ABSTRACT High-speed videos were used to analyse whether and how adults of a winged species of scorpion fly (Mecoptera, Panorpa communis ) jump and determine whether they use the same mechanism as that of the only other mecopteran known to jump, the wingless snow flea, Boreus hyemalis . Adult females are longer and heavier than males and have longer legs, but of the same relative proportions. The middle legs are 20% longer and the hind legs 60% longer than the front legs. A jump starts with the middle and hind legs in variable positions, but together, by depressing their coxo-trochanteral and extending their femoro-tibial joints, they accelerate the body in 16–19 ms to mean take-off velocities of 0.7–0.8 m s −1 ; performances in males and females were not significantly different. Depression of the wings accompanies these leg movements, but clipping them does not affect jump performance. Smooth transition to flapping flight occurs once airborne with little loss of energy to body rotation. Ninety percent of the jumps analysed occurred without an observable stimulus; the remaining 10% were in response to a mechanical touch. The performance of these jumps was not significantly different. In its fastest jumps, a scorpion fly experiences an acceleration of 10 g , expends 23 µJ of energy and requires a power output less than 250 W kg −1 of muscle that can be met by direct muscle contractions without invoking an indirect power amplification mechanism. The jumping mechanism is like that of snow fleas.

ABSTRACT The order Hemiptera includes jumping insects with the fastest take-off velocities, all generated by catapult mechanisms. It also contains the large family Miridae or plant bugs. Here, we analysed the jumping strategies and mechanisms of six mirid species from high-speed videos and from the anatomy of their propulsive legs, and conclude that they use a different mechanism in which jumps are powered by the direct contractions of muscles. Three strategies were identified. First, jumping was propelled only by movements of the middle and hind legs, which were, respectively, 140% and 190% longer than the front legs. In three species with masses ranging from 3.4 to 12.2 mg, depression of the coxo-trochanteral and extension of femoro-tibial joints accelerated the body in 8–17 ms to take-off velocities of 0.5–0.8 m s −1 . The middle legs lost ground contact 5–6 ms before take-off so that the hind legs generated the final propulsion. The power requirements could be met by the direct muscle contractions so that catapult mechanisms were not implicated. Second, other species combined the same leg movements with wing beating to generate take-off during a wing downstroke. Third, up to four wingbeat cycles preceded take-off and were not assisted by leg movements. Take-off velocities were reduced and acceleration times lengthened. Other species from the same habitat did not jump. The lower take-off velocities achieved by powering jumping by direct muscle contractions may be offset by eliminating the time taken to load catapult mechanisms.

ABSTRACT Many insects such as fleas, froghoppers and grasshoppers use a catapult mechanism to jump, and a direct consequence of this is that their take-off velocities are independent of their mass. In contrast, insects such as mantises, caddis flies and bush crickets propel their jumps by direct muscle contractions. What constrains the jumping performance of insects that use this second mechanism? To answer this question, the jumping performance of the mantis Stagmomantis theophila was measured through all its developmental stages, from 5 mg first instar nymphs to 1200 mg adults. Older and heavier mantises have longer hind and middle legs and higher take-off velocities than younger and lighter mantises. The length of the propulsive hind and middle legs scaled approximately isometrically with body mass (exponent=0.29 and 0.32, respectively). The front legs, which do not contribute to propulsion, scaled with an exponent of 0.37. Take-off velocity increased with increasing body mass (exponent=0.12). Time to accelerate increased and maximum acceleration decreased, but the measured power that a given mass of jumping muscle produced remained constant throughout all stages. Mathematical models were used to distinguish between three possible limitations to the scaling relationships: first, an energy - limited model (which explains catapult jumpers); second, a power-limited model; and third, an acceleration - limited model. Only the model limited by muscle power explained the experimental data. Therefore, the two biomechanical mechanisms impose different limitations on jumping: those involving direct muscle contractions (mantises) are constrained by muscle power, whereas those involving catapult mechanisms are constrained by muscle energy.

