The Nervous Control of Ventilation in Dragonfly Larvae

There has recently been a revival of interest in the co-ordination of arthropod movements from the point of view of the relative roles of the central nervous system and peripheral sense organs in producing patterned, rhythmic bursts of activity. Thus, in the flight system of the locust it has been demonstrated by Wilson & Gettrup (1963) that a general level of afferent input is required, while the role of the specific proprioceptors is to control the frequency of the wingbeat cycle. A great deal of work has been carried out on this system (Wilson, 1961, 1964; Wilson & Weis-Fogh, 1962; Weis-Fogh, 1964). As has been shown for the swimmeret movements of the crayfish (Hughes & Wiersma, 1960) there is evidence for the release of central patterns of activity by the excitation of descending command fibres from some of the higher centres. More recent work on this system has confirmed this mechanism (Wiersma & Ikeda, 1964) and has also shown that the abdominal cord of the crayfish can maintain persistent bursts of electrical activity in the segmental motor roots which innervate the swimmeret muscles when completely isolated (Ikeda & Wiersma, 1964). These bursts are closely similar to those present in the intact preparation.

Because of their rhythmic nature the control of the respiratory movements of animals by the nervous system is of interest from this point of view. Among the insects, however, there has been relatively little neurophysiological work. The classical work of Adrian (1930, 1931) on the electrical activity in isolated ganglia in a caterpillar and in the water beetle Dytiscus suggested that the nervous basis was essentially a central one which could be modified by peripheral inputs. Miller (1960a, b, c; 1962, 1964) has carried out some recent work on this problem in the locust and adult dragonfly. He too has emphasized the importance of central mechanisms in co-ordinating the ventilating movements of these insects. Huber (1960) has demonstrated that electrical stimulation of certain areas in the suboesophageal ganglion of the cricket will produce an increase in ventilation of the tracheal system, whereas stimulation of other areas will produce inhibition.

The dragonfly larva has recently been used in several physiological investigations (Hughes, 1953, 1958; Fielden, 1960, 1963a, Fielden & Hughes, 1962; Mill, 1963) and because of the previous work that has been done on its respiratory mechanisms it seemed likely to provide good preparations for the study of this problem. Many previous investigators have studied the effect of oxygen and carbon dioxide in the water on the frequency and pattern of ventilation and have postulated the presence of ‘centres’ within certain ganglia which are sensitive to changes in the tension of these two gases (e.g. Wallengren, 1914; Stahn, 1928).

As a preliminary to the investigations described below a study was made of the different patterns of ventilation found in these insects (Hughes & Mill, 1966) and this revealed a variety of muscular activities among different species and individuals and also in a given individual at different times. In general the division into three types made by Tonner (1936) was confirmed and was extended in several ways.

In the present work experiments were carried out using neurophysiological methods to determine the nervous basis of these movements, particularly in relation to the co-ordination of the main respiratory rhythm. The results have given information about the co-ordination of the intra- and intersegmental musculature and have also been of value in confirming the expiratory and inspiratory function of certain muscles.

In most experiments large larvae of Aeshna juncea were used, and in some experiments larvae of Anax imperator and Libellula sp., but only when this is stated specifically in the text.

The abdomen of the larva was dissected from the dorsal surface ; following a longitudinal median incision the terga were pinned out on each side, restraining the exoskeleton as little as possible so as to allow some movement to be effected by the respiratory musculature. The preparation was immersed in chilled saline (Fielden & Hughes, 1962). The gut and dorsal trachea were removed and also part of the ventral trachea. Care was taken to avoid damaging the two large transverse muscles at the anterior ends of segments 5 and 6. Some damage to the dorsal attachments of the former was always unavoidable. The longitudinal sternal muscles were normally removed from those segments from which it was desired to record. During dissection of the preparation it was washed several times with chilled saline to remove any digestive juices, etc., which may have been spilled.

Recordings were made principally from the second segmental nerve roots, which innervate the respiratory dorso-ventral muscles. A pair of fine platinum electrodes was used and the nerve was either left intact or severed distally. Nervous activity was amplified using a Tektronix 122 preamplifier and a Telequipment oscilloscope. Many of the recordings were stored on tape prior to taking photographic records. Platinum electrodes were also used to record extracellular potentials from the respiratory muscles and to stimulate various nerves. In the latter case a stimulus-isolation unit was used.

