Flight-motor-driven respiratory air flow in the hawkmoth Manduca sexta

Investigations of insect respiration using modern sensors and recording techniques have been performed predominantly in resting insects and in developing instars (Gunlinson and Harrison, 1997; Hadley, 1994; Harrison, 1997; Hetz et al., 1993; Hetz et al., 1999; Kestler, 1984; Lighton, 1996; Slama, 1976, 1988; Wasserthal, 1976; Wasserthal, 1981; Wasserthal, 1982; Wasserthal, 1996; Wasserthal, 1999). Although flying is one of the most fundamental characteristics of higher insects, only relatively little experimental work has focused on the supply mechanism for the flight motor in insects. Because of methodological difficulties, measurements of intact unnarcotized insects during activity, especially during flight, are almost lacking.

In insects such as locusts, hymenopterans, scarabaeid beetles and hawkmoths, the abdominal pumping movements observed during flight are assumed to contribute to respiratory gas exchange in the flight muscles (Snodgrass, 1935; Fraenkel, 1932a), but the effects of this abdominal pumping on tracheal ventilation of the thorax have rarely been measured. Our knowledge regarding respiration during insect flight is based mainly on the work of P. L. Miller (Miller, 1960; Miller, 1966; Miller, 1974; Miller, 1981) and Weis-Fogh (Weis-Fogh, 1964a; Weis-Fogh, 1964b; Weis-Fogh, 1967).

In locusts, haemolymph pressure data gave rise to the autoventilation model (Weis Fogh, 1964a; Weis-Fogh, 1967). Air is assumed to be sucked in and blown out equally through all the thoracic spiracles by deformations of the air sacs around and between the flight muscles as a consequence of wing movements. In locusts, a unidirectional flow between the anterior body and the abdomen caused by abdominal pumping was measured that contributed to the tidal flow of autoventilation (Miller, 1960; Weis Fogh, 1964a; Weis-Fogh, 1967).

In flying cerambycid beetles, the volume of air passively entering the exposed anterior spiracles and passing through the primary tracheae was analyzed and termed ‘through-draught ventilation’ (Miller, 1966). In these beetles, this mechanism is assumed to be combined with autoventilation of the smaller secondary tracheae.

As powerful fliers, hawkmoths are capable of increasing their metabolic rate by up to 148-fold from rest to full flight (Bartholomew and Casey, 1978). They therefore have, like hummingbirds (Berger and Hart, 1972), one of the highest recorded metabolic rates. It is unclear how hawkmoths manage to meet the increased O₂ demands during flight with their tracheal supply system, and the mechanism of air flow remains unresolved. Remarkably, the O₂ content of the flight muscles during steady flight exceeds even the resting level (Komai, 1998). It is the aim of the present study to contribute to the understanding of this very efficient supply mechanism. Special care was taken to use healthy moths and to avoid invasive techniques and unphysiological conditions as far as possible.

Since commercially obtainable Manduca sexta often show genetic deficiencies as a result of inbreeding, I used offspring derived from a laboratory stock from the Department of Animal Physiology, Marburg, Germany, that has been ‘replenished’ with stocks from Seattle and Cologne. Only individuals that were relatively long-lived and strong flyers were used in experiments. The adults were kept in a flight room prior to the experiments and were allowed to feed from artificial flowers containing a 15% honey solution. For experiments, 5- to 10-day-old healthy moths (N=6; body mass range 1.4–2.6g) were selected.

Moths were suspended at the descaled mesoscutellum with elastic layers of ‘Pattex’ (Henkel, Düsseldorf, Germany) from a rigid rod. Narcosis was avoided throughout all procedures. Instead, preparatory steps were carried out very gently, allowing recovery periods of several hours. To stimulate flight, the normal nocturnal activity rhythm was used; dimming of the room light was sufficient. The moths frequently flew for several hours without further encouragement, such as using a fan. The moths remained suspended for the entire experimental period over 3–5 days. They were fed ad libitum with a 12–18% honey solution at the beginning and sometimes again at the end of the recording session, which lasted several hours. At the end of the experiments, the moths were still healthy and could be separated from the mounting rod. All moths were used first for tracheal pressure measurements and then for CO₂ emission measurements. The experiments were performed in a Faraday cage at 20±1°C.