Rapid locomotion in the aquatic larvae and pupae of the phantom midge Chaoborus crystallinus was analysed. A 10-mm long larva moved sporadically by rapidly curling into a tight circle and then unfurling. The most common movement (70% of all movements) was a body rotation of 332±22 deg (mean ± s.d.) that lasted 63±19 ms and reached a peak velocity of 0.07±0.02 m s −1 . If the head unfurled earlier in the cycle, the rotation was smaller and the larva dived downwards. A distinct category of single rotations of approximately 180 deg (8%) resulted in a larva finishing with its head pointing in the opposite direction. A sequence of rotational movements (22%) resulted in more extensive displacements. The area of the tail fan was reduced by folding during part of a cycle. It was made of a row of 26 radiating filaments with interlacing hairs between adjacent filaments and resilin at their ventral midline articulations with the body. The fan sprang back passively to its splayed position after being forcibly folded. Reducing the area of the fan by 80% decreased angular rotation and impaired stability so that 33% of movements ended with the body upside down. A 6 mm long pupa also moved by curling and unfurling motions of the head and tail that lasted 215±19 ms and generated slower velocities of 0.03±0.01 m s −1 . The pupal tail fan was membranous, oriented differently, had resilin at its articulations and its area could be changed.

SUMMARY The kinematics and jumping performance of treehoppers (Hemiptera, Auchenorrhyncha, Membracidae) were analysed from high speed images. The eight species analysed had an 11-fold range of body mass (3.8-41 mg) and a 2-fold range of body length (4.1-8.4 mm). Body shape was dominated by a prothoracic helmet that projected dorsally and posteriorly over the body, and in some species forwards to form a protruding horn. Jumping was propelled by rapid depression of the trochantera of the hindlegs. The hindlegs were only 30-60% longer than the front and middle legs, and 47-94% the length of the body in different species. They were slung beneath the body and moved together in the same plane. In preparation for a jump, the hindlegs were initially levated and rotated forwards so that the femora were pressed into indentations of the coxae. The tibiae were flexed about the femora and the tarsi were placed on the ground directly beneath the lateral edges of the abdomen. Movements of the front and middle legs adjusted the angle of the body relative to the ground, but for most treehoppers this angle was small, so that the body was almost parallel to the ground. The rapid depression of the hindlegs accelerated the body to take-off in 1.2 ms in the lighter treehoppers and 3.7 ms in the heavier ones. Take-off velocities of 2.1-2.7 m s -1 were achieved and were not correlated with body mass. In the best jumps, these performances involved accelerations of 560-2450 m s -2 ( g forces of 47-250), an energy expenditure of 13.5-101 μJ, a power output of 12-32 mW and exerted a force of 9.5-29 mN. The power output per mass of muscle far exceeds the maximum active contractile limit of normal muscle. Such requirements indicate that treehoppers must be using a power amplification mechanism in a catapult-like action. Some jumps were preceded by flapping movements of the wings, but the propulsive movements of the hindlegs were crucial in achieving take-off.

SUMMARY Jumping mechanisms and performance were analysed in three species of psyllids (Hemiptera, Sternorrhyncha) that ranged from 2 to 4 mm in body length and from 0.7 to 2.8 mg in mass. Jumping was propelled by rapid movements of the short hind legs, which were only 10–20% longer than the other legs and 61–77% of body length. Power was provided by large thoracic muscles that depressed the trochantera so that the two hind legs moved in parallel planes on either side of the body. These movements accelerated the body to take-off in 0.9 ms in the smallest psyllid and 1.7 ms in the largest, but in all species imparted a rapid forward rotation so that at take-off the head pointed downwards, subtending angles of approximately −60 deg relative to the ground. The front legs thus supported the body just before take-off and either lost contact with the ground at the same time as, or even after, the hind legs. In the best jumps from the horizontal, take-off velocity reached 2.7 m s −1 and the trajectory was steep at 62–80 deg. Once airborne, the body spun rapidly at rates of up to 336 Hz in the pitch plane. In many jumps, the wings did not open to provide stabilisation, but some jumps led directly to sustained flight. In their best jumps, the smallest species experienced a force of 637 g . The largest species had an energy requirement of 13 μJ, a power output of 13 mW and exerted a force of nearly 10 mN. In a rare jumping strategy seen in only two of 211 jumps analysed, the femoro-tibial joints extended further and resulted in the head pointing upwards at take-off and the spin rate being greatly reduced.

