Response Properties Of Interneurons Of The Cricket Cercal Sensory System Are Conserved In Spite Of Changes In Peripheral Receptors During Maturation

When an animal matures, it often experiences quantitative and qualitative changes in its sensory inputs (Easter, 1983). Consequently, the central nervous system (CNS) must continually adjust to such changes in order to interpret incoming sensory information accurately. How the CNS manages to conserve its functions during maturation of the animal is a major question in developmental neurobiology. The cereal sensory system of the cricket (Murphey and Chiba, 1990) provides a model system where dynamic matching between the growing periphery and the maturing CNS can be studied at the level of single cells. Moreover, matching in this system seems to be particularly important because it supports such essential activities as escape from predators (Stabel et al. 1985; Gnatzy and Heusslein, 1986; Gnatzy and Kämper, 1990) and communications with other members of the species (Kämper and Dambach, 1981, 1985).

During postembryonic development, a larval cricket grows in body length from approximately 2 mm to 20 mm through a series of nine molts, and reaches the adult stage after about 60 days at 30°C. During this period, a new set of cereal filiform hair sensilla is added at each molt, increasing the total amount of input to the nervous system (Palka and Edwards, 1974; Knyazev and Popov, 1981). In addition, the external structures of the receptors enlarge significantly (Knyazev and Popov, 1981; and data presented below), resulting in a changing quality of input to the CNS (Kanou et al. 1988; Weber and Kämper, 1990; G. Kämper, in preparation). In spite of such developmental changes at the periphery, escape behavior of the orthopteroid insects remains coordinated throughout postembryonic development (Dagan and Volman, 1982). Kanou et al. (1988) showed for the third-instar larvae that frequency threshold curves, an aspect of the response properties, of the interneurons are similar to those in the adults. Their study provides the first physiological evidence that the CNS is successfully accommodating to the ever-changing input from the periphery at the level of the output from the first-order interneurons.

In the present study we confirm the constancy of basic response properties in these interneurons and also describe quantitative changes in the peripheral receptors during postembryonic development of the cricket. The number of peripheral receptors in the cereal sensory system increases almost 40-fold, and the sizes of some identified receptor hairs increase by a factor of 10. Even though these peripheral changes imply dramatically changing sensory input to the CNS, various aspects of the interneuron response properties remain relatively constant during the same period. We propose that dynamic mechanisms are responsible for the relative constancy within the CNS. In particular, we suggest that an important role is played by the previously described systematic rearrangement of the sensory neuron-to-interneuron synapses (Chiba et al. 1988; Chiba and Murphey, 1991) for maintaining a functional CNS in a maturing cricket.

Female house crickets (Acheta domesticus L.) of various postembryonic stages (i.e. sixth instar, eighth instar and adult) were used. Larval crickets were obtained from Flukers Cricket Farm (Baton Rouge, Louisiana, USA) and reared in a standard laboratory colony (Chiba and Murphey, 1991). At the time of experiments, the crickets were 2– 24 h post-ecdysis.

The preparation consisted of an isolated abdomen as described by Shepherd and Murphey (1986). Prior to dissection, the crickets were anesthetized on ice for 1-4 h. The cerci were positioned at a natural angle with approximately 50-60° between them (Fig. 1B). We recorded from the soma of the medial giant interneuron (MGI) or interneuron 10-3 (10-3) (Shepherd and Murphey, 1986; Chiba and Murphey, 1991) in either right or left hemiganglion, and stimulated the cerci with air currents (see below) from a direction that produced the maximal response for that particular interneuron (Fig. 1A; also see Fig. 1C for a sample recording). The identity of an interneuron was confirmed by its response properties, especially the phase of neuronal excitation relative to the phase of low-frequency air particle oscillation (Kämper, 1984), and by injecting the fluorescent dye 6-carboxy fluorescein (Kodak). In later experiments, soma location and the physiological criteria alone were sufficient and dye injection was not required. Recordings with resting potentials greater than −60 mV (mean−68 mV) were used for analyses. All recordings were made at 20±1°C.

