaniform sensilla of the wings. A study of the nervous pathways connecting the campaniform sensilla and the twist-controlling muscles in the desert locust, Schistocerca gregaria F., is presented below. A preliminary account has already been published (Gettrup, 1965 a).
When investigating a system as complex as the flight system the principal difficulty is that only the sensory input and the motor output are easily accessible, whereas central activity and patterning are almost impossible to study directly for the time being because so many active units are packed into so close a space within the thoracic ganglia. The only practical approach therefore seems to be an input—output analysis of the system as a whole.
In most experiments the locusts were suspended from an aerodynamic balance and flew against a wind from an open-jet wind tunnel (Weis-Fogh, 1956a). During periods of constant lift the body angle was changed in a controlled way, the sensory input was interfered with and the corresponding motor activity was recorded. Both sensory and motor events undergo cyclic changes at wing-beat frequency, and the principal method was to divide the flight period into discrete intervals and to study the effect of abolishing the afferent response by means of an anodic block placed within each of these intervals in succession. During free flights, another approach was to observe the control of rotation around the three body axes. Instability in pitch, yaw and roll was recorded in intact and in deafferentated locusts.
In Schistocerca the campaniform sensilla are situated on the wings as shown in Fig. 1. In accordance with Zachwilichowski (1934) campaniform sensilla were found to be absent from the upper surface of the wings. Most sensilla occur on the proximal parts of the subcostal veins in both forewings and hindwings, although in the fore-wings a few are located on the lower surface of the costa.
The sensilla are arranged in groups of different orientation, two groups in each forewing and one group in each hindwing. In Schistocerca the number of groups is low when compared with insects possessing greater manoeuvrability, namely higher Diptera, in which 6 groups are present both in the forewings and in the halteres. In the desert locust the sensilla within the distal group are uniformly orientated ; that is, the long axes of their ovally shaped cuticular parts are in parallel. In the proximal group of the forewing the sensilla are distributed with their long axes in a fan-like arrangement (Fig. 2). Angles of up to 65° have been measured between the long axes when seen in ventral view, but the actual figure may be as high as 78° in consequence of the way in which the sensilla are spread over the saddle-shaped proximal part of the costa and the subcosta (Fig. 2). A similar dispersion of direction of orientation has been recorded in Eristalis (Diptera) by Pringle (1948), who suggested that directional sensitivity was used for detecting and recording twisting forces.
The fan arrangement may not be the important feature. Thus, according to Pringle (1938, 1948), within any small area of the cuticle the compression lines or the extension lines run approximately in parallel. Therefore in order to display a directional spectrum of deformations in a structure in which cuticular compression lines may diverge by as much as 90° parallel sensilla must be present in several groups of different orientation whereas a single group of sensilla arranged in a fan on similarly shaped cuticle would suffice and would occupy a smaller area. The proximal group may be especially suited for recording wing twisting by reason of their proximity to the attachment of the controller-depressor muscles.
It is characteristic that the distal group consists of 62-65 sensilla as compared with 20-21 in the proximal group, whereas the hindwing group consists of 65-72 sensilla. Thus there are about 95 receptors in each forewing and about 70 in each hindwing,. but this is still low if compared with Coleoptera, Hymenoptera and Deptera, in which the corresponding numbers may be 6-7 times as high. On the other hand bad fliers such as the cockroaches may have only 20-40 sensilla, usually restricted to the forewings (Pringle, 1957).
It was not within the scope of this study to investigate the fine structure of the campaniform sensilla (e.g. Pflugstaedt, 1912; Thurm, 1964). However, some scattered observations and measurements were made of the shape and size of the cuticular parts, which are oval, dome-shaped structures, each placed as a lid over a hollow tube through the cuticle (Fig. 3). The lengths of the domes within a group display a normal statistical distribution. The two forewing groups were found to differ slightly from each other in average length of dome—9·4 μ (S.D. 1·9 μ) and 12·1 μ (S.D. 3·3 μ), respectively—and to be different in this respect from the hind wing group, 9·7 μ (S.D. 1·8 μ). The difference between the forewing groups is statistically significant and may indicate that the two groups respond to deformations of different magnitude (cf. Pringle, 1938).