SUMMARY If a hindleg of a locust slips during jumping, or misses its target during kicking, energy generated by the two extensor tibiae muscles is no longer expended in raising the body or striking a target. How, then, is the energy in a jump (4100–4800 μJ) or kick (1700 μJ) dissipated? A specialised buckling region found in the proximal hind-tibia where the bending moment is high, but not present in the other legs, buckled and allowed the distal part of the tibia to extend. In jumps when a hindleg slipped, it bent by a mean of 23±14 deg at a velocity of 13.4±9.5 deg ms –1 ; in kicks that failed to contact a target it bent by 32±16 deg at a velocity of 32.9±9.5 deg ms –1 . It also buckled 8.5±4.0 deg at a rate of 0.063±0.005 deg ms –1 when the tibia was prevented from flexing fully about the femur in preparation for both these movements. By experimentally buckling this region through 40 deg at velocities of 0.001–0.65 deg ms –1 , we showed that one hindleg could store about 870 μJ on bending, of which 210 μJ was dissipated back to the leg on release. A band of blue fluorescence was revealed at the buckling region under UV illumination that had the two key signatures of the elastic protein resilin. A group of campaniform sensilla 300 μm proximal to the buckling region responded to imposed buckling movements. The features of the buckling region show that it can act as a shock absorber as proposed previously when jumping and kicking movements go wrong.

SUMMARY Pygmy mole crickets live in burrows at the edge of water and jump powerfully to avoid predators such as the larvae and adults of tiger beetles that inhabit the same microhabitat. Adults are 5–6 mm long and weigh 8 mg. The hind legs are dominated by enormous femora containing the jumping muscles and are 131% longer than the body. The ratio of leg lengths is: 1:2.1:4.5 (front:middle:hind, respectively). The hind tarsi are reduced and their role is supplanted by two pairs of tibial spurs that can rotate through 180 deg. During horizontal walking the hind legs are normally held off the ground. Jumps are propelled by extension of the hind tibiae about the femora at angular velocities of 68,000 deg s −1 in 2.2 ms, as revealed by images captured at rates of 5000 s −1 . The two hind legs usually move together but can move asynchronously, and many jumps are propelled by just one hind leg. The take-off angle is steep and once airborne the body rotates backwards about its transverse axis (pitch) at rates of 100 Hz or higher. The take-off velocity, used to define the best jumps, can reach 5.4 m s −1 , propelling the insect to heights of 700 mm and distances of 1420 mm with an acceleration of 306 g . The head and pronotum are jerked rapidly as the body is accelerated. Jumping on average uses 116 μJ of energy, requires a power output of 50 mW and exerts a force of 20 mN. In jumps powered by one hind leg the figures are about 40% less.

SUMMARY Many animals move so fast that there is no time for sensory feedback to correct possible errors. The biomechanics of the limbs participating in such movements appear to be configured to simplify neural control. To test this general principle, we analysed how froghopper insects control the azimuth direction of their rapid jumps, using high speed video of the natural movements and modelling to understand the mechanics of the hind legs. We show that froghoppers control azimuth by altering the initial orientation of the hind tibiae; their mean angle relative to the midline closely predicts the take-off azimuth. This applies to jumps powered by both hind legs, or by one hind leg. Modelling suggests that moving the two hind legs at different times relative to each other could also control azimuth, but measurements of natural jumping showed that the movements of the hind legs were synchronised to within 32 μs of each other. The maximum timing difference observed (67 μs) would only allow control of azimuth over 0.4 deg. to either side of the midline. Increasing the timing differences between the hind legs is also energetically inefficient because it decreases the energy available and causes losses of energy to body spin; froghoppers with just one hind leg spin six times faster than intact ones. Take-off velocities also fall. The mechanism of azimuth control results from the mechanics of the hind legs and the resulting force vectors of their tibiae. This enables froghoppers to have a simple transform between initial body position and motion trajectory, therefore potentially simplifying neural control.