The wingbeat was recorded by projecting the shadow of the left or right or both moving wings onto one or two silicon photocells (Conrad 55 mm×20 mm or Telefunken BPY 10 mm×3 mm) installed on the bottom of the Perspex specimen chamber. For high-resolution analysis of wing position in relation to the course of a tracheal pressure pulse, the shadows produced by the wing were projected on two vertically arranged sensors. One sensor was shaded by the wing during upstroke, the other one was shaded during downstroke. Both sensors show the complete wingbeat cycle, but the arrangement was optimal when the curves of both sensors were symmetrical. Only the curve of the sensor shaded during the downstroke was used for the documentation of wing movement (see Fig.6). For this high temporal resolution, the sampling rate was increased to 40kHz; in contrast, 400Hz was used during most long-term measurements. The photocell voltage signal showed a delay of 2.5–3ms relative to the pressure pulse which was accounted for in the analysis of the recordings (e.g. in Fig.6). For observations of thoracic deformation and spiracular activity, the left side of the thorax was descaled and illuminated using an industrial strobelight (Drello, Mönchen Gladbach, Germany).

The cuticle around both first (=anterior=mesothoracic=SpI) spiracles was carefully descaled, and polyethylene tubes (1 mm external and 0.5 mm internal diameter) were tightly glued using ‘Fixogum’ rubber cement (Marabu, Tamm, Germany) around the perimeter of the intact spiracle. The tube from each side could be connected independently directly to a pressure sensor (Sensym SCXL 004 DN) or both tubes could be connected to the smaller compartment of a split-specimen chamber and their outputs combined for either pressure or CO₂ emission measurements (Fig.1). Blocking both anterior spiracles with pressure sensors at the same time induced intermittent flight, probably as a result of anoxia (see Fig.5A). Long-term pressure measurements were therefore conducted with only one of the anterior spiracles connected to the pressure sensor or with the sensors intermittently disconnected. Optimal flight temperature was assumed to be attained in the experiments after the transition from shivering to full flight; in a recent study using Agrius cingulatus and Hippotion celerio, contact thermometric measurements at the aorta have shown optimal flight temperature to be reached after 3–4min of shivering and to be maintained during typical flight phases of 3–6min, even when they were interrupted by short pauses (L. T. Wasserthal, in preparation).

As the second (posterior thoracic=metathoracic=SpII) spiracles are integrated into the wing hinge and are not accessible during flight, it was impossible to fix tubes to these spiracles. The intratracheal pressure in the posterior thorax was therefore measured at the mesoscutellar air sac on the rear of the mesotergal pulsatile organ (POII) on top of the dorsal longitudinal muscles (DLMs) (Fig.2). For this purpose, an ‘artificial spiracle’ was created by puncturing the cuticle and the underlying air sac. A polyethylene tube was connected to the hole in the air sac and fixed to the mounting rod to prevent positional changes of the tube with respect to the air sac that could cause artificial pressure pulses. The electronic pressure data were calibrated using a mechanical barometer (2500Pa total scale). The low-pass filter effect of the pressure sensors with increasing pulse frequencies was analyzed using a woofer-loudspeaker that broadcast certain frequencies inside the specimen chamber calibrated with a Bruel & Kjaer 0.5inch microphone to a 2331 sound level meter. The amplitude correction factor for two sensors was 1.37 and 1.86, respectively, at a pressure pulse frequency of 26Hz. Scales in the figures represent the corrected pressure amplitudes.

CO₂ measurements were performed in a split-specimen chamber with a controlled constant-speed air flow (60lh⁻¹) and adjustable pressure at a temperature of 20±1°C (Fig.1). For smaller individuals (mass 1.4 and 1.6g, N=2), a specimen chamber with a working volume of 83 mm×94 mm×62 mm was used. For larger moths (mass 1.9–2.6g, N=4), a cylindrical specimen chamber with an inner diameter of 112 mm and a height of 65 mm was used. To obtain the best possible CO₂ measurements, it was important to keep the chamber volume as small as possible. Care was taken that the movements of the wings were not restricted by the chamber wall or by the mounting rod or attached sensors. Flight was initiated some hours after the moths had been mounted in the chamber. Lateral openings in the chamber allowed the moth to be fed and the tissue patch provided for foot contact to be manipulated to quieten the moth at the end of a flight session. The tubes leading from the anterior spiracles (SpI) were connected to the small anterior compartment of the split-specimen chamber. The posterior thoracic (SpII) and abdominal (Spabd) spiracles opened into the larger posterior specimen compartment.