A few preparations were made of the isolated ventral nerve cord and recordings made from this in a moist chamber with wick electrodes. An integrator was occasionally used to give a picture of the total activity in these preparations.

Respiration involves pumping water in and out of the hind gut (which is modified to serve a respiratory function) through the anus and this is achieved by the action of both extrinsic and intrinsic muscles (Hughes & Mill, 1966).

The innervation of the abdominal muscles of the dragonfly larva associated with the body wall has already been described (Mill, 1965), but so far it has only been possible to ascribe a respiratory function to certain muscles on the basis of functional anatomy (e.g. Tonner, 1936; Mill, 1965). The present investigation has shown the extent to which these conjectures were justified and proof of the function of both expiratory and inspiratory muscles will be presented in subsequent sections.

In each abdominal segment from the fourth to the ninth the dorso-ventral muscle is divided into an anterior and a posterior part and a ‘respiratory’ dorso-ventral muscle. This latter lies between and internal to the anterior and posterior portions of the muscle and is the principal expiratory muscle of the extrinsic musculature. It is innervated by a branch from the corresponding second segmental nerve, which also innervates the anterior part of the dorso-ventral muscle, the posterior part being innervated by the third segmental nerve. When these respiratory dorso-ventral muscles contract the sterna are lifted and the volume of the abdomen is thus reduced, causing water to be expelled from the respiratory chambers.

At the anterior end both of the fifth and of the sixth abdominal segments there is a large transverse muscle; these are the diaphragm and the sub-intestinal transverse muscle respectively and are the inspiratory muscles of the extrinsic musculature. The diaphragm, first described by Amans (1881), forms a functional division between anterior and posterior parts of the abdomen (Hughes & Mill, 1966). It is innervated by a branch of the fifth abdominal median nerve on either side. The sub-intestinal transverse muscle is innervated on either side by a branch of the sixth abdominal median nerve. In addition it is thought to be innervated by a branch of the seventh abdominal median nerve on either side. This branch runs anteriorly from the anterior border of segment 7 across the pleural region of segment 6 and also innervates the lateral dilator muscles of the branchial chamber.

The remainder of the presumed respiratory musculature, namely the various dilator muscles of the branchial chamber, vestibule and valves, and the intrinsic musculature of these respiratory chambers, will not be dealt with in this paper, which is concerned primarily with body-wall movements during respiration.

There are several different types of respiration (Tonner, 1936; Hughes & Mill, 1966), and the remainder of this paper will be concerned with the ventilatory movements concerned in one of these, namely ‘normal respiration’.

In normal respiration expiration involves removing water from the branchial chamber via the vestibule and anus, and this is achieved by two processes working in phase. One of these concerns the contraction of the intrinsic musculature of the branchial chamber, causing an increase in pressure within it and decreasing its volume; while the other process involves a lifting of the sterna by the respiratory dorso-ventral muscles thereby enhancing the above effects.

In inspiration water is drawn into the branchial chamber via the anus and vestibule. This is achieved by two complementary processes, one due to the natural elasticity of the cuticle and the other muscular. The muscles involved are the various dilators of the branchial chamber, vestibule and valves and the two large transverse abdominal muscles (the diaphragm and the sub-intestinal transverse muscle). These transverse muscles, working in conjunction, together with the natural elasticity of the sclerites, restore the sterna to their extended position (Tonner, 1936).

The lifting of the sterna associated with expiration must involve the segmental dorso-ventral muscles of the abdomen, which lie on either side of the mid-line with their origins on the terga and their insertions on the ventral edge of the pleurites. Indeed, in preparations made as described above the respiratory dorso-ventral muscles can be observed to contract and relax rhythmically at the same frequency as movements of the respiratory chambers. There is a single pair of these muscles in each segment from the fourth to the ninth and each is innervated by a branch from the corresponding second lateral nerve. The activity within these nerves and its relation with the expiratory muscles will now be considered.