The campaniform sensilla communicate with the C.N.S. through nerve I A of Ewer (1953). This nerve also receives axons from the setiform sensilla of the wing, amongst them those of the tegula. During its course towards the wing nerve IA first provides a branch to the tegula and, at the wind hinge, it divides into an anterior branch which innervates the sense organs of the costa and the subcosta and a posterior branch which supplies the radius and posterior regions of the wing (Zachwilichowski, 1934; and present work).
Accessory sensory fields and nervous pathways within the flight system have been described in several previous papers, especially those of Weis-Fogh (1956b), Wilson (1961), Gettrup (1962) and Guthrie (1964). Further anatomical information can be found in Snodgrass (1929), Misra (1950), Wilson & Weis-Fogh (1962) and Gettrup & Wilson (1964).
THE SENSORY RESPONSE
Flight records of the afferent discharge in the nerve IA of the mesothorax were made by Wilson (1961). The firing, including activity from hair sensilla and campaniform sensilla, occurred mainly during the downstroke, whereas the greater part of the upstroke was silent. The recorded discharge was dominated by potentials which derive from the tegula, and clues as to the importance of the campaniform sensilla during various phases of the flight period could not be made. Records have now been obtained of the flight responses in both mesothorax and metathorax of locusts deprived of their tegulae. Mesothoracic records were very noisy and in this study they have only been used for determining the time of occurrence of the discharge within the flight period.
The recording technique was as follows. In the mesothorax nerve I A was hooked on a monopolar platinum electrode (0·2 mm.) which was introduced through an opening in the prothoracic shield. A reference electrode was placed in the abdomen. The tegulae were destroyed by cauterization and the gut was removed, but otherwise the animal was intact. The top position of the forewing was recorded by means of a stroboscope connected to the oscilloscope. In the metathorax the technique was a little different. The abdomen was cut off and the pterothorax was sealed with petroleum jelly. The reference electrode was placed in the prothorax. The recording electrode was introduced from behind. The firing of the first basalar muscle of the metathorax (127 of Snodgrass, 1929) occurs almost at the top position of the hindwing and was used as a time signal.
As to timing in relation to the wing beat, records from the meso-and metathorax show the same characteristics, apart from minor differences. During normal flight at a body angle of 5° activity occurs during the entire downstroke of the forewing, whereas the first three-quarters of the upstroke is silent; i.e. during normal flight the campaniform sensilla start to fire about 6 msec, before the wing reaches its top position and stop again at the bottom position. Maximum activity is attained during the first part of the downstroke. In the hindwing firing starts earlier during the upstroke, about 12 msec, before top position in a normal wingstroke, and ceases at bottom position or just before (usually not more than 5-6 msec.). In both segments activity of small units may persist at the noise level during the upstroke.
The discharge is composed of firings from campaniform sensilla as well as from hair sensilla, but several findings indicate that the campaniform sensilla communicate with the C.N.S. through the bigger axons. At the wing hinge the anterior branch of nerve IA was found to contain 30-35 axons exceeding a diameter of 2·5 μ, whereas the posterior branch only contained about 10 axons within this size group. Occasionally axon diameters of campaniform sensilla were measured close to their origin from the soma, and they were often found to exceed 2·5 μ. Supporting evidence is supplied by experiments in which the entire wing was deafferentated by cutting the nerves in the wing except for those to the region of the campaniform sensilla; it was then found that records from intact and deafferentated locusts were substantially similar, and that the few detectable differences in the overall patterns originated from the fall-out of small to very small units.
Several reflexes are mediated through nerve I A. In adult locusts 1 or 2 weeks old the discharge from the tegula seems to have a tonic effect upon wing-beat frequency, similar to the effect of the stretch receptor at the wing hinge (Gettrup, 1962; Wilson & Gettrup, 1963). Regulation of forewing twisting by grading the activity of the controller-depressors was found to require intact campaniform sensilla (Gettrup & Wilson, 1964). However, the ablation was rather unspecific and no attempt was made to determine to what extent the different groups participated. The author has now proved that during a normal flight performance at constant lift the hindwing groups are essential for the regulation of the important twisting of the forewings. On the other hand, intact forewing groups are necessary for any steady flight performance. A full description will be given later in this paper.
