The Derivation of the Motor Command to the Spiracles of the Locust

Among insects several patterns of motor activity have been studied in which the sensory input bears no simple relation to the output. Of these most is known of the locust flight system as a result of the penetrating studies of Wilson (1961, 1964) and Wilson & Weis-Fogh (1962). Wilson & Wyman (1965) applied electrical shocks with random timing to the locust central nerve cord and produced patterned oscillations in groups of motoneurones which supply the flight muscles, similar but at lower frequency to those which occur naturally in flight. The ventilatory mechanisms of some insects show similarities with the locust flight system in so far as maintained stimulation, either electrical or by the application of gases, can produce patterned oscillations in neurones which bring about pumping movements (Huber, 1960; Mill & Hughes, 1966; Miller, 1966). Unlike the flight system, however, such activity may also persist in the absence of input. The spiracle motoneurones too show persistent activity in the absence of input so that in the intact insect they need to be periodically switched off by other cells to allow the spiracle to open. In common with many cells in the arthropod cord their ‘resting’ state is apparently one of continuing activity. They are comparable with the spontaneous units of unknown function which have been studied in the crayfish cord (Preston & Kennedy, 1962; Kennedy & Preston, 1963; Biederman, 1964; Takeda & Kennedy, 1965; Biederman-Thorson, 1966), and some of their properties are also comparable with those of the neurones of the crustacean cardiac ganglion (summary by Hagiwara, 1961).

In the present study an attempt has been made to distinguish between the contributions of a pacemaker, which is an integral part of the motoneurone, and of synaptic input to the motor patterns which emerge under different conditions in the intact insect. Spiracle nerves contain only a small number of axons and it is normally feasible to distinguish their separate activity in extracellular records from the whole nerve. This, together with the fact that functionally related components of the system are distributed in different ganglia, makes the preparation a suitable one for a study of this nature. Moreover, the number of different commands issued is limited since, unlike most motor systems, the spiracle muscles are engaged in only one form of activity, namely, opening or closing the valves. Some attempt at an interpretation of the results in terms of synaptic properties and the organization of the axon branches in the neuropile has been made, but confirmation must await direct measurements from this region. Recent studies of the electrical activity in the neuropile of insects (Iwasaki & Wilson, 1966; Roeder, 1966) suggest that this approach will soon bear fruit.

In normal ventilation the first (prothoracic) spiracle of the locust, Schistocerca gregaria, opens for a short period during the inspiratory stroke of the abdomen and is closed for the remainder of the cycle. In co-operation with other spiracles this activity permits the abdominal pump to produce a uni-directional ventilating airstream which passes posteriorly through the insect. The valve movements are brought about by two muscles, an opener and a closer. The closer receives two motor axons from the median dorsal nerve of the prothoracic ganglion (GI), while the opener is supplied by three motor axons; two arise in GI and accompany the nerves to the closer, while the third arises in the mesothoracic ganglion (GII) and proceeds to the opener in 11N ₁ and then joins the median nerve close to the spiracle (Fig. 1) (Miller, 1960a, 1965). In spiracle 2 of the same species, Hoyle (1959) has shown that all muscle fibres are doubly innervated by a ‘slow’ and a ‘fast’ axon, whose effects on the muscle differ only by a small amount. Unlike many motor systems both contribute at roughly equal frequencies to normal activity, and their differing slow and fast properties may not be important.

Schistocerca gregaria Forskâl and Locusta migratoria migratorioides (Reiche and Fairme), both kindly supplied by the Anti-Locust Research Centre, were used in these studies. Most experiments were performed on teneral adults of Schistocerca but mature insects have also been examined. The results are derived from a total of 84 preparations. The temperature was normally 20-22° C.

