An interneurone of unusual morphology is tuned to the female song frequency in the bushcricket Ancistrura nigrovittata (Orthoptera, Phaneropteridae)

Information in acoustic signals is contained in at least two variables, carrier frequency and temporal pattern (=amplitude modulation). These may be processed together or independently. For example, frequency is important for long-range communication in the Australian cicada Cystosoma saundersii (female cicadas approaching a singing male by flight), while temporal pattern is important during short-range communication (females sitting close to a courting male; Doolan and Young, 1989). In the cricket Teleogryllus oceanicus, however, a conspecific temporal pattern presented to a flying cricket at the conspecific frequency elicits positive phonotaxis, while the same pattern at ultrasonic frequencies elicits negative phonotaxis (e.g. Moiseff et al., 1978; Nolen and Hoy, 1986). Acoustically active insects are especially suited for investigating the neural basis of song detection and song recognition, since the number of nerve cells potentially involved in these tasks is small compared with that in vertebrates and manipulation of single cells may have drastic behavioural effects (e.g. Schildberger and Hörner, 1988).

The subject of this paper is the duetting bushcricket Ancistrura nigrovittata, in which the male produces a song consisting of a chirp (eight pulses of 7 ms at intervals of 22 ms) followed by a trigger syllable approximately 350 ms later (Heller and von Helversen, 1986; see Fig. 2A). A responsive female replies with a short click (of a few milliseconds) at a constant latency of 30–45 ms in response to the trigger syllable. A female song is approximately 10 dB less intense (75–85 dB SPL peak) than a male song (measured at a distance of 10 cm). A conspicuous feature of the song in this species is the difference in song frequency between males (14–16 kHz) and females (28–30 kHz). This difference is paralleled by a difference in behavioural tuning curves for eliciting a female reply (lowest threshold at 12–16 kHz) and a male phonotactic response (lowest threshold at 24–28 kHz; Dobler et al., 1994). Evidently, the two behavioural responses that are tuned to the song frequency of the opposite sex depend on different neuronal substrates. No sex-specific differences correlated to the behavioural differences have yet been detected, either in the hearing organ or at the prothoracic level (Stumpner, 1996a, 1997, 1998; A. Stumpner, unpublished observations). An interneurone in the prothoracic ganglion (the termination site of auditory receptor cells) with an axon ascending to the brain and with activity showing a close correlation to the dependence of the female response threshold for male song frequency has been identified (Stumpner, 1997). All available evidence suggests that this neurone is homologous to AN1 in crickets, which has been shown to be crucial for female phonotaxis to the male calling song (Schildberger and Hörner, 1988; Atkins et al., 1992; Stumpner et al., 1995). The only other ascending auditory interneurone that has been reliably identified in crickets so far is AN2, which seems to be involved in both calling song phonotaxis and negative phonotaxis away from ultrasound during flight (Nolen and Hoy, 1984; Schildberger and Hörner, 1988; Atkins et al., 1992; Henley et al., 1992).

Might a neurone of A. nigrovittata homologous to AN2 in crickets be tuned to the (ultrasonic) female reply song? In addition to AN1, there are at least two more ascending neurones in A. nigrovittata that are potentially homologous to AN2 (Stumpner, 1996b). None of these ascending neurones shows a frequency tuning corresponding to the behavioural tuning of male phonotaxis (A. Stumpner, unpublished observations). One (neurone AN2) is most sensitive to 20 kHz, while the other (neurone AN3) shows a similar tuning to that of the whole tympanic nerve. Instead, activity in a neurone with a completely different morphology that has not been described in any other insect species was found to show the closest correlation to this behavioural response. The physiological features of this neurone will be described in detail in this report. Its morphological features will be described and compared with those of a typical auditory interneurone elsewhere (Stumpner, 1999).

Both sexes of the bushcricket Ancistrura nigrovittata (Brunner von Wattenwyl) were used in this investigation. Experimental animals were individuals either wild-captured from Northern Greece or taken from first or second laboratory-reared generations.