SUMMARY The hind legs of Issus (Hemiptera, Issidae) move in the same plane underneath the body, an arrangement that means they must also move synchronously to power jumping. Moreover, they move so quickly that energy must be stored before a jump and then released suddenly. High speed imaging and analysis of the mechanics of the proximal joints of the hind legs show that mechanical mechanisms ensure both synchrony of movements and energy storage. The hind trochantera move first in jumping and are synchronised to within 30 μs. Synchrony is achieved by mechanical interactions between small protrusions from each trochantera which fluoresce bright blue under specific wavelengths of ultra-violet light and which touch at the midline when the legs are cocked before a jump. In dead Issus , a depression force applied to a cocked hind leg, or to the tendon of its trochanteral depressor muscle causes a simultaneous depression of both hind legs. The protrusion of the hind leg that moves first nudges the other hind leg so that both move synchronously. Contractions of the trochanteral depressor muscles that precede a jump bend the metathoracic pleural arches of the internal skeleton. Large areas of these bow-shaped structures fluoresce bright blue in ultraviolet light, and the intensity of this fluorescence depends on the pH of the bathing saline. These are key signatures of the rubber-like protein resilin. The remainder of a pleural arch consists of stiff cuticle. Bending these composite structures stores energy and their recoil powers jumping.

SUMMARY To assess the effect of leg length on jumping ability in small insects, the jumping movements and performance of a sub-family of leafhopper insects(Hemiptera, Auchenorrhyncha, Cicadellidae, Ulopinae) with short hind legs were analysed and compared with other long-legged cicadellids (Hemiptera,Auchenorrhyncha, Cicadellidae). Two species with the same jumping characteristics but distinctively different body shapes were analysed: Ulopa , which had an average body length of 3 mm and was squat, and Cephalelus , which had an average body length of 13 mm with an elongated body and head. In both, the hind legs were only 1.4 times longer than the front legs compared with 1.9–2.3 times in other cicadellid leafhoppers. When the length of the hind legs was normalised relative to the cube root of their body mass, their hind legs had a value of 1–1.1 compared with 1.6–2.3 in other cicadellids. The hind legs of Cephalelus were only 20% of the body length. The propulsion for a jump was delivered by rapid and synchronous rotation of the hind legs about their coxo-trochanteral joints in a three-phase movement, as revealed by high-speed sequences of images captured at rates of 5000 s –1 . The hind tarsi were initially placed outside the lateral margins of the body and not apposed to each other beneath the body as in long-legged leafhoppers. The hind legs were accelerated in 1.5 ms ( Ulopa ) and 2 ms( Cephalelus ) and thus more quickly than in the long-legged cicadellids. In their best jumps these movements propelled Ulopa to a take-off velocity of 2.3 m s –1 and Cephalelus to 2 m s –1 , which matches that of the long-legged cicadellids. Both short-legged species had the same mean take-off angle of 56° but Cephalelus adopted a lower angle of the body relative to the ground(mean 15°) than Ulopa (mean 56°). Once airborne, Cephalelus pitched slowly and rolled quickly about its long axis and Ulopa rotated quickly about both axes. To achieve their best performances Ulopa expended 7 μJ of energy, generated a power output of 7 mW, and exerted a force of 6 mN; Cephalelus expended 23μJ of energy, generated a power output of 12 mW and exerted a force of 11 mN. There was no correlation between leg length and take-off velocity in the long- and short-legged species, but longer legged leafhoppers had longer take-off times and generated lower ground reaction forces than short-legged leafhoppers, possibly allowing the longer legged leafhoppers to jump from less stiff substrates.