Records were taken from the transverse nerves to the spiracles by conventional methods which have already been described (Miller, 1960a). Electrical stimulation of nerves was carried out through fine platinum wire electrodes via an isolation unit, using square-wave pulses of 0·2 msec, duration at various frequencies and intensities. Stimulation of parts of a connective between GI and GII was carried out after it had been desheathed and split into a number of small bundles, according to the method recommended by Fielden & Hughes (1962). One connective was cut through completely, while the other was usually split into six to twelve bundles. After removal of the tough perineurium there was often a temporary conduction block, probably due to unavoidable stretching of the fibres, but recovery usually followed in less than 1 min. as indicated by the re-establishment of synchrony between spiracle 1 and abdominal pumping. Recordings taken from split bundles showed that the activity of recognizable units persisted for at least 1 hr. when the preparation was bathed in Hoyle’s locust saline. Split bundles were lifted into air on the stimulating electrodes and surrounded by a small amount of petroleum jelly to prevent desiccation and minimize stimulus spread. The recording electrodes on the spiracle nerves were similarly treated and then reimmersed in saline (Fig. 1). No deleterious effects were observed as a result of the use of petroleum jelly (cf. Narahashi, 1963).

As Fielden & Hughes (1962) have pointed out, a disadvantage of this technique is that in successive preparations there is no means of identifying the same bundles of fibres except by their activity. A variety of effects is to be expected therefore when different bundles are stimulated electrically, according to their component fibres. On the termination of an experiment the bundle was squeezed between forceps on the ganglion side, subsequent shocks then being found to be without effect on the spiracle nerves. This indicated that the stimulated area was normally close to the electrodes, although with very high intensities of current it was sometimes possible to stimulate the motoneurones directly. In all preparations care was taken to minimize damage to the tracheae which supply the ganglia.

Observations on the natural patterns of activity. A brief description of the natural patterns of activity has already been given (Miller, 1960a, 1965): only activity in the two motor units to the closer muscle of spiracle 1 will be described here although similar patterns are distributed to spiracle 2. In favourable preparations their separate activity can be distinguished by the different spike heights, or one can be examined alone after damage of the axon of the other proximal to the electrodes. Three main patterns of activity can be distinguished (Figs. 2, 3).

Pattern 1 consists of high-frequency bursts of impulses occurring at the same frequency in both axons, between 100 and 250/sec./axon, and lasting about 100 msec. The two units fire either synchronously or alternately and the burst normally initiates contraction of the muscle and closure of the spiracle. The pattern comprises smooth regular firing, sometimes at a slowly declining frequency, and is usually quite distinct from pattern 2 which follows. Such bursts may occur synchronously in spiracles 1 and 2, but can occur in either alone ; in the former case there is no synchronization of individual spikes in the two spiracles (Fig. 2c, d).

Pattern 2 normally follows 1 although it may be absent during fast ventilation. When pumping is slow or stops temporarily, pattern 2 continues in the spiracle nerves and maintains a tetanus in the closer muscle until the next inspiratory stroke starts. Again both axons participate, firing at approximately equal frequencies (20-50 impulses/sec./axon). Pattern 2 may comprise either a, regular trains of impulses, often in doublets (Figs. 2,a, b ; 3B, C), synchronized in the two units ; or b, less regular firing in which sometimes components of a or c can be recognized; or c, regular evenly spaced but independent trains of impulses in the two axons, usually at different frequencies so that a slower beat-frequency appears (Figs. 2 e,f, g; 3D). Pattern 2c is identical with free running which always appears after removal of the metathoracic ganglion (GUI), or section of the connectives between GI and GII. Pattern 2 c is more commonly seen in intact Locusta than in Schistocerca, although in the latter 2a is sometimes replaced by 2 c at faster frequencies of ventilation. Different individuals, however, tend to remain true to their firing plan; for example, one Schistocerca showed pattern 2a before and after decapitation, removal of the abdomen and section of lateral thoracic nerve s, even though ventilation frequency was altered considerably.

When pattern 2 a appears there is often approximate 1:1 synchronization of im-pulses between spiracles 1 and 2 (Figs. 46, c and 5). This has not been observed during 2b or c, or under any conditions after removal of GUI. Since at high frequencies patterns 2 a, b and c all produce a maintained tetanus of the closer muscle it may be functionally unimportant which firing plan is followed. However, tension measurements of the muscle contractions have not been made and close-paired impulses such as occur in 2 a are well known in other situations to cause tensions higher than are produced by the same frequency evenly spaced (Wilson & Davies, 1965); Hoyle (1959) has shown this to be the case in spiracle 2 of the locust.