The apparatus for generation of air particle oscillation (Kämper, 1984) consisted of a wind tunnel with a loudspeaker (Realistic 4” woofer) attached at one end and the specimen-holding platform placed at the other (Fig. 1B). It allowed precise control of both the direction and velocity of air movement (sinusoidal wave) at the cerci, by using a computer (IBM PC-XT) and a scientific software package (Asyst2.01, Macmillan Software). The computer system generated precisely shaped alternating voltage waves (e.g. Fig. 1C). Digital-to-analog conversion rate was 10 kHz. The speaker output, measured at the open end of the tunnel after signal amplification (Realistic MPA-90), was calibrated with a microphone sensitive to air particle velocity (after Bennet-Clark, 1984, modified). The microphone was pre-calibrated using a laser anemometer (Spectra Physics 215-2, DISA).

To stimulate the cereal sensory neurons, the distal half (4 –5 mm) of the cercus was amputated and two insulated silver wire electrodes (20 μm diameter) were inserted about 2 mm into the cercus (Fig. 2A). The insulation for one wire was stripped off to maximize the electrode-to-cercal tissue contact. The electrodes were connected to a digital stimulator (Getting DS1) through an isolation unit (Grass SIU5). For each preparation, we adjusted the strength of a 300– 350 μs positive square pulse (1–5 V) to just below the firing threshold of the interneuron (see Fig. 2B for a sample recording).

Results of all statistical analyses presented in this study are of two-tailed t-tests.

One important change during postembryonic development of the cricket is the increase in the number of sensory receptors such as filiform hair sensilla on the cercus. Prior to each of the nine larval molts, a new cohort of sensilla is added to those already existing at that stage (Palka and Edwards, 1974). We examined a defined area located dorsolaterally and proximally on the cercus at different stages of development using scanning electron microscopy. In the sixth and eighth instar and the adult, we also examined a second area situated laterally on the cercus and distal to the first one. The density of sensilla within the areas remained constant throughout development at 2.0–2.7 hairs per 10⁶μm² (approximated by the mean value of 2.35). Thus, through extrapolation of the data and assuming a cone shape of the cercus, we estimated the total number of filiform hair sensilla on each cercus (Fig. 3). It increases at an average rate of 55% per instar stage, almost 40-fold, from 55, the known number in the first instar, to approximately 2000 in the adult (see Bentley, 1975; Knyazev and Popov, 1981).

Another significant change during postembryonic development is growth of individual receptor hairs. We measured the lengths of seven identified filiform hair sensilla (Shepherd and Murphey, 1986; Walthall and Murphey, 1986; Shepherd et al. 1988) at various developmental stages. Four of them, 3X, 4X, 6X and 10X (Fig. 4A), belong to a subclass located within a dorsomedial longitudinal strip on the cercus and share a similar directional sensitivity to air current from the posterior of the animal. Three others, 2C, 2B and 3C (Fig. 4B), belong to another subclass located on the dorsolateral and proximal surface of the cercus and have directional sensitivity to air current from the posterior-dorsal side of the animal. The seven sensilla represent numerous others that appear on the cercus at different stages of development, and the number in each name (e.g. ‘3’ in ‘3X’) indicates the postembryonic stage at which the sensillum initially appears on the cercus (Shepherd et al. 1988). All filiform sensilla examined here and elsewhere (Shepherd et al. 1988) persist throughout development following their initial appearances. Each filiform hair is 50–150μm long (mean 105 μm, N = 70) at first appearance. Each time the cricket molts, a new hair is secreted (Gnatzy, 1978), which, on the average, is 150μm longer than the old one. For example, filiform hair 3X initially appears on the cercus in the third instar and is about 120 μm in length; by the time the cricket reaches the adult stage, this hair is about 1200 μm long after repeated step-wise increases coupled to the molting cycles (Fig. 4A). In addition to variation in the initial lengths (e.g. compare 3X in Fig. 4A and 2C in Fig. 4B), the growth rates can vary among the hairs. For example, 2C and 2B initially appear in the second instar but grow to be very different in size in the adult; 2C is approximately 1200μm long and 2B approximately 700μm (Fig. 4B). Nevertheless, in general, long hairs in the adult are not long when they are first born, and many filiform hairs that are initially among the shortest (approximately 100 μm) grow to be among the longest (>1000 μm). Since the oscillation characteristics of the hair depend crucially on its size (Shimozawa and Kanou, 1984; Kämper and Kleindienst, 1990), the growth of the hair during development changes the quality of the sensory information its sensory neuron conveys to the CNS (Kanou et al. 1988; Weber and Kämper, 1990; G. Kämper, in preparation).

These two kinds of changes imply ever-changing input, both qualitative and quantitative, to the CNS during development. Therefore, we next tested whether response properties of first-order interneurons change during development.