THE MOTOR OUTPUT
The twisting of the wings has been described in detail by Weis-Fogh (1956b) and Gettrup & Wilson (1964). Wing twisting is controlled by 10 motor units in each segment (i.e. 5 motor units per wing), but only 3 units are important—the single unit of the first basalar muscle and the two units of the subalar muscle (Wilson & Weis-Fogh, 1962).
In animals exhibiting the constant-lift reaction active regulation of twist has been demonstrated only in the forewings, whereas all attempts to find a similar response in the hind wings have failed so far (Gettrup & Wilson, 1964). Changes in forewing twisting are mainly due to changes in activity of the subalar muscle (99 of Snodgrass). Its activity patterns at different body angles are described by Wilson & Weis-Fogh (1962), together with those of the basalars (97 and 98 of Snodgrass, 1929).
The contraction of the subalar muscle is graded in several ways. Usually, during standard flights and at a body angle of 15°, only one unit is active. A change in body angle from 15 to 0° is met by recruitment of the second unit and, in addition, the first may fire twice. The total number of firings, which depends upon the relative lift, is generally increased by 1, more infrequently by 2 firings per downstroke (Fig. 4). The power output of a single unit is graded by changing the distance in time between two firings within the same flight period (Neville & Weis-Fogh, 1963).
When the body angle is changed, or when flight starts, a constant output to the subalar muscle is attained only after 4-6 sec. The start is characterized by high but irregular activity of both units. Single, double, and even triple firing is often seen, and the number of firings changes from cycle to cycle (Fig. 5). Occasionally, a higher degree of constancy is achieved during the first seconds; that is, the burst pattern does not change from one cycle to the next, but several cycles may pass before a pattern change occurs.
The instability patterns of corresponding motor units from each of the four subalar muscles are not correlated.
INTERFERENCE WITH SENSORY INPUT
As to the relative importance of the two pterothoracic segments in relation to the twisting of the wings little was known except for the observation that sectioning of connectives and recurrent nerves between the respective ganglia results in cessation of wing twisting; but since this is associated with a very considerable decrease in power output it may be a result of a general decrease in the degree of excitation (Wilson, 1961). On the other hand, in higher Diptera control of rotation around the three body axes depends upon the halteres being intact. Forewing receptors may also be involved, but their task may be quite different (Fraenkel & Pringle, 1938; Pringle, 1948; Faust, 1952).
Partial interference with the input from the campaniform sensilla has now been investigated in Schistocerca. The technique was either to cauterize the groups of sensilla or to cut away their superficial cuticular parts. The locusts were suspended from an aerodynamic balance (Weis-Fogh, 1956a) and their ability to regulate forewing twisting was tested before and after operation. Additional support has come from experiments in which the campaniform sensilla were intact but immobilized to some extent by means of a small aluminium bar glued to the ventral base of the subcostal vein. These locusts behaved in the same way as the ones with cauterized sensilla.
Changes in wing twisting were recorded photographically (cf. Gettrup & Wilson, 1964) and could be caused by imposing a new body angle upon an animal performing the constant-lift reaction. A change in the setting of the wing was recorded as the change in the angle of a wing marker, placed as a chord perpendicular to the long axis of the wing approximately where the vannal fold meets the posterior edge of the wing. In this way the rear flap remains free to operate. The degree of twist regulation is expressed as the twist-control ratio, R. The ratio R is the change in the angle of the chord, measured in the plane of symmetry of the body when the wing is horizontal during the downstroke, divided by the imposed change in body angle. For body angles between o and 15° normal regulation of forewing twisting is characterized by an approximately constant chord angle regardless of body pitch. The true angles of attack were not measured. R is then about o when regulation is normal and about 1 when it is completely abolished. The results are presented as histograms in which R is displayed along the abscissa and the statistical frequency of observation on the ordinate.
Fig. 6 summarizes the results of 19 experiments of 115 observations in which all campaniform sensilla of either forewings or hindwings were destroyed by cautery.