In summary, therefore, pattern 2 comprises activity which is either largely synchronized (impulse for impulse) in both axons and in adjacent segments, or which is independent in each axon, or again which may be irregular and possibly a mixture of the first two.

Pattern 3 comprises short periods of no activity in the nerves when the muscle relaxes and the spiracle opens coincident with the inspiratory stroke. During weak ventilation, activity in both axons of spiracle 2 may continue at a low frequency, but this does not normally occur in the spiracle 1 nerve. As a result spiracle 2 may remain more or less closed throughout the cycle and inspiration takes place only through spiracle 1, spiracles 3 and 4 also remaining closed. At such times spiracle 10 acts as the only expiratory spiracle and gaseous exchange is achieved with a minimum of spiracle opening. Similarly in resting dragonflies spiracles 1 and 10 alone are operational.

In order to produce effective uni-directional ventilation, spiracle movements must be precisely synchronized with the pumping strokes and these strokes must take place in all parts of the abdomen nearly simultaneously. Activity is co-ordinated in the locust by GIII which lies one-tenth of the distance between spiracles 1 and 10. To bring about synchronized movements it must therefore excite remote regions through fast-conducting pathways with minimum delays or operate through a system with graded delays. Co-ordination in the abdominal cord is believed to be brought about by fast-conducting interneurones (Miller, 1966) similar to those which occur in the cricket (Huber, 1960), and a further set has been postulated which runs anteriorly to harness the activity of the thoracic spiracles to the pumping rhythm (Miller, 1965).

Some measurements have been made of the delays between the occurrence of events in the ventilatory cycle in different parts of the insect. For example, bursts of impulses in lateral nerves from GUI to expiratory muscles commence up to 10msec, before corresponding high-frequency bursts in the spiracle 1 nerve. Comparable bursts leave the last abdominal ganglion in nerves to expiratory muscles about 60 msec, after the start of activity in the spiracle 1 nerve, while a similar delay intervenes between the end of closer activity in the spiracle 1 nerve and the start of the high-frequency burst in the nerve to spiracle 10 (Fig. 4a). In dragonfly larvae Mill & Hughes (1966) found a relatively long delay, over 50 msec, between the expiratory bursts in adjacent abdominal segments, and exceptionally a delay of as much as 379 msec, between segments 5 and 7. Ventilation would therefore presumably have more of a peristaltic nature than in locusts.

The onset of high-frequency bursts in the spiracle 2 nerve precedes that of corresponding bursts in the spiracle 1 nerve by 5-7 msec. When impulse synchronization occurs between these spiracles (pattern 2a), spiracle 2 normally leads by about 5 msec., although considerable variation is seen (Figs. 4, 5). Such synchronization does not involve all the spikes ; for example, in one preparation a train in the spiracle 1 nerve comprised a total of 246 spikes (both axons) of which 61 % were within 5 msec, of a corresponding spike in the spiracle 2 nerve. The frequency was 25/sec./axon, and since most spikes occurred synchronously in the two axons of spiracle 1 (cf. Fig. 3 B, C), the impulse intervals for the whole nerve were about o and 40 msec. Correspondingly in the spiracle 2 nerve, firing at 2o/sec./axon, 68 % of the spikes were synchronized with spiracle 1 spikes, and many of the remainder were only a little more than 5 msec, from a spiracle 1 spike. In contrast, when pattern 2 c occurs, less than 3 % of the spikes are precisely synchronized in the two axons of spiracle 1 (cf. Fig. 3D) and a similarly small proportion is synchronized between spiracles. Pattern 2 c is thus characterized by a lack of intra- and intersegmental coupling while these are prominent in pattern 2 a.

Free running occurs indefinitely in the motoneurones to the closer muscles of spiracles 1 and 2 after section of the connectives between GII and GUI, and, as in pattern 2c, there is no interaction or coupling between the units. Section of one GI-II connective and the contralateral GII-III connective produces continuous free running in spiracle 1 while spiracle 2 remains synchronized with ventilation; the command interneurones therefore run through GII from GUI to G I. In a preparation in which GI is connected only with the head, stimulation of the head, mechanically or by light, accelerates free running which then slowly returns to its original frequency. Comparable stimulation of GI does not have this effect. If the head becomes hypoxic, the frequency of free running declines and it may cease. Since spiracle 1 supplies the head directly with tracheae, control of its activity by the head is to be expected. After decapitation free running becomes insensitive to mechanical stimulation applied at any point.