The electrophysiological methods used have been described in detail by Stumpner (1997). In summary, the procedure was as follows: an experimental animal was briefly anaesthetised with CO₂ and fixed ventral-side-up to a plastic holder with a wax–resin mixture. The forelegs were fixed in a normal (inverse) standing position. The prothoracic ganglion was exposed and stabilized using a Ni–Cr spoon from below and a steel ring from above. Prior to starting some of the experiments, a fragment of collagenase (Sigma) was placed on the air-exposed ganglion for 90 s followed by several washes with saline to facilitate penetration of the ganglionic sheath by the glass capillary. Exposed tissue was bathed with Fielden’s saline (Fielden, 1960). Thick-walled borosilicate glass capillaries were either filled with neurobiotin (5 % w/v in 1 mol l⁻¹ potassium acetate; Vector) or in few cases with Lucifer Yellow CH (5 % w/v in 0.5 mol l⁻¹ LiCl; Sigma). Electrode resistances ranged from 80 to 160 MΩ. Recordings were amplified using a direct-current amplifier, continuously displayed on screen and stored on a digital recorder (Sony). Following physiological characterization, neurobiotin or Lucifer Yellow was ionophoretically injected for 3–55 min. In the course of some experiments, one leg was cut to eliminate auditory input from one ear. The physiological results presented in this paper are based on recordings of six neurones from females and 15 neurones from males. The morphological data are based on four neurones from females and 12 neurones from males and on 25 receptor cells from both sexes.

The mounted bushcricket was placed in an anechoic chamber with dynamic speakers (Dynaudio DF 21) on both the left and right at a distance of 37 cm from the animal. Acoustic stimuli were produced with a computer-controlled stimulator (Lang et al., 1993). In a standard test series, five 50 ms stimuli (1.5 ms rising and falling ramp) were presented at 250 ms intervals with frequencies ranging between 3 and 46 kHz and intensities between 30 and 90 dB SPL (10 dB increments) from the more sensitive side. Alternatively, ten 3 ms stimuli (0.5 ms rising and falling ramp) were presented at 100 ms intervals using the same frequencies as above and intensities between 30 and 70 dB SPL (5 dB increments). For directional tests, 100 ms stimuli (16 kHz, 2 ms ramps) were presented at 250 ms intervals at intensities ranging from 30 to 90 dB SPL with five repetitions from the left side followed by the same number from the right side at any given intensity. Alternatively, the standard test series were presented from either side. The system was calibrated with a continuous sine wave using a Bruel & Kjaer amplifier (2610) and Bruel & Kjaer microphones (0.5 inch or 0.25 inch). A dB SPL value of a stimulus therefore refers to a continuous tone of the same amplitude. Repeated measurements gave an accuracy of 2 dB.

Digitised data were evaluated using the NEUROLAB program (Hedwig and Knepper, 1992). Threshold values for excitatory tuning curves were defined as the intensity of sound of the respective frequency eliciting one spike above spontaneous activity in three out of five stimuli. Thresholds of excitatory postsysnaptic potentials (EPSPs) occurring below the spiking threshold were calculated with an accuracy of 5 dB (for series with 10 dB increments) or 2.5 dB (for series with 5 dB increments). The calculated value depended on how many stimuli elicited an EPSP.

This paper deals with a neurone whose soma is found in the seventh abdominal ganglion, that is the first ganglion anterior to the terminal (cercal) ganglion (Fig. 1). This neurone is called AN5-AG7 (the fifth ascending neurone known from bushcrickets with the soma in the seventh abdominal ganglion). For practical reasons, it will be designated AN5 in the following. The ascending axon of AN5 passes all free abdominal ganglia on the side contralateral to the soma and gives off a very few branches with a beaded appearance in each ganglion. In the metathoracic and mesothoracic ganglia, arborizations are also relatively sparse and have a beaded appearance. The densest arborizations are found in the prothoracic ganglion, where they terminate in most areas of the auditory neuropile (=medial ventral association centre; Pflüger et al., 1988) with a concentration in the posterior portion (Fig. 1). Auditory sensory cells which have characteristic frequencies between 24 and 34 kHz terminate in this posterior area of the neuropile. The ascending axon enters the brain dorsally and terminates with many endings in the ventral frontal half of the protocerebrum. No differences in the morphology of AN5 were found between males and females.

All recordings were made in the dendritic region of the prothoracic ganglion in one of the thicker branches. All data presented in the figures are from males; however, no differences were found in physiology between males and females. Spontaneous activity in AN5 often followed penetration. In a considerable number of penetrations, the spontaneous activity increased continuously until all spiking eventually ceased and only graded potentials and abortive action potentials were seen. Such recordings were not used for physiological data. In the remaining cases, either there was no spontaneous activity from the start or the frequency of spontaneous spikes stayed constant or decreased continuously.