SUMMARY Nonspiking interneurons are important components of the premotor circuitry in the thoracic ganglia of insects. Their action on postsynaptic neurons appears to be predominantly inhibitory, but it is not known which transmitter(s) they use. Here, we demonstrate that many but not all nonspiking local interneurons in the locust metathoracic ganglion are immunopositive for GABA (γ-aminobutyric acid). Interneurons were impaled with intracellular microelectrodes and were shown physiologically to be nonspiking. They were further characterized by defining their effects on known leg motor neurons when their membrane potential was manipulated by current injection. Lucifer Yellow was then injected into these interneurons to reveal their cell bodies and the morphology of their branches. Some could be recognised as individuals by comparison with previous detailed descriptions. Ganglia were then processed for GABA immunohistochemistry. Fifteen of the 17 nonspiking interneurons studied were immunopositive for GABA, but two were not. The results suggest that the majority of these interneurons might exert their well-characterized effects on other neurons through the release of GABA but that some appear to use a transmitter other than GABA. These nonspiking interneurons are therefore not an homogeneous population with regard to their putative transmitter.

We used the cobalt-backfilling method to map the somata of neurones with axons that project in the two paired lateral nerves of the abdominal neuromeres of the locust Schistocerca gregaria with the objective of expanding and bringing together the incomplete and scattered information on these efferent neurones. We compared somata sizes and positions, and the pathways of primary neurites, with information from previous studies on individual, or groups of, abdominal neurones and we identify many of the somata we mapped. The stained somata belong to paired motor neurones and paired neurosecretory neurones, to unpaired neuromodulatory neurones (dorsal unpaired median, DUM, neurones) and unpaired bilaterally projecting neurones. In different neuromeres, the total number of somata with axons in these lateral nerves ranges from 73 to 106. Within an individual segmental neuromere, approximately 25 % of the somata belong to neurones with axons in nerve 1 (N1) and 35 % to those with axons in nerve 2 (N2) of that segment, while the remaining 40 % belong to neurones with axons in N1 of the next posterior segment. This basic pattern is repeated in all abdominal neuromeres, with differences in the percentages depending on whether the neuromeres are pregenital fused, pregenital unfused or genital. Nerve 1 contains the axons of 26–37 neurones with central somata in different neuromeres, of which 40 % are in the segmental neuromere and 60 % in the next anterior neuromere. In the segmental neuromere, 15 % of somata are ipsilateral to the nerve, 30 % are at the midline and 55 % are contralateral, whereas in the next anterior neuromere, 70 % are ipsilateral, 10 % are at the midline and 20 % are contralateral. Nerve 2 contains the axons of 11–28 neurones in different neuromeres, all of which have somata in the same segmental neuromere from which the nerve projects. Of these, approximately 70 % are ipsilateral, 30 % at the midline and none contralateral, except for the first abdominal and eighth male abdominal neuromeres, where one and two somata, respectively, are contralateral.