Electrical stimulation has been applied to the spiracle nerves antidromically in an attempt to locate the pacemakers responsible for free running. Stimulation was applied to one transverse nerve while records were taken from the contralateral nerve. This method takes advantage of the fact that the median nerve and its contained motor axons divide into left and right transverse nerves before running to the spiracles. Single shocks evoked responses (S spikes) in all four motor axons (two supplying the closer and two the opener) which could be distinguished readily by their different heights and conduction rates (Miller, 1960 a), and these were recorded together with natural impulses produced by the pacemaker (P spikes). When S and P spikes collided in the transverse nerve, no S spikes were recorded, but when collision took place in the median nerve no P spikes were recorded. Otherwise it was found that S spikes were always followed by at least the normal spike interval before the next P spikes appeared, regardless of the position of the S spikes in the natural P cycle. Thus both pacemaker cycles could be re-set by antidromic spikes which must therefore be able to invade the pacemaker loci (Figs. 6, 7). It can be concluded that each motoneurone contains an independent pacemaker, perhaps with impulse production and recovery cycle taking place at the same locus, and there is no need to postulate an antecedent pacemaking cell. The alternative theory of a closed loop of cells which has been discussed by Kennedy & Preston (1963) can be discounted since there is no evidence for the necessary long refractory periods. Kennedy & Preston reached the same conclusion for the cells they were studying in crayfish ganglia. Biederman-Thorson (1966) has shown that many of the spontaneously active cells in crustacean ganglia derive their activity from intrinsic pacemakers, even when their firing pattern is not markedly regular, such cells having ‘noisy’ pacemakers.

In some records the interval between the S spike and the subsequent P spike in the spiracle motor nerve was slightly longer than the normal pacemaking interval. This was most marked when the S spike followed closely after the preceding P spike and some examples are shown in Fig. 7. The juxtaposition of P and S spikes may produce a prolonged post-excitatory depression, or the delay may represent the activity of a negative feed-back mechanism which controls pacemaker frequency (cf. Bullock & Horridge, 1965, p. 314). However, the possible effects of stimulus spread cannot be excluded.

Normal pacemaker activity can be studied most readily in preparations in which one axon has been damaged proximal to the electrodes and the activity of the other alone is recorded. A histogram of the activity of one pacemaker recorded in this way is shown in Fig. 8, where it can be seen that the mean interval is 38 msec, with extreme values at 28 and 46 msec.

Antidromic stimulation of the transverse nerve has been applied to preparations with an intact nerve cord in an attempt to determine the contribution of the motoneurone pacemakers to normal activity. When Locusta ventilates rapidly pattern 2b may be distinguished, but if ventilation is slowed by perfusion with 50% oxygen, or by decapitation, pattern 2b is often replaced by 2c; that is, firing becomes more regular. During periods of 2 c firing, antidromic stimulation re-sets the cycle as in free running. Again the close proximity of P and subsequent S spikes sometimes results in a lengthened interval before the next P spike (Fig. 9 c and d). Pattern 2 c, therefore, like free running, seems to be solely due to the activity of the pacemakers which act independently of each other and of those in GII.

When pattern 2 a appears, however, antidromic stimulation produces S spikes which are followed by natural spikes after a more variable interval. Moreover simultaneous records from spiracles 1 and 2, stimulation being applied to spiracle 1 alone, show that such natural spikes are often matched by spikes in the spiracle 2 nerve (Fig. 9a, b). The spiracle 2 records act as a control for those from spiracle 1 and they suggest that S spikes are interpolated into the firing train without affecting the timing of subsequent spikes. Although spiracle 1 may be firing couplets at regular intervals, the timing at least of the first spike is not affected by the S spikes, and it can be concluded that during such firing the oscillation with the major period is determined by synaptic input and that the pacemakers are probably inactive.