Constant response properties of the interneurons during development

We characterized three aspects of the response properties of interneurons MGI and 10-3 during postembryonic development of the cricket: (1) frequency response, (2) intensity response, and (3) rate of response decrement. Our analyses showed that the response properties of these interneurons remain virtually unchanged during development despite the changes at the periphery described above.

Frequency response of MGI and 10-3

Fig. 5 shows examples of intracellular recordings from MGI and 10-3 in the sixth instar and adult. We varied the frequency of the sinusoidal air current within the range 10–200 Hz, while keeping the stimulus intensity constant at a peak velocity of 5.0 mm s^{− 1} at the cerci. In the adult, MGI and 10-3 differ in their relative responsiveness to different stimulus frequencies. MGI responds poorly to low frequencies (10 and 30 Hz), evident from the low numbers of action potentials, but very well to high frequencies (100 and 200Hz). In contrast, 10-3 responds to the entire range of frequency with roughly an equal number of action potentials. In the sixth instar, about half way through postembryonic development, the differences between MGI and 10-3 are equally apparent.

We quantified the data by counting the number of action potentials for each stimulus (Fig. 6A). MGI produces significantly fewer action potentials in response to low frequencies (10 and 30Hz) than does 10–, but its response to high frequencies (100 and 200Hz) is greater than that of 10-3 (P<0.01). The responsiveness of a given interneuron, in terms of the relative number of action potentials, is not significantly different between the sixth instar and adult. Thus, in spite of the large changes occurring in the peripheral nervous system during this interval, each interneuron continues to convey virtually the same information concerning stimulus frequency. In a separate experiment, the same tests were performed using the eighth instars and adults (Fig. 6B). Since the data for the adults in the two sets of experiments differ slightly for some unknown reason (compare Fig. 6A and 6B), we did not combine the two sets of data. These differences may be seasonal, since we collected the data at different times of the year (October 1988 for the data in Fig. 6B, and July and August 1989 for the data in Fig. 6A). Also, they may be technical. We used two different wind tunnels and platforms for the specimens, which were constructed and calibrated similarly, but the highly sensitive cricket sensory system may have detected differences that our calibration instruments did not. Nevertheless, the two data sets in Fig. 6 show similar and internally consistent pictures. MGI and 10-3 are distinct from each other at all stages tested, and the characteristic shapes of their tuning curves remain virtually unchanged between the juvenile and adult stages.

By varying the stimulus intensity in a random order, we estimated firing thresholds of MGI and 10-3 at 10 and 100 Hz. Firing threshold was defined as the stimulus intensity that causes each interneuron to generate action potentials at a rate significantly (P<0.05) higher than background. The comparison of the firing frequencies between stimulated and unstimulated (background) conditions was necessary for 10-3, which normally fires at a low frequency (4–5 Hz) during recording sessions. Ambient noise level, as measured through a sound level meter placed at the open end of the wind tunnel, was approximately 58 dB SPL, or 0.04 mm s^{− 1} air velocity, and this may have been above the firing threshold of 10-3.

Consistent with the previous study by Kanou and Shimozawa (1984), MGI has higher thresholds for both 10 and 100 Hz than does 10-3 (P<0.01). Furthermore, the thresholds of MGI at the two frequencies differ by approximately one logarithmic scale. This is true both as a group as well as within individual specimens (P<0.01). In contrast, 10-3 is almost equally sensitive to both 10 Hz and 100 Hz. The difference between these two interneurons is equally apparent in the sixth instar (P<0.01). For both interneurons we found no significant changes in their threshold sensitivity between the sixth instar and adult stages. Interestingly, at 100Hz, although MGI has a higher firing threshold than does 10-3, its suprathreshold response at 5.0mms^{− 1} peak velocity is nearly double that of 10-3 (Fig. 6A,B). This suggested that MGI, once above its threshold, may respond more sensitively than 10-3 to changes in stimulus intensity. To test this possibility we next characterized the intensity response of MGI and 10-3.