For CO₂ measurement, the air flow from either the anterior (Fig.1A) or the posterior compartment (Fig.1B) was conveyed directly to an infrared gas analyzer URAS 3G (Hartmann & Brown) while the air from the other compartment was conveyed to a CO₂-absorbing vessel containing NaOH. The pressure within both compartments was normally adjusted to the same value (between 10 and 100Pa). The detection limit was approximately 0.05μls⁻¹ in the smaller specimen chamber and 0.1μls⁻¹ in the larger one; the response time of the system was 1.4±0.2s. The baseline of the CO₂ analyser and system was checked for drift after each experiment without the moth. In some experiments, the pressure of the posterior chamber was slightly increased to avoid an artificially high pressure at the anterior spiracles that would have prevented CO₂ emission at SpI and, thus, might have caused unnaturally high CO₂ outputs at SpII. A higher pressure in the posterior chamber should produce increased outflow through the anterior spiracles. The CO₂ output was calibrated for each pressure regime and specimen chamber after the experiments using a motor-driven 50ml syringe simulating the release of a constant volume of pure CO₂ at different rates (0.1–20μls⁻¹). Data were acquired using an eight-channel MacLab interface, and calculations were performed using Chart 3.63 software on Power Macintosh computers.

After descaling and cleaning, the thoracic spiracles were examined using a binocular microscope and photographed using a custom-made light scanning microscope and a field-emission scanning microscope at 2kV accelerating voltage (Hitachi S 800). For scanning electron microscopy, specimens were air-dried, gold-coated for 3min under argon plasma at 25mA and 2kV (Hummer JR) and glued with colloidal silver to aluminium stubs.

In the hawkmoth Manduca sexta, intratracheal pressure pulses at the anterior spiracles and mesoscutellar air sac are synchronous with the wingbeat (see Fig.4, Fig.6). Mean frequency is 27±3Hz (mean ± s.e.m., N=6). During shivering, the mean pressure oscillates around ambient with an amplitude ranging between approximately 20 and 100Pa (Fig.3A). During steady flight, the pressure amplitude increases in conjunction with the increased wingbeat amplitude to between 50 and 250Pa at the anterior spiracles (SpI) and between 100 and 450Pa at the mesoscutellar air sac (Fig.3B, see Fig.6).

At the anterior thoracic spiracles, the mean pressure is between −10 and −70Pa (Fig.3, Fig.4, Fig.5A, Fig.6). In contrast, in the mesoscutellar air sac during steady flight, the mean pressure is between +20 and +50Pa above atmospheric pressure (Fig.3, Fig.4, Fig.5B, Fig.6). The resulting pressure gradient ΔP between the anterior spiracles and the posterior dorsal air sacs is 30–120Pa. This suggests that inspiration of fresh air at the anterior spiracles and expiration at the posterior spiracles will occur.

A single wingstroke in Manduca sexta lasts between 32 and 41ms. Analysis of pressure changes during a single wing stroke showed that the intratracheal pressure minimum at both measuring sites coincided with the second half of the downstroke (Fig.6). The pressure maximum occurred during the last third of the upstroke. The maximum and minimum pressures at SpI and the mesoscutellar air sac often coincide but may also be slightly shifted in time, with the maximum at the mesoscutellar air sac occurring 1.2–2.5ms earlier than that at SpI. The pressure minima of the mesoscutellar air sac and anterior spiracles could be shifted in both directions by 0.8–1.3ms.

To investigate whether the observed pressure difference between SpI and the mesoscutellar air sac affects the pattern of air flow in the thorax, CO₂ measurements in the split-specimen chamber were performed with the same air flow speed and pressure in both chambers. No CO₂ emission was recorded from the anterior spiracles, which opened into the anterior chamber (Fig.1A, Fig.7A). All recorded CO₂ emission was from the posterior spiracles (Fig.1B, Fig.7B). Even when the pressure in the posterior chamber was higher (ΔP up to 25Pa) than in the anterior chamber, the CO₂-containing air stream was not reversed (Fig.8, Fig.9). Only when the pressure in the posterior chamber was artificially raised to more than 25Pa above that of the anterior chamber was CO₂ expired at the anterior spiracles. The moths were only capable of short periods of flight under these imposed pressure differences (Fig.9). Under these conditions, CO₂ emission increased at the anterior spiracles during pauses between flights and decreased again during flight periods with a latency of 1–1.5s. A complete reversal of CO₂ output from the anterior spiracles could not be achieved by the application of a higher pressure to the posterior chamber.