In conclusion, therefore, pattern 2 c, identical with the free running which occurs in preparations with G III removed, is derived from motoneurone pacemakers, while pattern 2 a, which is often accompanied by impulse synchronization between spiracles 1 and 2 and never occurs in the absence of G III, may be produced by a command inter-neurone arising in GIII and synaptically activating the motoneurones of spiracles 1 and 2; it may simultaneously suppress the activity of their pacemakers. Pattern 2b may sometimes be a mixture of a and c resulting from low activity in the command interneurone, or due to the effect of irregular suppression of pacemaker activity by inhibitory fibres. The spiracle motoneurones therefore seem to act either as segmental oscillators whose frequency can be adjusted, or as follower cells which are driven by a dominant oscillator in GUI.

As an alternative approach electrical stimulation has been applied to split bundles of the connectives between GI and GII, after removal of GUI, in an attempt to activate the command interneurones and examine their effect on the segmental pacemaker activity. As previously mentioned the technique suffers from the major defect that the same fibres cannot be stimulated in successive preparations and that unavoidably a mass of axons is excited by each shock. Thus responses may be brought about via the natural pathways or by the activation of large numbers of axons none of which acts as the spiracle command fibre in the intact insect. However, the failure to affect the spiracle motoneruones by the simultaneous stimulation of large numbers of lateral afferent nerves, and the different effects of stimulation of different bundles of split connectives, suggest that massive and non-specific input does not account for the observed results.

Rowell (1964) has shown that dissociation of parts of the central nervous system of a locust drastically modifies the threshold of some reflex activities and the following preparations were therefore employed in the experiments:

1. Whole nervous system intact.

2. Central nervous system intact: all lateral thoracic nerves cut.

3. Connectives between GII and GUI cut; lateral nerves of GI and GII either all cut or left intact.

4. Connectives between GI and GII cut; GI lateral nerves all cut or intact.

5. All the above preparations decapitated.

With the exceptions to be mentioned below, the choice of preparation did not affect the results. Although decapitation has marked effects on pacemaker activity, all the responses to connective stimulation could be obtained with or without the head. It was found most convenient to employ preparation 3 in which the frequency of free running could be kept low by the maintenance of hypoxic conditions in the head. Delicate control of the frequency was obtainable by opening and closing the cut ends of cephalic tracheae or by perfusion with hypoxic gases.

Single shocks applied either to a whole connective between GI and GII, or between GII and GUI, or to some bundles of the split connectives between GI and GII, produced a 1:1 response in both axons of spiracles 1 and 2 (Fig. 10). Assuming that the natural route has been invoked, excitation of spiracle 2 probably takes place antidromically when stimulation is applied to the GI-II connective but orthodromically when applied to the GII-III connective. The simultaneous excitation of both spiracles by GI-II stimulation suggests the presence of a single command interneurone supplying both, an hypothesis already postulated to account for impulse synchronization between these spiracles. By stimulating at either end of the connective a conduction speed for the interneurone of 0·7-1·0 m./sec. can be measured. With conduction in the spiracle motor axons taking place at 1 · 1-1·5 m./sec. and a total latency of 12-15 msec. (Fig. 13), a ganglionic delay of 4-7 msec, can be postulated. This suggests the existence of at least one synapse. Since conduction velocity is probably slowed by the use of petroleum jelly, the central delay may be slightly shorter. Knights (1965) has pointed out that in dragonfly larvae an increased intensity of stimulation reduces central delays, part of the effect being due to spatial summation and part possibly to greater stimulus spread. Slight reductions of latency were obtained by increasing the intensity in spiracle motoneurone preparations, and with very high intensities it was possible to stimulate the motoneurones directly (Fig. 12c). High frequencies of stimulation on the other hand increased the latency, presumably due to synaptic fatigue as was found by Knights. Occasionally at low intensities of stimulation the latency was as great as 50 msec., and exceptionally 115 msec., when polysynaptic pathways were probably involved.