The segmental ganglia of the locust contain efferent neuromodulatory neurones with cell bodies at the dorsal midline and axons that supply muscles and other tissue on both sides of the body. These are the dorsal unpaired median (DUM) neurones. Intracellular recordings were made from pairs of known metathoracic efferent DUM neurones in locusts in which all nerves were intact and in isolated metathoracic ganglia. The 19 metathoracic, efferent DUM neurones were identified according to the nerve roots through which their axons emerge from the ganglion. The synaptic potentials in these DUM neurones have been analysed to investigate how these neurones are activated and how their spikes are controlled. The degree of correlation between the synaptic potentials in particular pairs of neurones was quantified using a correlation analysis. This allowed the population of DUM neurones to be divided into three subsets that also map onto an anatomical grouping based on the distribution of their axons in the lateral nerves: (i) DUM1 neurones (DUMDL and DUM1b); (ii) DUM3 and DUM3,4 neurones; and (iii) DUM3,4,5, DUM5b neurones and DUMETi. Individual neurones within each subset showed strong correlations between their synaptic potentials, in both intact locusts and isolated ganglia, and tended to spike at the same time. Neurones in different subsets had few synaptic potentials in common and tended to spike independently. The persistence of common synaptic potentials in neurones of the three subsets in isolated ganglia indicates that they are derived from neurones within the metathoracic ganglion. The DUM neurones that had many common synaptic potentials in a quiescent locust responded in similar ways to mechanosensory stimulation of different parts of the body. DUM3,4, 5 and DUM5 neurones gave the clearest and most consistent responses to stimulation of mechanoreceptors on either hind leg. DUM3 and DUM3, 4 neurones responded variably, but usually with a hyperpolarisation. DUM1 neurones were rarely excited by mechanosensory stimuli but, like the preceding group, their responses were dependent upon whether the locust was moving its legs. These results lend further support to the idea that there is a subdivision of action amongst this population of DUM neurones, with those supplying the same targets being driven by the same presynaptic local neurones.

The flexor tibiae muscle of a locust hind leg consists of 10-11 pairs of fibre bundles in the main body of the muscle and a distal pair of bundles that form the accessory flexor muscle, all of which insert onto a common tendon. It is much smaller than the antagonistic extensor tibiae muscle and yet it is innervated by nine excitatory motor neurons, compared with only two for the extensor. To determine the pattern of innervation within the muscle by individual motor neurons, branches of the nerve (N5B2) that supplies the different muscle bundles were backfilled to reveal somata in the metathoracic ganglion. This showed that different muscle bundles are innervated by different numbers of excitatory motor neurons. Physiological mapping of the innervation was then carried out by intracellular recordings from the somata of flexor motor neurons in the metathoracic ganglion using microelectrodes. Spikes were evoked in these neurons by the injection of current, and matching junctional potentials were sought in fibres throughout the muscle using a second intracellular electrode. Each motor neuron innervates only a restricted array of muscle fibres and, although some innervate a larger array than others, none innervates fibres throughout the muscle. Some motor neurons innervate only proximal fibres and others only more distal fibres, so that the most proximal and most distal bundles of muscle fibres are innervated by non-overlapping sets of motor neurons. More motor neurons innervate proximal bundles than distal ones, and there are some asymmetries in the number of motor neurons innervating corresponding bundles on either side of the tendon. Individual motor neurons cause slow, fast or intermediate movements of the tibia, but their patterns of innervation overlap in the different muscle bundles. Furthermore, individual muscle fibres may also be innervated by motor neurons with different properties.

Locusts are usually quiescent at night, but this state can be interrupted by spontaneous periods of motor activity, or arousals, that can also be induced by exposure to light stimuli. To investigate whether repeated arousing stimulation has any lasting effect on behaviour, locusts were confronted at night with a series of 1 s light stimuli. Groups of three stimuli at intervals of 60 s were repeated 11 times at 10 min intervals during the first experimental night, and three stimuli at intervals of 90 s were repeated at 15 min intervals during the next night. Arousals and the effects of stimulation were monitored as changes in the spike activity of muscles in the basal part (the scapus) of the right antenna. In the early part of the night preceding the presentation of the light stimuli, neither 60 s nor 90 s periods were present as significant peaks in spontaneous changes in spike activity. The initial stimulus of a series evoked an arousal response that habituated on repetition of the stimulus. The end of the series of stimuli was followed by changes in spike activity that tended to have the same periodicity as the preceding stimuli. Furthermore, a single light stimulus at the end of the night evoked changes in spike activity that again tended to have the same periodicity as the preceding entraining stimuli. Repeated stimulation may therefore establish a memory trace for the period of stimulation that can be recalled either spontaneously or by the application of an appropriate external stimulus.