The anterior spiracle (SpI), the mesothoracic spiracle, is situated in the flexible intersegmental membrane between the prothorax and mesothorax. It is hidden under scales and its orifice is protected from dust or particles by a pair of peritrema lamellae (Fig.10A). Each lamella consists of approximately six rows of cuticular processes connected by anastomoses, thus forming a three-dimensional meshwork (Fig.10C,D). The position of this spiracle is not significantly affected by the wingstroke. Closing and opening of the inner valve can be observed only after removal of the two peritremal filter lamellae (Fig.10B). The pressure curves (e.g. Fig.3) suggest that these valves are held open during steady flight. Changes in amplitude of the pressure signal are correlated with changes in wingbeat amplitude (Fig.5A), with no superimposed effect of valve opening or closing observed.

The posterior thoracic spiracle (SpII) has no peritrema filter apparatus, instead having a single external valve flap at the anterior side of the orifice and spinose perispiracular cuticle (Fig.11). This valve is assumed to be closed by a ventral muscle and opened as a result of cuticle elasticity (Nüesch, 1953; Eaton, 1988; Nikam and Khole, 1989). However, it has been overlooked that the functioning of this spiracle is affected by wing movements. It is located deep within the subalar intersegmental cleft, which is subjected to positional changes during flight. These movements and their influence upon SpII were observed under stroboscopic light and viewed from slightly below. The wing movements of the flying moth were ‘frozen’ by interference between the flash rate and the wing-stroke frequency (approximately 27Hz), and changes in the position and accessibility of the metathoracic spiracle were visualized. The intersegmental cleft becomes constricted during downward bending of the wings by adduction of the metathorax to the mesothorax (Fig.11B, Fig.12A). SpII becomes enclosed in the subalar cleft, and the external valve lip is probably pressed against the opposite soft intersegmental cuticle, closing the tracheal opening. In some individuals, a closing reflex of the spiracular lip could be induced by passive downward bending of the wings, and its wide reopening by wing lifting was clearly observed. During the upstroke and when the wing oscillates in the more upright position, as during shivering, the spiracle was visible with the external flap open (Fig.11A,C, Fig.12B).

Hawkmoths are well known for their rapid and long-lasting flight. Regarded as among the most powerful flying insects, they are reminiscent of hummingbirds. The metabolic rate of Manduca sexta of 237mWg⁻¹bodymass (Casey, 1976) is similar to that of the hummingbird Amazillia fimbriata (232mWg⁻¹; Berger and Hart, 1972). The present results go some way towards explaining the high O₂ content found in the active flight muscle (Komai, 1998), exceeding even the resting level in Agrius convolvuli. During steady flight, air is inspired through the anterior spiracles and expired through the posterior thoracic spiracles. Although the pressure maxima at the anterior spiracles are slightly positive (e.g. see Fig.4), the negative mean pressure explains why CO₂ emission was not recorded from the anterior spiracles without further manipulation. The mean pressure difference between the anterior spiracles and the posterior air sacs (which are directly connected to the posterior thoracic spiracles) is obviously responsible for the unidirectional air flow. This air flow is so strong that an imposed pressure difference of more than 25Pa was necessary to reverse it. Only under such an artificially increased counter-pressure in the posterior specimen chamber is CO₂ release at the anterior spiracles induced. Under these conditions, only intermittent flight occurred; the rate of CO₂ emission increased during the flight pauses by diffusion and decreased during flight phases (Fig.9). This shows that the flight motor worked against the diffusive CO₂ outflow from the anterior spiracles even under this extreme artificial pressure difference; it was not possible to redirect all CO₂ emission through the anterior spiracles during flight. It cannot, as yet, be determined whether all the CO₂ measured was released from the posterior thoracic spiracles alone. Some CO₂ may also be emitted from the first abdominal spiracles, which are connected to the thoracic tracheae. However, the first abdominal spiracles have an even denser peritrema filter structure than the anterior thoracic spiracles, suggesting that they might serve for inspiration rather than for expiration.