Single shocks sometimes gave rise to responses which were interpolated into the free-running rhythm of spiracle 1 (in preparations without GUI) hardly affecting the timing of natural spikes in either axon. Examples are shown in Figs, 10 and 11. In other preparations, however, where different bundles of axons were stimulated, the evoked response was sometimes followed either by a temporary acceleration of free running, or by a post-excitatory depression which lasted for twice the normal spike interval. ‘Inhibitory’ effects of this nature are discussed further below.

In preparations 1 and 2, with G III intact, a similar 1:1 response could be obtained during the expiratory phase when the motoneurones were naturally active, but stimulation failed to produce a response during inspiration when the units remained silent. Thus when the activity of the motoneurones is inhibited by GUI during inspiration, electrical stimulation is without effect; but when inhibition is produced as a result of hypoxic conditions in the head, in preparations without GUI, electrical stimulation is still effective. The inhibitory command interneurone from GUI must therefore act both on the pacemakers and at the synaptic regions of the cells. Head hypoxia, however, affects the pacemakers alone and thus does not prevent the spiracle from making movements synchronized with ventilation under conditions of respiratory stress, since it can still be driven by GUI.

In a few preparations stimulation of small bundles of the connective produced double firing in each axon, the spikes in each couplet being 10-15 msec, apart (Fig. 12f), at a time when the normal spike interval was 80 msec. Such double firing is reminiscent of that which sometimes occurs in flight motoneurones of the locust where Wilson (1964) has recognized an intrinsic refractory oscillatory mechanism. Double firing in response to a single input is also a feature of some Aplysia ganglion cells when they are fatigued or hyperpolarized (Tauc, 1962). Paired impulses occur in natural trains during pattern 2 a firing in spiracle motoneurones of the locust, but it is not known if they also arise from a single input (cf. Hoyle, 1959). Electrical stimulation may of course be able to produce them by the activation of two central paths with different conduction rates.

Threshold stimulation of the connective produces either a response in both spiracle motoneurones or in neither. A response in one alone can be invoked only when its partner is in the refractory period of a preceding P spike. Normally the two units fire synchronously in response to stimulation, but in some preparations their latencies may vary by as much as 20 msec, even when pacemaker activity is suppressed (Fig. 12 d, e). The command interneurone may synapse with both motoneurones, but if only one is caused to fire by a command its activity may in turn excite the other. No such interaction evidently occurs between the pacemaker zones.

By the stimulation of different split bundles of the connective between GII and GUI, several ‘inhibitory’ effects have been produced. In a few preparations in which low-intensity shocks were delivered at 5/sec., the complete inhibition of free running was obtained in the absence of any other response. An increase in intensity produced a 1:1 response and then with a further increase the response was followed by a high frequency of free running. A gradual reduction of intensity produced the same sequence in reverse.

Stimulation at about 40 shocks/sec may produce a 1:1 response with every second shock, but after several seconds the response gradually disappears as the synapse fatigues. After a few seconds rest it can again be obtained. At 50/sec. no response is obtained, free running is inhibited and the nerves remain silent (Fig. 14); inhibition may outlast the period of stimulation by several seconds.

High-frequency stimulation may produce a cessation of synaptic responses as a result of post-excitatory depression or fatigue and this may also affect the pacemaker zone. Low-frequency stimulation (5/sec) at low intensities, however, probably causes inhibition by the activation of a specific inhibitory command fibre. It has not so far been possible to obtain in one and the same preparation separate connective bundles whose stimulation produces either excitation alone or inhibition alone, although each has been obtained on different occasions.

When stimulation is repeated on preparations with GUI intact, the stimulation of one GI-II connective at 50/sec., the other being intact, produces again a complete inhibition of all activity in spiracles 1 and 2. In addition it induces a marked acceleration of ventilation and probably excites the flight motoneurones (cf. Wilson, 1964). Inhibition of spiracle 2 may therefore be brought about by the normal mechanism which operates during flight, but that of spiracle 1, which remains active in flight (Miller, 1960b), is possibly the result of stimulation of an inhibitory interneurone.