Rhythmic bursts of impulses occur in the second segmental nerve roots of abdominal ganglia 5–8 at least (Pl. 1, fig. 1a). These are efferent in nature as can be demonstrated by sectioning the nerve either distal to the electrodes, when the bursts are still recorded, or proximal to them, when the bursts cease. The time between the start of consecutive bursts varies between preparations but is normally within the range of 1·5−2·5 sec., which corresponds to a frequency of 24–40 bursts/min., and thus compares favourably with the usual range of respiratory frequency observed in the intact animal (23–48 bursts/min.) by Hughes & Mill (1966). However, the interval is fairly constant in any one preparation at a given time. The individual burst is composed of a single unit which typically starts to fire at a fairly low frequency (in the range of 10-20 impulses per second), gradually increases to a plateau with a maximum of up to 100 impulses/sec., and then falls off rapidly at the end of the burst, which has a sharp cut-off (Text-fig. 1). The rising phase can be divided into two stages: an initial fast rise in frequency, followed by a more slowly rising phase. The duration of the burst and the number of times the expiratory neurone fires within the burst also vary between preparations (Text-fig. 2), but again are relatively constant in consecutive bursts in a given preparation (Text-fig. 1). The duration can range from about 0·2 sec. up to almost 1 sec., while the number of impulses per burst is typically between 7 and 30. Factors affecting these two variables will be dealt with later.

The respiratory dorso-ventral muscle was observed to contract simultaneously with each burst in the second segmental nerve, and simultaneous recording from the muscle and nerve shows that a 1:1 relationship exists between the nerve impulses and the muscle action potentials (Text-fig. 3). In addition, as the frequency of the nerve impulses in the burst increases, facilitation of the muscle action potentials occurs. Evidently this unit involves an innervation of the slow type. Whether or not this muscle has a fast-fibre innervation has not yet been determined. The use of a triggered time-base (Text-fig. 3b) shows the 1:1 relationship very clearly and also the presence of a very small nerve potential preceding the development of each muscle action potential. In this preparation there is a constant delay of about 6 msec. Electrical recordings have shown the presence of other motor fibres in this nerve which, however, also innervates the anterior part of the dorso-ventral muscle as well as partially innervating two other segmental muscles (Mill, 1965). As far as respiration is concerned activity in only one fibre has been recorded associated with muscle action potentials in the respiratory dorso-ventral muscle. With the nerve cut distal to the recording electrodes bursts of impulses still occur but the muscle, of course, ceases to contract.

The presence of only a single motor unit in the expiratory burst offers the opportunity of investigating the pattern of the burst. Analyses of the expiratory bursts were made by measuring the pulse interval of the unit. The variables concerned are the duration of the burst, the number of impulses in the burst and the change in frequency of the unit during the burst; also the frequency of repetition of the burst, i.e. the respiratory frequency. The basic form of the expiratory burst is described above and in Text-fig. 1.

The minimum pulse interval normally attained by the unit is relatively constant irrespective of the duration of the burst (Text-fig. 2; Table 1), and it is the rate at which this minimum value is achieved that is the main factor in determining the duration. Thus the minimum pulse interval attained by the unit in the four bursts shown in Text-fig. 2 (all from different preparations of the second root of the seventh ganglion) is ▴ = 10, ○ = 13, • = 14 and ▄ = 14 msec., with burst durations of, 200, 230, 579 and 879 msec, respectively. Presumably this ultimate frequency is that required to enable the respiratory dorso-ventral muscle to overcome the natural elasticity of the cuticle and also the water pressure in the respiratory chambers.

Neither the minimum pulse interval nor the burst duration nor the number of impulses per burst are related to respiratory frequency, at least over the normal range of the latter (Table 1). However, the burst duration (d) and the number of impulses per burst (n) both vary considerably and the one is directly proportional to the other. The expression d/n, that is the mean pulse interval within the expiratory burst, in a given segment rises steadily with increase in duration of the burst, again over the range of normal respiratory frequencies. The approximate relationship between these two variables in the seventh segment can be expressed in the form d = (n + 2)².

Unusually large values of d/n may occur under certain circumstances and the values for one such preparation, which has a very low respiratory frequency, are presented in Table 1. This will be discussed later.

Occasionally bursts of longer duration occur, but these do not appear to affect the regular rhythm (Pl. 1, fig. 1b). As it has been shown that ‘gulping’ respiration appears to be independent of the rhythm of normal respiration, although the respiratory dorso-ventral muscles are involved in both (Hughes & Mill, 1966), it is suggested that this longer type of burst may be associated in some way with gulping movements. However, the latter are normally of much longer duration than the burst shown.