We determined the intensity response of MGI and 10-3 by varying the stimulus intensity above the firing threshold while maintaining the stimulus frequency constant at 100Hz (see Fig. 7A for sample recordings). We counted the number of action potentials during each stimulus and plotted the results with respect to air particle velocity (Fig. 7B). Analyses such as this have been a standard way of characterizing the cereal sensory interneurons (e.g. Palka and Edwards, 1974; Matsumoto and Murphey, 1977). Even though the increase in response for each interneuron is almost linear when plotted on semi-logarithmic coordinates, the two interneurons are clearly different (Fig. 7B). MGI has a much steeper slope than does 10-3 (P<0.01). Importantly, these differences between MGI and 10-3 are conserved in juvenile and adult crickets at intensities above 10mms⁻¹. At these intensities, however, at least some hairs already touch their sockets (Kämper and Kleindienst, 1990) and may stimulate other receptors such as campaniform sensilla. Also, it should be noted that the response of adults is somewhat higher than that of the larvae.

The sample recordings in Fig. 5 show another aspect of differentiation among the interneurons: the rate of response decrement during stimulation is much faster in MGI than in 10-3. Although the idea that MGI’s response pattern is phasic and that of 10-3 is tonic is not new (e.g. Levine and Murphey, 1980; Kanou and Shimozawa, 1984), we describe three experiments that quantitatively demonstrate the difference between the two interneurons. The data provide further evidence that various independent aspects of interneuronal response properties are conserved during development.

First, we examined the response of the interneurons during 200-ms stimulus pulses. An individual response was then divided into ten 20-ms bins, and the number of action potentials in each bin was counted. The response decrement in MGI can be approximated well by a straight line when plotted on semilogarithmic coordinates (Fig. 8B inset). With a 200Hz stimulus, MGI goes from the fastest firing rate of about 200 Hz in the first bin to near zero within 200 ms (Fig. 8B). This is true for both the sixth instar and the adult (Fig. 8B). In contrast, the response of 10-3 shows little decrement during the same stimulation (Fig. 8B, inset). The two interneurons clearly differ both in the sixth instar and in the adult (P<0.01). With a 100 Hz stimulus, the two interneurons differ clearly in the adult (P<0.01) but not so much in the sixth instar (Fig. 8A). In particular, we noted that the initial response of MGI is not vigorous in the sixth instar (see the first and second data points of MGI in the sixth instar in Fig. 8B). Therefore, the prominent phasic response of MGI with 100 Hz stimuli appears to emerge during the later half of postembryonic development.

Second, we compared the rates of response decrement during multi-pulse stimulation (see Fig. 9A for sample recordings). We presented ten stimulus pulses of 200 ms duration, each at 5s inter-stimulus intervals (ISI) and counted the number of action potentials during each pulse. The data, when plotted on semilogarithmic coordinates, demonstrate the difference between MGI and 10-3 in yet another way (Fig. 9B). Consistent with the above analyses, MGI decreases its responsiveness more quickly during repetitive stimulation than does 10-3

(P<0.01). This is true in both the eighth instar and the adult. The difference between MGI and 10-3 is apparent within a wide range of frequencies from 10 to 200 Hz.

Third, we used direct electrical stimulation, rather than air current stimulation, to activate the cereal sensory neurons (see Fig. 10A for sample recordings). So far, the difference in the rate of response decrement between MGI and 10-3 has been analyzed using the number of action potentials generated by these interneurons. An assumption has been that the temporal change in the production of action potentials reflects a change in the synaptic inputs to interneurons. In this experiment, however, we assessed the validity of this hypothesis by monitoring the subthreshold response of MGI and 10-3, while stimulating the sensory axons directly (Fig. 10). Consistent with the above two sets of analyses, which used action potential number (Figs 8 and 9), the compound excitatory postsynaptic potential (EPSP) amplitude decreases more rapidly in MGI than in 10-3 (P<0.05) (Fig. 10B). Therefore, this is the third piece of evidence demonstrating the difference in the rates of response decrement in MGI and 10-3. Owing to technical difficulties, we could not obtain comparable recordings from the sixth-instar crickets.

The differences in the response decrement may be intrinsic to the interneurons or they may reflect diversity and interactions among the synaptic inputs to different interneurons. More specifically, MGI may be able to generate action potentials tonically if the synaptic inputs maintain its membrane potential at a steady level during repetitive activation, or it may still slow down action potential generation even when its cell membrane is continually depolarized. To distinguish these two possibilities, we used an intracellular recording electrode inserted into the dendrites to current-clamp the membrane potential of MGI and observed the patterns of action potentials generated (Fig. 11). The results clearly show that MGI is capable of sustaining a constant rate of action potential generation when its membrane potential is held constant at a suprathreshold level (Fig. 11A). A decreasing rate of action potential generation, a pattern similar to those during air stimulation of the filiform receptors (e.g. Fig. 5), results only when the shape of the current-clamp simulates the decreasing compound EPSP (Fig. 11B). Therefore, the response decrement in MGI seems largely to be due either to the nature of the synapses between the sensory neurons and the interneurons or to interactions between these inputs and other excitatory and inhibitory synaptic inputs characteristic of MGI.