Following stimulation at high or low frequencies there is often a prolonged period during which free running continues at an enhanced frequency; it may last for more than 15 min. and can occur in preparations with or without the head (Figs. 15 and 16). Stimulation of bundles of the split GI-II connective has occasionally produced acceleration with no other effect. This can take place via GII, since it will also occur in the spiracle 1 motoneurones after section of the connective between stimulating electrodes and G I. Stimulation of the lateral nerves of GI does not produce an enhancement of frequency. Comparable long-lasting effects were found in cockroach ganglia by Luco (1963) after electrical stimulation, and by Kennedy and Preston (1963) in crayfish ganglion cells.

In order to distinguish between the activities of the postulated command inter-neurones, one connective between GI and GII was progressively blocked either by mechanical damage or by cooling. The other connective was previously cut and the changes in activity of the spiracle 1 nerve were examined. A mechanical block was applied by cutting in turn each bundle of a split connective. The cold block was applied by lifting the connective on a metal hook which was cooled by evaporating ether under reduced pressure in a closed vessel. Such treatment produced a complete block of all activity but was fully reversible when the hook was allowed to return to room temperature. It was thus possible to apply the block several times to the same preparation. Some difficulty in the interpretation of the results is encountered since ventilation was modified considerably by the treatment and this must be distinguished from the selective inactivation of interneurones which supply the spiracle motoneurones.

Five deviations from the normal firing plan were observed during the gradual application of a block, although not all in the same preparation; they are shown in Table 1.

No. 2 (in Table 1) probably results from the general slowing of ventilation which is produced by the block, while nos. 4 and 5 are likely to be the result of the complete loss of contact between GI and GUI. However, nos. 1 and 3 are firing plans which are peculiar to this treatment and have not been seen under different circumstances. No. i may be explained by the selective block of an inhibitory intemeurone and no. 3 by that of an excitatory interneurone which produces pattern za.

The spiracle motoneurones represent a system which is to some extent functionally independent of the ganglia in which they are found. There is no evidence for diffuse connexions of the type which Fielden (1960) has described for motor units in the ganglia of the dragonfly abdomen. Nevertheless, their activity is subject to considerable variation and different motor patterns may be used on different occasions to achieve a maintained closure of the spiracle. Similarly, Ewing & Manning (1966) found that in different individuals of the same species of cockroach, and in different species, similar leg movements were not always achieved by the same motor patterns.

The activity of the spiracle motoneurones is compounded from two sources, the segmental and the metathoracic ganglion, and it may be further affected by head ganglia. The contribution of the segmental ganglion (GI for spiracle 1) is a regular smooth pattern of firing, usually independent in the two units, brought about by the activity of pacemakers which lie in some part of the motoneurone. The metathoracic ganglion drives the motor units via command interneurones and in doing so may suppress the activity of the pacemakers. This driven activity comprises a high-frequency burst (pattern 1) at the beginning of expiration and a maintained barrage of firing for the remainder of the stroke (pattern 2d). The regularity of firing and smooth decline in frequency of the high-frequency burst suggest that there is a sudden and maintained depolarization of synaptic regions in both units which produce the rapid discharge in spike-initiating zones with pacemaker properties, possibly some way from the synapses. Spiracles 1 and 2 appear to be controlled independently since either alone can produce a high-frequency burst, or it can occur in both at different frequencies. Pattern 2 a, which often follows the burst, comprises regular trains of impulses usually synchronized in the two axons and sometimes occurring in couplets in each. Such activity is often synchronized, impulse for impulse in spiracles 1 and 2, suggesting that one command interneurone excites both on a 1:1 basis. Antidromic stimulation of the transverse nerve does not affect this firing plan and local pacemaker action seems to play no part at least in the establishment of the major periods. Such an interneurone may fire at 20/sec. when the overall frequency in each motor axon is 40/sec., if the second spike of each couplet occurs as a result of an oscillation within the motoneurone. Wilson (1964) has shown this to be the case in the double firing of flight motoneurones, but the evidence is inconclusive for the spiracles.