When presented with a model of duetting A. nigrovittata comprising a 16 kHz, 60 dB SPL male song and a 28 kHz, 60 dB SPL female reply 35 ms after the male ‘trigger syllable’, AN5 responded preferentially or exclusively to the female reply (Fig. 2A). The intensities were chosen assuming that the female is at a distance of 0.5–1 m and that the male has a reduced sensitivity during song production (see Brush et al., 1985; Wolf and von Helversen, 1986; Hedwig et al., 1988). The response was usually a burst of several spikes on top of an EPSP. In addition, inhibitory postsynaptic potentials (IPSPs) were observed in some of the recordings during the 16 kHz syllables of the male song. Inhibition became very obvious after signal-averaging or in peristimulus time histograms (PSTHs) and appeared as a suppression of spontaneous spiking during the male song (Fig. 2A). Often the first syllable of the male song elicited some excitation (seen as a small EPSP in the averaged recording shown in Fig. 2A); this EPSP was sometimes large enough to trigger spiking. In Fig. 2A, both the PSTH and the averaged recording include all stimuli of a test series (in which the male song model was always the same, while the female song model was varied in frequency). It becomes obvious that the only prominent excitation takes place immediately after the ‘female reply’. Fig. 2B shows PSTHs of AN5 responses when the carrier frequency of the female response was varied. At 28 kHz, the modal song frequency, a burst of spikes was elicited, with the initial spike frequency exceeding 300 Hz. A low frequency (16 kHz) elicited no spikes, while a high frequency (38 kHz) elicited a shorter burst of spikes with a lower maximum spike frequency than at 28 kHz.

Carrier frequencies between 12 kHz and 42 kHz were tested, and the strongest mean response of AN5 was found between 24 and 28 kHz (Fig. 2C). A steep decline in spike number in response to low frequencies and a decline with much more interindividual variability in response to higher frequencies were observed. A superposition of the mean response function to varied song frequency and of the behavioural threshold for male phonotaxis (Dobler et al., 1994) shows a close correlation: the frequencies with the lowest behavioural threshold correspond to the frequencies eliciting the strongest response of AN5 (Fig. 2D).

Responses of AN5 to series of short clicks at a frequency of 28 kHz (imitating a female reply) and increasing in intensity show a peaked curve (Fig. 3A) with a steep increase in response strength at low intensities. A minimum in the latency of the response is found at the same intensity as that eliciting the strongest response. This indicates that the decrease in spike number that occurs at higher intensities is due to an inhibition that reduces spike number and increases latency. Only at the highest intensity tested in this programme did the response strength increase again and latency decrease. The latencies of action potentials varied considerably among individuals and were always clearly longer than the latencies of EPSPs. Thus, EPSP latencies averaged 13.9±0.7 ms (mean ± S.E.M., range 12.3–18.0 ms, five males) at 28 kHz, 70 dB SPL, while spike latencies of the same individuals in response to the same stimuli were 18.2±0.6 ms (range 16.1–19.4 ms). The shortest EPSPs started 12 ms after stimulus onset. The mean spike latency of auditory receptors tuned to 24–34 kHz was 12.4±0.4 ms (mean ± S.E.M., range 10.3–14.9 ms, 28 kHz, 70 dB SPL, 14 cells). A scan through carrier frequencies between 12 kHz and 38 kHz at various intensities (Fig. 3B) shows peaks of responses at 24 kHz or 28 kHz at all intensities (except for 35 dB SPL, where the peak response was at 20 kHz). The response tuning is comparatively narrow and broadens somewhat only at 70 dB SPL.

Frequency threshold curves derived from these data show a minimum between 20 and 30 kHz (Fig. 3C, in comparison with male behavioural tuning). Since clear EPSPs without action potentials were often visible at lower intensities, thresholds for both EPSPs and action potentials were determined independently. The dependence of responses upon frequency can be divided into two distinct portions. At frequencies above 20 kHz, EPSP thresholds and spike thresholds are identical and closely match the behavioural threshold. Below 24 kHz, however, there is a clear difference between EPSP threshold and spike threshold (averaging 20 dB). The EPSP thresholds lie below the behavioural thresholds while the spike thresholds lie above, but all three functions run nearly parallel between 12 and 20 kHz.

All stimulation programmes described so far were presented from the side on which AN5 was penetrated, which is the soma-contralateral side. This side was chosen because the direction of the stimulus affects the responses of AN5. First, soma-contralateral stimulation at 28 kHz resulted in a 10–15 dB lower threshold than soma-ipsilateral stimulation (Fig. 4A). Second, with the type of stimuli used for this test (100 ms, 28 kHz), there was a steady increase in response strength as intensity increased. However, saturation was observed at 90 % response strength with soma-contralateral stimuli and at 60 % with soma-ipsilateral stimuli, indicating the existence of soma-ipsilateral inhibition. This effect is seen even more clearly using the same test, but with short clicks as stimuli (Fig. 4B). As described previously, these short stimuli result in a peaked intensity/response function with soma-contralateral stimuli (Fig. 3A). In contrast, soma-ipsilateral stimulation evoked only very weak spiking responses at all intensities tested. This result can only be explained by inhibition.