In summary, MGI and 10-3 differ from each other in the rate of response decrement. This difference between the two interneurons, as with the other two aspects of the neuronal response properties analyzed (i.e. the frequency responses and the intensity responses), is apparent at least throughout the last half of postembryonic development. In some cases, however, there is evidence for an increase in the response level (MGI in Fig. 6B and 10-3 in Figs 7B and 8B) and a change in the initial response (MGI in Fig. 8A) during development. This might indicate a refinement of properties of the cereal system.

In this study, we found that each of the two cricket interneurons examined, MGI and 10-3, conserves its basic response characteristics, i.e. the frequency response, intensity response and rate of response decrement, during the last half of postembryonic development. This constancy within the CNS is manifested in spite of large changes in the number and properties of the peripheral receptors.

How does the cricket nervous system accommodate to the developmental changes in the peripheral receptors? Two mechanisms have been proposed by previous studies, and both seem to contribute. First, partial accommodation is accomplished peripherally by adjusting the transducer mechanisms of the receptors during development and, second, a major accommodation results within the CNS through developmentally plastic synapses between the sensory neurons and the interneurons (Fig. 12).

Kanou et al. (1988) first showed that the elasticity of the cereal cuticle decreases as the cricket matures. As a result, filiform hairs in later developmental stages are harder to move than those in earlier stages (Kämper, 1990, in preparation). For example, Kanou et al. (1988) reported that a sensillum with a 500-um hair in the third instar would oscillate as easily as a 1000-μm hair in the adult. Consistent with this, the oscillation amplitude of individual hairs in a sound field measured at different larval stages decreases with age, even through the deflection force increases with hair length (G. Kämper, in preparation). Therefore, a developmental change in cuticular elasticity accommodates in part for the size increase of the hair. However, the cuticle elasticity factor alone is insufficient to account for the full scale of changes accompanying the increase in hair size.

There is good evidence that developmental plasticity of synaptic connections between the sensory neurons and the first-order interneurons is at play as an additional mechanism during development to maintain the constancy of the input-output relationships (Chiba et al. 1988; Murphey and Chiba, 1990; Chiba and Murphey, 1991). When certain sensory neurons are young and thus are associated with relatively short hairs, they form strong synapses with MGI. But as they become older they presumably switch their synaptic partners progressively from MGI to 10-3. Consequently, MGI keeps receiving its major inputs from sensory neurons associated with relatively short filiform hairs, while 10-3’s major inputs are from those associated with relatively long hairs throughout development (Shimozawa and Kanou, 1984; Shepherd et al. 1988). Therefore, synaptic rearrangement could maintain the quality of the sensory inputs to each interneuron during development. This is thought to be a major reason that the outputs from these interneurons remain virtually unchanged.

The relative importance of the above two factors, and others yet to be fully described, cannot be assessed at the moment, although simulation studies in progress may help address this question.

The interneurons of the cereal sensory system of orthopteroid insects (e.g. crickets, cockroaches and locusts) represent the major neuronal pathways connecting the wind-sensitive filiform hair sensilla of the cerci to the motor centers in the thoracic ganglia of the animal (Roeder, 1948; Ritzmann et al. 1982; Camhi, 1984; Kanou and Shimozawa, 1985; Boyan and Ball, 1990). Our study confirms that each interneuron has a distinct set of response properties (Bacon and Murphey, 1984; Kämper, 1984; Kanou and Shimozawa, 1984; Murphey and Chiba, 1990). With its characteristic frequency tuning (Fig. 6) and intensity response (Fig. 7) and its rapid response decrement (Figs 8, 9 and 10), MGI is suited to detect a sudden and transient wind, such as the one caused by an approaching predator (Gnatzy and Kamper, 1990). In contrast, 10-3 responds optimally and continually to weak and low-frequency air currents (based on the data presented in Figs 6-10) which, in the natural environment, could be either ambient wind or courtship songs (Kämper and Dambach, 1981, 1985). The cricket cereal sensory system apparently utilizes parallel processing by these interneurons and others which provide an array of uniquely specialized pathways.