Observations on intact and operated locusts are indicated by broken and continuous lines respectively. The frequency diagrams relate to steadily flying animals. When only the hindwings are interfered with the frequency maxima of samples before and after operation are clearly separated, and the experiments therefore demonstrate that the twist regulation of the forewings is abolished when the campaniform sensilla of the hindwings are destroyed. In locusts with cauterized forewing groups the situation is more complicated, and the frequency diagram does not disclose any characteristic features except for a tendency for either-or rather than uniform distribution within the range considered. It is difficult to interpret this because of the probable interference with the feed-back routes of the motor system, but it is safe to conclude that intact hindwing receptors are essential for the regulation of forewing twisting which is characteristic for the constant-lift reaction.
THE IMPORTANCE OF THE VARIOUS PHASES OF THE SENSORY INPUT
The relative importance of the different phases of the wing stroke was investigated by application at flight frequency of a periodic anodic block to the sensilla of the hindwing groups. The blocking pulses had a duration of one-sixth of the flight period (i.e. about 10 msec, in the case of a normal flight of a big female). In the hindwings the sensory response extends over two-thirds of the stroke period (i.e. about 40 msec.) and the pulse length was found adequate in order to strike a balance between time resolution and the effect upon the motor output of reduced afferent input. The forewing twisting was recorded photographically (Fig. 7b).
The blocking electrodes (0·1 mm. diameter platinum) were placed co-axially within the base of the subcostal vein of each hindwing (Figs. 3, 7b), whereas the cathodes were fixed inside the metathorax anterior to each wing hinge and near to the pleural process. Furthermore, recording electrodes were placed in the muscles 127, 129 and 99. The locust was grounded by an electrode in the abdomen. It was essential that these 7-8 electrodes were placed in position without further operation since a very long flight performance (5-6 hr.) is necessary to make all the measurements. The procedure was as follows. A stimulator (Disa Multistim) was triggered from the animal, using the action potential of muscle 129 (the subalar muscle) as the trigger source. The delay between trigger and output was adjustable. The stimulator had two independent channels, one for the anodic block and another for the triggering of a flash-lamp (General Radio Strobolume) used for the photographic recording of forewing twisting.
In a preliminary series of experiments records of the afferent discharge were made simultaneously with the application of the anodic block. It was shown that the sensory response could be almost completely abolished during the passage of a square-wave pulse representing 0·3 × 10-6 coulomb. The pulse length was 10 msec, and the rate of repetition was matched to flight frequency (17-20 cyc./sec.). Oxygen or hydrogen bubbles were not generated. In any case 0·3 × 10-6 coulomb transmitted across 2·5 mm.2 anode and 0·5 mm.2 (cathode) is far below the critical value for the nucleation of oxygen and hydrogen bubbles (cf. Rowell, 1963).
The experiments are difficult to perform and out of 15-20 attempts only 4 were successful. Only one locust performed the constant lift reaction long enough (namely for at least 4 hr.) for each of the six intervals to be tested, but fragmentary support has come from three other flights. None of these three locusts performed well and only one locust flew at constant lift for extended periods of time. In this animal R attained the values 0·9, 0·9 and 0·3 during the first three intervals (2, 3 and 4) of the downstroke and 0·1 during the first part of the upstroke, in the sixth interval. In the two other experiments the values of R were lower in the first and higher in the third interval from the top position. The most complete set of results from a single experiment is seen in Fig. 7,c, in which the vertical axis displays decreasing regulation of forewing twist. The horizontal axis represents one wingstroke period, the top position being indicated by an asterisk. An anodic block amounting to one-sixth of the flight period was moved around within the wing stroke cycle. The twist control ratio R is shown for each of the six wing-stroke intervals in Fig. 7 a. It is evident that the effect of the block on the twisting of the forewings changes significantly from phase to phase. The afferent discharge during the end of the upstroke and the first two-thirds of the downstroke is clearly more important than that during the remaining part of the cycle.
The twist-control ratio reaches its maximum of 0·83 during the first part of the downstroke (period 3), showing that the short anodic block can abolish the regulation almost completely (i.e. the discharge during these 10 msec, is of decisive importance for the constant-lift reaction). It is also characteristic that the period during which anodic blocking is effective coincides with the time when the lift produced by the hind-wings is increasing and reaches its maximum (Jensen, 1956). Also, from other experiments direct records of the response in the hindwing nerve show that the sensory discharge appears within the same period and reaches its maximum during the first half of the downstroke (Fig. 8). The flight period was divided into the six intervals of Fig. 7 a, c, and the number of sensory firings was counted within each of these intervals. Furthermore, some alteration in pattern of the sensory input was found to follow imposed changes in body angle, and these changes were limited to the four periods in which an anodic block was effective in producing lack of twist regulation.