Electrical stimulation of split bundles of the GI-II connective can produce synchronized impulses in each of the four axons which supply the closer muscles of spiracles 1 and 2. Motoneurones supplying the opener muscle of spiracle 1 do not respond in this way. Inhibitory effects and acceleration of pacemaker activity have also been achieved, but these cannot be assigned with certainty to the activation of specific command interneurones. Low frequencies of stimulation can inhibit pacemaker activity ; at higher intensities there is often a 1:1 excitation, whereas higher frequencies of stimulation abolish all activity. Different effects are produced in successive preparations presumably because different combinations of axons are included in the split bundles of the connective. In some preparations 1:1 spikes produced by stimulation of the cord are interpolated into the free-running cycles of the pacemakers without affecting their timing. This suggests that the synaptic region and pacemaker zone may be in different parts of the motoneurone, perhaps in different axon branches, and that the pacemaker is not invaded by synaptically derived spikes; there may thus be two spike-initiating regions. Comparable functional separation of the different regions of the motor units and interneurones in the central nervous system of the crayfish has been demonstrated by Takeda & Kennedy (1965) where they state that the axon branches ‘take on almost the integrative role of separate internuncial elements’. A similar situation has been demonstrated by Tauc & Hughes (1963) in certain Aplysia neurones where not all branches of an axon are invaded by a spike. In other locust preparations, however, 1:1 responses in the spiracle nerves may be followed by temporary depressions before normal pacemaker firing is resumed. This may result from the simultaneous stimulation of excitatory and inhibitory interneurones, the latter with a slower conducting speed, which are both included in some bundles but not in others. But alternatively there may be weak interaction between synaptic and pace-maker regions, and differences between preparations may sometimes depend on their ages or on the extent of damage to tracheae during dissection. The possible effects of stimulus spread acting directly on synapses cannot be excluded.

Pattern 2c, like free running, comprises regular and independent firing in the two motoneurones and it is probable that GIII plays no part in its production beyond exercising a general effect on the frequency. Such activity can be re-phased by antidromic stimulation of the transverse nerve before or after the removal of GIII and is quite independent in adjacent spiracles.

In adult dragonflies spiracle closure is maintained by a pattern of firing in the two axons which closely resembles free running in the locust. Activity corresponding to pattern 2 a has not been seen and it appears that the dragonfly lacks the interneurone responsible for this pattern (Miller, 1965). The presence of an excitatory command interneurone responsible for pattern 2 a in the locust may be explained by the need to maintain spiracle closure at times when free running is slowed by hypoxic conditions in the head. Such activity will then permit abdominal pumping to maintain an efficient through-conduction of air and prevent ventilation from becoming tidal, which probably happens in dragonflies under such conditions.

A tentative model for the arrangement and connexions of one motoneurone supplying the closer muscle of spiracle 1 is shown in Fig. 17. It is depicted with separate pacemaker and synaptic regions and with two inhibitory and three excitatory interneurones which make synaptic connexions. The interneurone responsible for pattern 2 a is shown as synapsing in a separate region of the motoneurone from that which produces the high-frequency burst of pattern 1 ; the latter is shown running to the pacemaker zone, but it may equally well act through a third discrete part of the cell which is able to initiate spike trains in response to a maintained depolarization. Several dozen different models could be drawn all accounting for the observed properties; the virtue of the present one is only that it provides a terse summary of the findings to date. It may be compared with the diagram of the neurone of the cardiac ganglion which is reproduced as Fig. 3-4 by Bullock & Horridge (1965). Indeed the spiracle motoneurones are comparable in several ways with the follower neurones of the crustacean cardiac ganglion (Maynard, 1955). Both are capable of spontaneous activity when isolated from the principal pacemaker (for the spiracle this resides in GUI), but in both such spontaneous activity may seldom appear in the intact animal. Bullock (Bullock & Horridge, 1965, p. 319) has suggested that such spontaneity may represent ‘a way of poising the cells so close to threshold that they will follow: (a) with a small synaptic input, or (b) with a short delay, or (c) in conjunction with the property of noninvasion of the soma and pacemaker potential regions by the spikes...with a high frequency, repetitive burst, or (d) any combination of the three.’ In the spiracle motoneurone such spontaneity is expressed at times, and can in some species contribute to the motor output, as well as possibly maintain the cell in a poised state of readiness.

I am indebted to the Science Research Council for financial support, and to the Anti-Locust Research Centre for the supply of locusts.