In conjunction with the evidence of the movements of the sterna during normal expiration and inspiration (Hughes & Mill, 1966) it can safely be inferred that the rhythmic bursts of motor impulses observed in the second lateral nerve roots of the posterior abdominal ganglia are indeed expiratory in nature, and that the respiratory dorso-ventral muscles are expiratory muscles.

The contraction of the respiratory dorso-ventral muscles on either side of the abdomen is simultaneous in any one segment and this is reflected in the expiratory bursts in the second segmental nerve roots (Pl. 1, fig. 1 b–d). Not only do these bursts occur simultaneously on opposite sides of a given ganglion, but their pattern is very similar both in the frequency of the individual units and in the overall form of the bursts (Text-fig. 4). The timing of the onset and end of the discharge is similar on the two sides and the individual action potentials normally show a very close relationship, which may or may not be 1:1 (Pl. 1, fig. b, c). The onset of the first spike of the bursts on one side may precede that on the other by up to about 35 msec.

In the crayfish a similar situation prevails in the first roots of those abdominal segments which bear swimmerets (Ikeda & Wiersma, 1964). Here the bursts on opposite sides of the ganglion start and stop at approximately the same time and the number of impulses in them is similar. Occasionally one or more bursts fail to appear on one side and this may occur naturally (Ikeda & Wiersma, 1964), or be brought about by stimulation of a specific interneurone in the ventral nerve cord (Wiersma & Ikeda, 1964). In both cases when the bursts reappear they are still in phase with those on the opposite side. This phenomenon has not been observed in the dragonfly larva where, in a normal preparation, the occurrence of an expiratory burst on one side is always accompanied by a similar burst on the contralateral side. This difference between the two preparations may be an expression of the different motor functions which these rhythmic bursts effect. Thus in the crayfish failure of the swimmerets to beat on one side has little effect on the mechanism of those beating on the contralateral side; whereas in the dragonfly larva failure of the dorso-ventral muscles to contract on one side may lead to an abnormal ventilatory movement.

The pattern of the individual expiratory burst is essentially the same in each second root from at least segment 5 to segment 8. Also, as one would expect, the frequency of the bursts is the same. However, in a given preparation the expiratory bursts in second roots of more anterior ganglia tend to have fewer impulses per burst and the bursts themselves tend to be of shorter duration than in more posterior second roots (Table 2). This difference also occurs to some extent between preparations but is somewhat masked by individual variations (Text-fig. 2). Furthermore, the expiratory burst in one segment starts after the burst in the segment next posterior to it. Thus the start of the respiratory burst in the fifth segment follows that in the sixth, and so on (Pl. 2, a−d). A similar phenomenon occurs in the rhythmic motor discharges to the swimmerets in different segments of the crayfish abdomen (Ikeda & Wiersma, 1964). Lack of synchrony of the expiratory bursts between segments has not been observed.

Several examples illustrating the intersegmental delay are presented in Table 2, from which it can be seen that the delay between segments 7 and 6 and between segments 6 and 5 is in the order of 50 msec., while that between segments 8 and 7 is rather longer, provided that d/n remains within the normal range. Two preparations are shown in which d/n is rather high and both of these show a delay of several times the normal length (see also Text-fig. 5,a, b). The range of delay in a single preparation is not great.

The units in the expiratory bursts from consecutive segments, although commencing at different times, i.e. posterior before anterior, reach a minimum pulse interval at approximately the same time as one another and may attain the same minimum pulse interval. Thus, in Text-fig. 5, 6, the minimum pulse interval in segments 5, 6 and 7, is 14, 11 and 11 msec, respectively. In addition the bursts end at approximately the same time.

Although slight variations in the mechanical effect in different segments and in different preparations may seem indicated by the above results it may, nevertheless, be less variable than this if there are compensatory differences in the neuromuscular characteristics.