The unidirectional air stream observed in steady flight is not operative during shivering, when the mean tracheal pressure at the anterior spiracles and the posterior air sac oscillates around atmospheric pressure and the amplitude of the pressure pulses at the anterior and the posterior spiracles was only one-third to one-quarter of the amplitude during steady flight (Fig.3). During shivering, the mechanism is similar to autoventilation with a two-way (tidal) flow, as observed in Schistocerca gregaria (Miller, 1960; Weis-Fogh, 1967). In locusts, wing movements and the corresponding up and down movements of the nota result in air moving into and out of the anterior and posterior thoracic spiracles. In Manduca sexta, this flow becomes directional only with the greater downstroke amplitude that occurs during steady flight (Fig.4, Fig.5). At the anterior spiracles, as wingbeat amplitude increases, there is an increase in pressure amplitude correlated with a decrease in the mean pressure (Fig.5A). At the mesoscutellar air sac, an increase in wingbeat amplitude is correlated with an increase in the mean pressure (Fig.5B).

Komai (Komai, 1998) states that ‘a hawkmoth has no shunt mechanism in the primary tracheae’ and that ‘it does not use unidirectional ventilation flow and it is not known whether other large insects use unidirectional flow’. However, the experiments in which an artificially high pressure was imposed on the posterior spiracles from outside and induced emission of CO₂ through the anterior spiracles (Fig.8, Fig.9) provide further evidence for a unidirectional air flow. The minimum counter-pressure in the posterior chamber (ΔP) necessary for an air flow reversal to overcome the internal pressure gradient produced by the moth was approximately 25Pa. This value corresponds to the lowest measured mean pressure gradient between the anterior spiracle and mesoscutellar air sac. This reversal of air flow must take place via longitudinal tracheae. Such a connection between the mesothoracic and metathoracic spiracles has been described in Manduca sexta (Eaton, 1988). Lateral primary tracheae with a simple spirally coiled intima could be traced from a complete histological series of transverse sections through the thorax and anterior abdomen of a smaller hawkmoth species Proserpinus proserpina (L. T. Wasserthal, unpublished data). These tracheal stems clearly connect the anterior spiracles with the air sacs of the mesothorax, the metathorax and the first abdominal segment and, thereby, the spiracles of these segments. Gas exchange during flight, under continuous mean negative pressure at the anterior spiracles, can only function with a corresponding mean positive pressure at the posterior spiracles.

The intratracheal pressure maximum occurred at the end of the upstroke, and the haemocoel pressure measured in the locust was also greatest at this point (Weis-Fogh, 1967). This does not support the suggestion that the air sacs are compressed by contraction of the dorsal longitudinal muscles (DLMs) during the downstroke in Agrius convolvuli (Komai, 1998). According to the generally accepted model of indirect flight muscle systems (Pringle, 1975; Chapman, 1998), the thorax of hawkmoths is deformed by the action of the flight apparatus as follows. In Manduca sexta, wing upstroke is correlated with a slight flattening of the thorax caused by contraction of the dorsoventral muscles (DVMs), thus compressing the large air sacs lying between the tergites and the DLMs. These air sacs are continuous with the branching intramuscular air sacs (Fig.2). During the downstroke, the thorax becomes slightly arched and shortened by contraction of the DLMs, increasing the volume of the posterior and dorsal thoracic air sacs. Thus, during the downstroke, the intratracheal pressure is negative (Fig.6). During the downstroke, the posterior thoracic spiracles are covered by the intersegmental cleft and are probably closed by the valve flap fitting against the opposite cuticle, while the anterior spiracles remain open (Fig.11B, Fig.12), so air can be inspired unhindered only through the anterior spiracles and will be sucked posteriorly and dorsally into the large air sacs. Thus, a relatively simple mechanism involving volume changes of the thoracic air sacs together with the prevention of air inflow into the posterior thoracic spiracles is responsible for the retrograde air flow through the pterothorax. Whether the closing muscle of the posterior thoracic spiracle (Eaton, 1988; Nikam and Khole, 1989) is also involved in the closing mechanism during the downstroke needs to be investigated. Active closing and opening could be observed to result from passive wing bending in some individuals. It is probable that contraction of this closing muscle is synchronized with contraction of the DLMs, but a passive mechanism of valve closure may also be involved because the valve flap has no perforations and is unlikely to remain open against inspiratory suction, especially when the flap is close to the opposite cuticle. The role of the first abdominal spiracle, which communicates with the thoracic tracheal system, in the air supply mechanism is unclear. The first abdominal spiracles have an even denser peritrema filter structure than the anterior thoracic spiracles, so they are likely to serve for inspiration rather than for expiration.