The change in activity of the subalar muscle of the forewing caused by altering the body angle builds up slowly, the process being accomplished within 100-150 wing strokes. During the transitory period it was characteristic that the final pattern appeared only now and then at the start but occurred with increasing frequency until, at the end, it became dominant and present in each cycle.
As to the total number of firings during a wing stroke the motor output of the subalar muscle in the mesothorax seemed to change with body angle in the opposite direction from that of the hindwing input ; at a low body angle (o°) sensory activity was lower than at high body angle (15°). This was seen in an experiment in which the body angle was changed 15 times in the one direction and 15 times in the opposite direction. The number of potentials from the mesothoracic subalar muscle was counted both when the hindwing sensilla were intact and when they were blocked. The sensory discharge from the hindwings has been analysed in a comparable flight. The results are seen in Fig. 9, where natural and blocked inputs and their corresponding motor patterns have been compared. A change in body angle of 15° alters both sensory and motor patterns, and the new sensory pattern represents less activity; that is, a smaller number of potentials when changing from high to low body angle (and higher activity when changing from low to high body angle). Each count in Fig. 9 is an average of counts from ten consecutive sensory records. If an anodic block is introduced during the first half of the downstroke where it is most effective and the body angle is changed 15°, neither the blocked afferent pattern, nor the new but blocked afferent pattern evoked by a change to high body angle, produces a new motor pattern. The motor pattern is identical with the original one, although 30-40 % of the unblocked afferent discharge at high body angle has been removed, which shows that input specifications additional to number of firings (e.g. a pattern) is required in order to produce the motor pattern changes associated with regulation of forewing twist. The result is independent of the direction of change of body angle.
OBSERVATIONS ON FREELY FLYING LOCUSTS
A freely flying locust advances through the air with a speed of about 3·5 m./sec. The corresponding flight frequency is 17 cyc./sec. ; that is, the locust travels 21 cm./wing beat. A sudden externally or internally caused change in pitch, yaw or roll results therefore in an observable change of the flight-path unless instantly corrected (i.e. within 1 or 2 cycles). Observations were made of the flight of intact and deafferen-tated locusts in order to decide whether the C.N.S. receives and makes quick use of sensory information from the wings. The procedure was to cauterize the campaniform sensilla and to observe the start of flight. In locusts flying in the wind-tunnel the start is characterized by an unstable output of the controller—depressor muscles. The freely flying locusts were caused to fly by removal of the support and flew towards a lighted wall 10 m. ahead. Angular instability was recorded in 10 flights of each animal before and after each deafferentation (Table 1). More than 550 flights were observed and the table is based upon 300 flights. The animals were selected by test flights of which several hundreds were made.
Locusts deprived of the metathoracic groups are still able to perform a well-co-ordinated stable flight, whereas cautery of the mesothoracic groups definitely interferes with control of angular movements around the three main body axes. Irreversible instability thus occurs, whereas the intact animal can correct angular deviations within a flight path of 20-30 cm., that is, within one or two flight periods. Destruction of the distal forewing group is still compatible with stable flight whereas the proximal group seems to be essential for active stabilization, although some ability to regulate in the pitching plane persists. Furthermore, it is interesting to note that blinding has an effect in combination with unilateral cautery of both forewing and hindwing groups (Table 1). In the latter experiments the compound eyes as well as the ocelli were painted with black cellulose lacquer.
The proprioceptors at the wing hinge (Gettrup, 1962) proved to be of minor importance as regards control of angular movements. Vision and hearing seem to have little influence in locusts with intact wing sense organs, although Haskell (i960), studying the flight of blinded locusts in the field, found that in some of the tested animals instability became apparent a few seconds after the start. Rainey & Ashall (1953) and Goodman (1959) found traces of a dorsal light reaction in these insects. Selection of locusts by test flights may account for the difference between these findings and the results reported in this paper (Table 1). Only intact animals which maintained an almost straight flight-path were used for further experimentation.