If a first segmental root is stimulated two events may occur in the ipsilateral second segmental root of the same or of other ganglia. First there is a very rapid response, involving at least one large motor neurone in the second roots. Second, if the stimulus occurs in the interval between two successive expiratory bursts a burst is elicited which is similar in form to a normal expiratory burst, and successively more anterior ganglia fire after one another in the same way as in normal respiration. The next expiratory burst now commences at a time after the elicited burst determined by the normal period between bursts in the preparation. Thus the elicited burst completely re-sets the respiratory rhythm, but does not affect the frequency of subsequent bursts (Text-fig. 6; Pl. 2, b−e).

If the stimulus occurs during the course of an expiratory burst only the first response is observed, the burst during which it occurs not being markedly affected (Pl. 2, c).

Slight differences in the elicited burst are apparent. Thus, in the experiment illustrated in Text-fig. 6 and Pl. 2 the burst in the second root of the seventh abdominal segment (the ipsilateral first root of the seventh segment being the one stimulated) is somewhat different in appearance from a normal burst initially, although this is probably due in part to masking of the expiratory unit by the large unit(s), but is of similar duration, and the expiratory unit has a closely similar time-course as regards its increase and subsequent decrease in frequency and in the minimum pulse interval attained. The difference is rather more marked in the recording from the ipsilateral second root of the fifth abdominal segment, where the expiratory neurone starts to fire rather earlier with respect to that in the seventh segment than is normal. However, other features of the burst remain similar.

The first root innervates all tactile sensilla on the tergum and pleuron as well as both dorsal stretch receptors and the oblique receptor. In the intact animal stimulation (e.g. tactile) which would cause afferent activity in the first segmental nerve causes a very rapid flexion of the abdomen towards the stimulated side and it is thought that this flexion, presumably brought about by the unilateral contraction of the dorso-ventral oblique muscles (which are also innervated by a branch from the second segmental nerves) is correlated with the observed rapid response in the second roots.

It was suggested by Tonner (1936) that the two transverse abdominal muscles, acting in conjunction, could assist in inspiration by depressing the sterna. These muscles, as mentioned previously, are innervated by branches of the median nerves and differ in this respect from the principal expiratory muscles, which are innervated by the paired segmental nerves.

The response in the sub-intestinal transverse muscle was studied either by recording from it directly or by recording from the branch of the seventh median abdominal nerve thought to innervate this muscle. As in the recordings from the expiratory neurones rhythmic bursts of impulses occurred. Each burst caused a contraction of the sub-intestinal transverse muscle and the contractions of this muscle alternated with those of the respiratory dorso-ventral muscles. These rhythmic bursts resembled the expiratory bursts in that only a single unit was present in each burst, but differed from them in their rather longer duration and in the absence of the regular patterning of the unit within the burst, characteristic of the expiratory bursts. The frequency of the unit is rather irregular throughout the burst. However, this irregularity can be smoothed out somewhat by plotting the mean pulse interval of each three consecutive spikes (Text-fig. 7). The resulting graph indicates a slight overall increase in frequency during the first half of the burst and a fairly constant overall frequency during the second half. The minimum pulse interval attained is not normally as low as that in the expiratory bursts.

Recordings of the above response in the sub-intestinal transverse muscle simultaneously with that in the second segmental nerve roots showed that the two types of burst alternated with one another and that there was no overlap between them. (Text-fig. 7; Pl. 1, fig. 2). From this it was concluded that contractions of the sub-intestinal transverse muscle are inspiratory in nature. It was also observed that there was a very close coupling between expiration and inspiration rather than the reverse, an inspiratory burst commencing only a short interval (of the order of 100–150 msec.) after the end of an expiratory burst. The pause between the end of the inspiratory burst and the start of the next expiratory burst was longer (more than twice as long in the example shown in Text-fig. 7). The duration of this interval would depend on the respiratory frequency, since the extended position of the sterna (that is, the post-inspiratory position) is the normal position of rest between ventilatory movements.

Notice the presence of two smaller units in the second root in the preparation shown in Pl. 1, fig. 2, which only start to fire at the end of, or following, the expiratory burst activity. They are in fact approximately covering the period when the activity of the inspiratory neurone is being recorded.

Recordings taken from the two ends of the sub-intestinal transverse muscle show a 1 : 1 relationship between the muscle action potentials (Text-fig. 8), which is in agreement with Zawarzin’s (1924) observation that the motor axons in the median nerve divide into two, one branch of each passing to either side.