The hair plate proprioceptors at the thoraco-coxal joint of insect limbs provide information about the movements of the most basal joint of the legs. The ventral coxal hair plate of a middle leg consists of group of 10-15 long hairs (70 microns) and 20-30 short hairs (30 microns). The long hairs are deflected by the trochantin as the leg is swung forward during the swing phase of walking, and their sensory neurones respond phasically during an imposed deflection and tonically if the deflection is maintained. Selective stimulation of the long hairs elicits a resistance reflex that rotates the coxa posteriorly and is similar to that occurring at the transition from the swing to the stance phase of walking. The motor neurones innervating the posterior rotator and adductor coxae muscles are excited, and those to the antagonistic anterior rotator muscle are inhibited. By contrast, selective stimulation of the short hairs leads only to a weak inhibition of the anterior rotator. The excitatory effects of the long hairs are mediated, in part, by direct connections between their sensory neurones and particular motor neurones. A spike in a sensory neurone elicits a short-latency depolarising postsynaptic potential (PSP) in posterior rotator and adductor motor neurones whose amplitude is enhanced by hyperpolarising current injected into the motor neurone. When the calcium in the saline is replaced with magnesium, the amplitude of the PSP is reduced gradually, and not abruptly as would be expected if an interneurone were interposed in the pathway. Several sensory neurones from long hairs converge to excite an individual motor neurone, evoking spikes in some motor neurones. The projections of the sensory neurones overlap with some of the branches of the motor neurones in the lateral association centre of the neuropile. It is suggested that these pathways would limit the extent of the swing phase of walking and contribute to the switch to the stance phase in a negative feedback loop that relieves the excitation of the hairs by rotating the coxa backwards.

The processing of mechanosensory signals responsible for the reflex adjustment of the posture or movement of the legs of the locust is described in terms of the actions and connections of identified neurones. Signals can be followed from the major classes of exteroceptors of a leg, through their various integrative stages in the central nervous system to their emergence as specific patterns in known motor neurones. Particular emphasis is placed on the integrative roles of two classes of local interneurones. The spiking local interneurones map the leg as a series of overlapping receptive fields and reverse the sign of the afferent input. The nonspiking local interneurones control the output of the motor neurones by the graded release of chemical transmitter and can adjust the gain of a local reflex depending on the position and movements of the joints of that leg. The reflex movements of one leg must not impair the stability of the animal and must therefore be influenced by events at the other legs. Populations of intersegmental interneurones convey sensory information from one segment to another to ensure such coordination. These interneurones do not produce stereotyped intersegmental reflexes but, instead, alter the performance of a local reflex in a distant leg by making synaptic connections with nonspiking local interneurones. These connections change the effectiveness of the outputs to the motor neurones and consequently the local reflex. The local interneurones therefore play a crucial role both in the production of local reflexes and in the integration of these actions with the movements of the other legs.

The innervation pattern of inhibitory motor neurones of the locust has been revealed by intracellular recording from their cell bodies in the meso- and metathoracic ganglion and simultaneous recording from muscle fibres in a middle, or in a hind leg. Three neurones in each ganglion, the common inhibitor (CI = CI1), the anterior inhibitor (AI = CI2), and the posterior inhibitor (PI = CI3) innervate several muscles in one leg and are thus common inhibitory neurones. Metathoracic CI innervates 13 muscles in one hind leg and mesothoracic CI innervates 12 muscles in one middle leg. The muscles are all in the proximal parts of the legs and move the coxa, the trochanter and the tibia. Metathoracic AI and PI innervate four muscles in the more distal parts of one hind leg that move the tibia, the tarsus and the unguis. None of these muscles is innervated by CI. Each inhibitor innervates muscles that have different and often antagonistic actions during movements of a leg. AI and PI receive many synaptic inputs in common and show similar patterns of spikes during imposed movements of a tibia. Tests fail, however, to reveal evidence for any electrical or synaptic coupling between them. A revised scheme of nomenclature for these inhibitory neurones is proposed.