The morphology of the posterior thoracic spiracle and its remarkable differences from the anterior spiracle have generated relatively little interest. The location of the posterior thoracic spiracle in the subalar cleft immediately below the wing hinge indicates that it is coupled to wing movements. In contrast to the extremely dense filter apparatus of the anterior spiracle (and of the first abdominal spiracle), the absence of any filter structures suggests that the posterior thoracic spiracle is adapted mainly for expiration. In contrast to all the other spiracles, it has no inner valve. It is only present after metamorphosis; caterpillars and pupae lack a functional spiracle on the metathoracic segment. The spinose margin of the valve lip and the spinose perispiracular cuticle could serve to prevent tight adhesion when the valve lip touches this soft cuticle.

Although the anterior spiracles are also open during the upstroke, expiration through the posterior spiracles may be easier because of their vicinity to the large posterior air sacs and because of the absence of filter structures around SpII, both of which should result in less resistance to the air stream.

In other insect orders with similar fundamental differences between the anterior and posterior thoracic spiracles, a similar unidirectional air stream might occur during flight. While, in resting insects, observations of spiracular closing and opening behaviour have led to the conclusion that respiratory air flow is unidirectional during abdominal pumping movements (for references, see Mill, 1985), only few studies using split-chamber experiments have measured the effects of such directional air streams (Fraenkel, 1932b; Bailey, 1954; Wasserthal, 1996). For flying insects, three mechanisms of tracheal ventilation have been described: autoventilation with tidal flow, abdominal pumping and passive stream by the ‘Fahrtwind’. The latter has been measured in cerambycid beetles and was termed the ‘through-draught mechanism’ (Miller, 1966). This retrograde air flow is suggested to be caused by the air stream passively entering the exposed anterior spiracles. A similar passive inflow of air into the anterior lepidopteran spiracles is unlikely because of the presence of dense scale layers and the peritrema filter apparatus. In flying locusts, a retrograde air stream produced by abdominal ventilatory movements is superimposed on the two-way autoventilation system (Miller, 1960; Weis-Fogh, 1967). In this case, all the thoracic spiracles are open during flight. The rate of flow through the spiracles is identical in both directions, so that no direction-sensitive valves are involved. The tidal flow of autoventilation mainly involves spiracles 2 and 3, while during abdominal pumping air is assumed to supply mainly the head via spiracles 1 (inspiration) and 5–10 (expiration) (Miller, 1960; Weis-Fogh, 1967). In Schistocerca gregaria, the haemolymph pressure amplitude near a ventral thoracic air sac was 100–150Pa generated by the flight motor and 300–500Pa generated by abdominal pumping (Weis Fogh, 1967). At 50–450Pa, the pressure pulse amplitudes in Manduca sexta are within the range of values for the locust, but they are produced only by the the flight motor with no measurable contribution from abdominal pumping. Abdominal ventilatory movements have been described in other hawkmoth species (Sphinx ligustri and Deilephila elpenor) during flight (Fraenkel, 1932a). It is likely that the effects of these movements are restricted to the abdomen, as in the giant silk moth Attacus atlas at rest, where they are correlated with CO₂ bursts recorded from the abdominal chamber (Wasserthal, 1996).

The present results in Manduca sexta reveal a new mechanism for the supply of oxygen to the flight muscles and extend the spectrum of respiratory air supply mechanisms discussed above. The flight motor itself, by increasing thoracic volume during the downstroke and the simultaneous automatic closure of the metathoracic spiracle in the subalar cleft, produces a retrograde airflow through the pterothorax, with CO₂ output occurring only at the posterior spiracles. The flight-motor-driven retrograde air flow is reminiscent of a turbo engine that sucks fresh air in at the anterior openings and releases the used air through the posterior openings. This efficient respiratory air supply may provide part of the explanation for the increase in O₂ levels in sphingid flight muscles during steady flight (Komai, 1998) and the physiological basis for their powerful and long-lasting flight characteristics and hovering ability.

I thank Dr Joachim Schachtner, Marburg, for supplying us with eggs of Manduca sexta and Witold Lapinsky for rearing the animals. I am grateful to Thomas Messingschlager for constructing the apparatus and to Alfred Schmiedl for optimizing the electronic equipment. I am indebted to Dr Stefan Hetz for a helpful hint concerning pressure sensor calibration. Dr Wiltrud Wasserthal contributed by reading and discussing the manuscript. Her help and the stimulating recommendations of an anynomous referee and the text Editor Alison Cooper are gratefully acknowledged. The electronic equipment was partly financed by a grant of Deutsche Forschungs-Gemeinschaft (Wa 258/4-1, 4-2).