It was observed that the diaphragm and sub-intestinal transverse muscle contract together and indeed, repeating the procedure outlined in the previous section (in this case with one pair of electrodes on the diaphragm and the other recording the expiratory bursts), it was found that the diaphragm contracted rhythmically and alternated with the contraction of the respiratory dorso-ventral muscles (Text-fig. 9), in the same way as did the sub-intestinal transverse muscle. Again there is a very close coupling between expiration and the diaphragm contractions, and recording simultaneously from the diaphragm and the sub-intestinal transverse muscle showed that the activity in these two muscles was virtually identical as regards its onset and duration. However, the presence of only a single unit in the recordings from the diaphragm is not quite as certain. It was concluded from the above that the diaphragm is also an inspiratory muscle.

Recording from the isolated nerve cords of a caterpillar and of Dytiscus, Adrian (1930, 1931) demonstrated the presence of an asynchronous discharge of nerve impulses soon after isolation. After a period of time the overall level of activity fell, revealing some regularly firing units and periodic bursts containing many units. The frequency of the bursts was between 5 and 15 per minute, which corresponds with the frequency of the respiratory rhythms of the intact insect. Similar phenomena have since been reported in crayfish (Prosser, 1934) and in Periplaneta (Roeder & Roeder, 1939). In Dytiscus the interval between bursts varies somewhat and the pattern within the bursts is not the same as when the sensory supply is intact (Hughes, 1949). Furthermore, the passage of carbon dioxide over the isolated preparation gave rise to a considerable increase of activity (Hughes & Shelton, unpublished).

In dragonfly larvae, preparations of the isolated ventral nerve cord showed a large number of mostly irregular units to be active. However, spontaneous rhythmic bursts of impulses, probably containing several units, were obtained in a libellulid larva. These bursts became particularly prominent some time after the preparation had been made and are shown in Text-fig. 10b about 1 hr. after isolating the ventral nerve cord. This is presumably at least in part due to the cessation of activity of other units, or to their recruitment in the bursts. Text-fig. 10c shows the integrated activity in the same preparation.

In a few preparations made of the isolated ventral nerve cord of Aeshna and Anax burst activity was not found to be normally present, but could be induced by the passage of carbon dioxide over the preparation (Text-fig. 10a) as in the case of Melanophis (Roeder, 1953). Carbon dioxide also had the more general effect of increasing all activity in the preparation. Occasionally single units could be detected in these preparations.

The frequency of the bursts in the libellulid preparation is about 40/min., which is at the lower limit of the observed range of respiratory frequency for this insect (Hughes & Mill, 1966). This also applies in the case of the preparation shown of Aeshna, which has a maximum frequency of only about 16 bursts/min. The corresponding lower limit is 12/min.

On the basis of these results certain distinct differences clearly exist between the expiratory and inspiratory muscles of the body wall and their innervation. The expiratory muscles are paired segmental muscles which have a very rich tracheal supply. On the other hand there are only two principal inspiratory muscles, both of which extend transversely across the abdomen from one side to the other, and they have an apparently normal supply of trachea. The expiratory muscles receive their innervation via the paired segmental nerves; whereas the inspiratory muscles are innervated by median nerves.

The difference in pattern between the highly organized expiratory and the rather irregular inspiratory bursts is of particular interest. The increase in frequency during the expiratory burst may be necessary to overcome the increasing restoring force due to elasticity of the cuticle which must occur as the sterna are raised and the terga become more convex. This would not be necessary in inspiration as the muscles concerned are assisted by this restoring force, and indeed only a slight overall increase in frequency occurs during the inspiratory burst and this only during the first half of the burst. Furthermore, the overall frequency is not so high as is reached in the expiratory burst.

The individual variation which exists both in the duration of the expiratory burst and in the number of impulses per burst may be due to a variety of factors, such as temperature and carbon-dioxide tension, which alter the physiological condition of the animal. The elasticity of the cuticle may also be of importance. Thus expiratory bursts may tend to be shorter in newly moulted larvae, and hence have fewer impulses in them, since a lower restoring force (of the cuticle) has to be overcome.