The existence of such a direction-dependent inhibition was directly demonstrated by means of lesion experiments. Fig. 5A shows averaged recordings of graded potentials (action potentials removed) in response to a white-noise stimulus from the soma-ipsilateral side. The stimulus elicited a prominent EPSP before the soma-contralateral input was cut; after this was cut, an IPSP predominated. However, some data from lesion experiments clearly demonstrated that the soma-ipsilateral ear provides not only inhibition but, depending on frequency, also some excitation. When tests were made at 24 kHz at various intensities, strong EPSPs were seen prior to lesioning, while a prominent EPSP occurred at low intensities (40 dB SPL) and a combined EPSP/IPSP at higher intensities (70 dB SPL) after lesioning (Fig. 5B). The low-intensity EPSPs with soma-ipsilateral stimulation before and following the lesion had approximately the same latency, indicating similar neuronal pathways. In summary, an effective soma-ipsilateral (direction-dependent) inhibition exists, but may be counteracted at low intensities by excitation from the same side.

Female A. nigrovittata respond to a male song with a remarkably constant latency of 30–40 ms after a male trigger syllable (Heller and von Helversen, 1986). Fig. 6A shows that the response of AN5 depends on the latencies of artificial female replies (60 dB SPL) following between 5 and 85 ms after the trigger syllable in a male model song (16 kHz, 70 dB SPL). Obviously, short latencies resulted in reduced responses of AN5, while latencies greater than 25 ms evoked responses of relatively constant strength. This is true for spike number per stimulus as well as for maximal spike frequency (Fig. 6B). The peristimulus time histograms and plots of instantaneous spike frequency shown in Fig. 6C make this effect even clearer. A female reply following 5 ms after the male trigger syllable evoked very few spikes with low spike frequency in both this animal and several other individuals, while a latency of 35 or 85 ms evoked a burst of spikes with a very high initial spike frequency.

Female replies are always very short, a few milliseconds at most. Does AN5 differentiate between a signal duration that compares with a female signal and a longer signal that might, for example, represent the song of a different species or a bat cry? The results presented in Fig. 4A,B indicate little dependence on stimulus duration. Moreover, the variability, which can be estimated from the standard errors, is quite large. A direct comparison of the actual numbers of action potentials elicited by three different programmes with stimulus durations of 3, 50 or 100 ms showed no large differences (Fig. 7A). The intensities shown are those for which there are data from all pulse durations. The mean spike numbers measured between 30 and 50 dB SPL are nearly identical, and only at the highest intensities do shorter stimuli elicit fewer spikes. The maximum mean spike number was approximately two, indicating a phasic response at stimulus onset. The recordings shown in Fig. 7B corroborate this interpretation. The longer stimulus elicited an EPSP that was terminated by a clearly visible inhibition, while no such inhibition was apparent in the response to the short stimulus. As a consequence, the spiking response was the same in both cases, and it can be concluded that AN5 is unable to code stimulus duration.

One prominent feature of the physiology of the AN5 neurone has yet to be mentioned. The responses of AN5 show a strong tendency to decrease in strength during a series of stimuli. In a test programme using short clicks with a repetition rate of approximately 10 Hz, the mean response of eight individuals was highest to the first stimulus and decreased quickly during the first four stimuli to level out during the following stimuli at 20–40 % of the initial response (Fig. 8A). Interestingly, the relative strength of this response decrement did not depend clearly on either intensity or frequency, since stimuli at 28 kHz between 45 and 70 dB SPL and stimuli at 16 kHz, 70 dB SPL all showed the same type of dependence (with the exception of the loudest stimulus at 28 kHz, which elicited the strongest habituation).

A test programme that modelled the song duets showed much lower habituation than other programmes (Fig. 8B). Despite the strong responses elicited by several of the stimuli (e.g. at 24 kHz), very little decrement of response was seen. This is probably because the responses were evaluated at intervals of approximately 950 ms, with additional stimuli at 16 kHz occurring in between (which, however, did not usually evoke excitatory responses, but may have had some dishabituating effect).

It should be noted that at least 20 % of the AN5 neurones from which recordings were made exhibited weak responses that were mostly subthreshold in response to all stimuli, potentially because of very strong habituation.