Some of the differences among interneurons presumably reflect differences in the response properties of the sensory neurons innervating each interneuron. For example, MGI has a low sensitivity to low stimulus frequencies of 10-100 Hz (Fig. 7). This is reflected in the high threshold of the short-hair-associated (young) sensory neurons (Shimozawa and Kanou, 1984), the major input to MGI (Shepherd et al. 1988). Therefore, the characteristic frequency response of an interneuron is thought to be a direct consequence of receiving inputs from the sensory neurons conveying such specific frequency responses.

Other differences may arise at the synapses between the sensory neurons and the interneurons. For example, the rapid response decrement in MGI (Fig. 8) may reflect the properties of the synapses for the following reasons. First, recordings from the sensory axons show little or no decrement during continual stimulation of the hairs, regardless of hair size (Palka et al. 1977; Shimozawa and Kanou, 1984; A. Weber and G. Kamper, unpublished data). Second, when the sensory axons are directly activated by electrical stimulation bypassing the sensory transducers, the compound EPSP amplitude decreases very rapidly in MGI (Fig. 10). Finally, when the membrane potential of MGI is depolarized at a steady level above the firing threshold by current injection, there is virtually no reduction in firing rate (Fig. 11A). Therefore, the pattern of response decrement in the interneurons (Figs 8 and 9) may be due to the nature of the individual synapses but not to the mechanisms of action potential generation in either the presynaptic or the postsynaptic cell. Further characterization of transmission at single identified synapses during repetitive activation is currently under way (G. Davis and R. Murphey, personal communication).

When and how do the interneurons acquire their unique sets of response properties? One possibility is that all interneurons begin life with similar response properties and that their diversity emerges as an animal matures. Alternatively, many of the properties may be intrinsic to the circuit from the beginning. Basic differences are noticeable very early (Kanou et al. 1988) and are present even at hatching (Blagburn, 1989). Probably, however, the differentiation of the interneurons is a progressive process during maturation of the nervous system. For example, the response properties of MGI and 10-3 in the sixth instar are slightly (but statistically significantly) less distinct from each other than in the adult (Fig. 8A), and 10-3 in juveniles responds with fewer action potentials to the same stimulus than in adults (Fig. 8B). Examples such as these allow us to speculate that the fully differentiated response properties of each interneuron emerge through continual refinement during development, possibly as a result of experience or neuronal activity (Murphey and Matsumoto, 1976; Matsumoto and Murphey, 1977; Shepherd and Murphey, 1986; Murphey and Chiba, 1990). Roles played by local interneurons (Kobashi and Yamaguchi, 1984; Bodnar et al. 1991) during refinement of the interneuron response properties are yet to be determined.

This study presents direct evidence that interneurons in the cricket conserve, to a large extent, their unique response properties during development in spite of significant changes in sensory inputs. Analogous situations exist in maturing vertebrates, but the evidence that vertebrate central neurons maintain constant response properties during development is indirect. For example, during growth of the retina in fish and frogs, photoreceptors whose receptive field is 90° away from the visual axis in the juvenile are gradually displaced by newly generated receptors towards the center (Easter, 1983). The shifted receptive field of the photoreceptors is presumably compensated for by the axons of retinal ganglion cells ‘sliding’ over the tectal surface (Gaze et al. 1979; Reh and Constantine-Paton, 1984). In some frogs (Xenopus), the two eyes migrate from lateral to frontal positions so that the binocular visual field shifts during maturation. The polysynaptic connections between the two tecta, which are responsible for binocular vision, shift their arborizations accordingly (Grant and Keating, 1986). In these examples, the best guess is that the ‘sliding’ projections provide the basis of the constant responses within the CNS. Behavioral tests by Northmore (1981) support this notion. In the insect system examined in this study, we obtained direct evidence that the central neurons (interneurons) maintain virtually constant response properties. As in vertebrate visual systems, this is thought to be achieved through developmental plasticity within the sensory pathway. Apparently, flexibility of the nervous system during development is a common feature among vertebrates and insects (Murphey, 1986). Moreover, we suggest that neural flexibility is a key to maintaining functions of the CNS in the face of various developmental changes at the periphery of an animal.

Special thanks to H. Hirsch, G. Lnenicka, R. Olberg, J. Schmidt and our colleagues at the Universities of Massachusetts and Ulm for their helpful comments on the manuscript. Supported by grants from the DFG to G.K. (Ka 662/2-2 and Ka 662/2-3) and the NSF BNS-87-19377 to R.K.M.