The twist-regulating system in locusts includes reflexes involved in control of lift and stability. The reflex for control of lift is intersegmental and slow whereas the reflexes for control of angular movements seem to be fast.
Sensory inputs from forewing and hindwing converge upon the same motor units in the mesothorax. The input from the hinwdings is of decisive importance in regulation of forewing twisting during periods of constant lift.
Several experimental facts indicate the existence of a feed-back within the mesothorax. Wind-tunnel experiments have shown that locusts in which the campaniform sensilla of the forewings have been destroyed display some lability and unsteadiness in adjustment. As described earlier in this paper, there is a tendency for either-or; e. a change in body angle of 15° may be followed by a change in wing twisting of either o or 15°. However, quite independently of the final result, wing twisting usually attains the correct value momentarily during the transitory period. Further support is provided by observations on free flights and these flights show that the control of angular movements in which the hindwing sensilla do not take part is fast relative to the intersegmental reflex of the constant-lift reaction. Nevertheless, regulation of forewing twisting associated with constant lift depends on a phasic input, as shown by the anodic block. There is some indication that a phasic input is necessary in stability reactions since partial or total prevention of deformation of the campaniform sensilla by means of pieces of aluminium can block this type of control.
The afferent discharge from the hind wings could not be traced in the connexions between the two pterothoracic ganglia. If not due to the recording technique this implies that integration starts in the metathoracic ganglion. This is in agreement with Guthrie (1964), who found that the fibres of IA end in synaptic regions within then-own ganglion. However, his methods were morphological and it could not be excluded that some fibres continue into the longitudinal fibre tracts and pass on to the next ganglion. Multisegmental connexions of fibres from single afferent neurones have been demonstrated electrophysiologically within the abdomen of crayfish (Kennedy & De Forest Mellon, 1964). Each sensory unit communicates several times with intemeurones covering more than one segment.
Reflex control of forewing twisting during periods of constant lift seems to depend on a mechanism by means of which the lift contributions from the four wings via the inputs from the campaniform sensilla are summated, ‘memorized* and compared. Mechanisms which do not make use of a central reference are probably not realized since in wind-tunnel experiments at constant flying speed and constant lift locusts possess some plasticity as regards control of lift, so that the size of a parameter may change from flight to flight, and also parameters which do not normally vary may occasionally do so. Both in the mesothorax and in the metathorax such wing stroke parameters as amplitude, stroke-plane angle and stroke angle change to some extent with body angle so that, although suggestive, omission of a central processing by direct reflex monitoring of forewing twisting from the hindwings (which provide 70 % of total lift) would hardly suffice. There is no indication so far as to how information is stored in the central nervous system. The response of the campaniform sensilla is further discussed by Gettrup (1965 b).
On the basis of free flights with intact animals it is difficult to estimate the duration of the control reactions for angular stability, but the time-constants of the reflexes controlling forewing twisting during the constant-lift reaction and during the stability reactions respectively may differ by one or two orders of magnitude, and it is doubtful whether the same integrative mechanisms operate in both cases. Within the lift-controlling system integration seems to depend on a phasic input but, as previously shown, input specifications other than quantity seem to be necessary. A similar slow reflex was found to control wing beat frequency (Wilson & Gettrup, 1963). This reflex operates by averaging a phase-independent sensory input (Wilson & Wyman, 1965), but, whereas the analogy with respect to time constants is obvious, it is rather uncertain whether the input can be phase-independent in the lift-controlling system. Nor is patterning of individual afferent bursts essential for the operation of the stretch reflex, whereas anodic blockings show that the intersegmental reflex for twist control depends upon a certain input pattern. The results obtained by the anodic blockings are still compatible with the presence of a loose, central coupling as proposed for the flight system by Gettrup (1963) and Weis-Fogh (1964).
I would like to thank Prof. T. Weis-Fogh and Prof. O. Sten-Knudsen for reading and discussing the manuscript and Prof. Weis-Fogh for free working conditions. The technical assistance of Mr K. Larsen and Mr V. Rasmussen is much appreciated. The work was supported by the Danish State Research Foundation. The locusts were supplied by the Anti-Locust Research Centre, London.