Respiratory frequency does not appear to be directly correlated with the ratio of the duration of the burst to the number of impulses contained within it (d/n) over the normal range experienced, but this ratio assumes an unusually large value under certain conditions, for example at unusually low respiratory frequencies. Thus the anomalous burst presented in Table i, which has a value for d/n of 55-8 at a respiratory frequency of 12-8 bursts/min. (12/min. is the observed minimum—Hughes & Mill, 1966), although it may be explained by the poor physical condition of the preparation, did behave consistently over a reasonable period of time, and this change in pattern may normally occur at low respiratory frequencies. Low respiratory rates may be associated with a small amplitude of ventilation, and this is also indicated by the high value of ‘minimum pulse interval’ (36 msec.). In other cases it is not so easy to find an explanation for the high values of d/n. Some, at least, are possibly due to deterioration of the preparation, which does tend to impose abnormal conditions on the respiratory rhythm. Another feature attendant on high values of d/n is the much slower propagation of the respiratory bursts from behind forwards (Table 2).

As the duration of one cycle of ventilation rarely exceeds 1·4 sec. and is normally less than I-I sec. (Hughes & Mill, 1966) it follows that the duration of individual respiratory bursts would not be expected to be curtailed below a frequency of respiration of around 43 or, normally, 55 bursts/sec. These values are close to the upper limits of the usual (48) and observed (57) frequency ranges obtained for normal respiration.

Text-fig. 11 illustrates a possible arrangment of the respiratory pathways in the abdomen and their control. The start of the respiratory cycle, in terms of the output from the central nervous system, is the expiratory burst which, it may be presumed, has been triggered in some way by a command from some higher centre. In the crayfish (Procambarus clarkii) rhythmic motor discharges to the swimmerets are present in the first segmental roots of the relevant abdominal ganglia (Hughes & Wiersma, 1960; Ikeda & Wiersma, 1964). These discharges, unlike the respiratory bursts in the dragonfly larva, contain more than one unit and the activities of these individual units overlap with one another. However, as inspiration is coupled to expiration in the dragonfly larva so is protraction coupled to retraction in the crayfish swimmeret, and it is thus the units causing retraction which start the cycle and which, therefore, are presumably directly responsive to the triggering mechanism. Hughes & Wiersma (1960) and Wiersma & Ikeda (1964) have demonstrated the existence of command intemeurones at various levels of the nerve cord, some of which, when stimulated, caused bilateral, rhythmic swimmeret movements, while stimulation of other inter-neurones bilaterally inhibited the rhythmic motor discharges to the swimmerets.

The degree of peripheral control of motor functions appears to vary in different systems. Thus in the wingbeat of grasshoppers only a general level of sensory excitability needs to be present to maintain the co-ordinated cycle of wing movements, the specific inputs from the associated proprioceptors only controlling the frequency (Wilson & Gettrup, 1963). However, in the case of the crayfish, rhythmic co-ordinated motor discharges still occur even under conditions of complete deafferentation accompanied by isolation from the rest of the central nervous system. Anything approaching this degree of isolation in the dragonfly larva disrupts the pattern of the respiratory bursts, but further experiments may show that peripheral control is of less importance than at present seems to be the case. The only proprioceptors affected during respiration are the so-called oblique receptors (Finlayson & Lowenstein, 1958; Mill, 1965).

Stimulation of a first segmental nerve can reset the rhythm of respiration as described above. This nerve contains afferent fibres from two dorsal stretch receptors and from the spines and hairs of the tergum and pleuron as well as from the oblique receptor. Miller (1960a; 1962) has demonstrated the presence of rhythmically-firing ventilation centres in the abdomen of the locust and adult dragonfly which are influenced by higher centres. In the former at least, frequency and amplitude of ventilation are not under the control of receptors outside the central nervous system.

A positive feedback system would presumably be necessary to cause an increase in the frequency within the expiratory burst and either this or the expiratory centre itself must ultimately be limited by some high-threshold system requiring temporal summation. A limiting system of some sort must also be required for the inspiratory burst and this may conceivably be provided by feedback from, for example, the appropriate oblique receptor, which will be stretched during inspiration or, alternatively, it may be part of the central network.

One of the authors (P. J.M.) wishes to thank Prof. C. F. A. Pantin and members of his department most sincerely for the hospitality shown to him during his stay in Cambridge ; also the Department of Scientific and Industrial Research and the North Atlantic Treaty Organization for the Fellowship during the tenure of which this work was carried out.