The general morphology of the AN5 neurone is reminiscent of that of the cercal giant neurones in crickets and other insects (e.g. Mendenhall and Murphey, 1974). These neurones process information derived mainly from wind-sensitive hairs. During communication between crickets, these hairs are stimulated by air currents evoked by wing movements (see, for example, Dambach et al., 1983), and this response may have developed early in evolution. Alternatively, AN5 might be serially homologous to those neurones in the thorax that process information from the chordotonal organs (from which the ear has evolved; Field and Matheson, 1998). In either case, the morphology must have changed dramatically during evolution. From the point of view of its function as an auditory filter neurone in A. nigrovittata, it is most interesting that prominent branches of AN5 are found only in the thoracic ganglia, particularly the prothoracic ganglion, and in the brain. Even without presenting physiological evidence, the fine smooth structure of the thin terminations in the prothoracic auditory neuropile indicate a predominantly postsynaptic function. These projections overlap with, and are concentrated in, the regions of terminations of those auditory receptors that are tuned to the female song frequency. A direct connection to receptors is, therefore, conceivable. The shortest latencies measured in high-frequency receptors are approximately 1.5 ms shorter than the latencies of EPSPs of AN5 in response to the same stimuli. This is of the same order of magnitude as those found in AN1, an auditory interneurone in A. nigrovittata that probably has direct connections to receptors (Stumpner, 1997; Hennig, 1988). The variations in EPSP latencies in AN5, however, are rather large, and spike latencies are considerably longer than those found in AN1. Other methods would, therefore, need to be applied to prove that direct connections to receptors are the standard excitatory input to AN5.

The interneuronal projections are not restricted to the termination sites of those receptors that are tuned to the female song frequency. This may indicate that AN5 receives excitatory input from more receptors than its excitatory tuning implies. A similar observation was made for AN1 (Stumpner, 1997, 1998). It is also possible, however, that these additional projection sites may be regions making contact with other auditory interneurones, e.g. those evoking inhibition in AN5. The protocerebral projections show the same structure as, and similar locations to, those of AN1 and other ascending auditory neurones (Stumpner, 1999). It is possible, therefore, that these structures feed into a network that evaluates information about the song and may, for example, be able to identify that a female reply has the appropriate frequency and temporal relationship with respect to the conspecific male song.

AN5 is predominantly excited by frequencies above 20 kHz and inhibited by lower frequencies. All kinds of stimuli elicit a phasic response. This makes AN5 of A. nigrovittata well suited to respond to a female stridulatory signal, which occurs exclusively in a duet with a male and, in fact, AN5 responds reliably to such a female song. A comparison between the neuronal threshold and the threshold for eliciting male phonotactic behaviour (Dobler et al., 1994), which occurs exclusively in response to a female reply, reveals a close correspondence in the range 24–38 kHz. At lower frequencies, however, the spiking thresholds of AN5 are consistently above the behavioural thresholds, while the thresholds for eliciting EPSPs are below the behavioural thresholds. Above 20 kHz, there is no difference between EPSP threshold and spiking threshold. Provided that AN5 is a necessary element in the neuronal network eliciting male phonotaxis, it is probable that, in a motivated male, a modulation takes place (within the prothoracic ganglion) that reduces some inhibitory influence on AN5 below 24 kHz. So far, no other prothoracic neurone among those identified in A. nigrovittata (four ascending, five local, one T-shaped and five descending; Stumpner, 1996b, 1997; A. Stumpner, unpublished results) with a better correspondence to the male behaviour is known.

In addition to this modulation of low-frequency inhibition, a more general modulation of AN5 responses seems to be indicated by the fact that a considerable percentage of animals had an AN5 neurone that exhibited few or no suprathreshold responses to acoustic stimuli. One might argue that the AN5 cells from which recordings were made in these individuals represent a second neurone type with a similar morphology. However, there is no indication (e.g. from multiple penetrations and staining) that more than one type of AN5 exists.

Artificial female replies 5–15 ms after a male trigger syllable evoke little response in AN5 compared with replies with more natural latencies. This suggests that AN5 is a temporal filter in A. nigrovittata and might be important for the recognition of female song. Unfortunately, behavioural experiments testing the efficiency of various female response latencies on male phonotaxis have yet to be made for this species. However, a ‘time window’ for an effective female reply with respect to its own song has been reported for several related bushcrickets species (Heller and von Helversen, 1986). The constant latencies of female replies in A. nigrovittata strongly suggest that a critical time window also exists in this species. The typical female response latency of 30–40 ms in A. nigrovittata is in good agreement with the ability of AN5 to respond well to artificial replies with a minimum latency of 25 ms.

The female response consists of very short clicks. Such clicks evoke a similar number of action potentials from AN5 as longer stimuli. Thus, AN5 seems to be well-adapted to detect such clicks, but would be unable to discriminate between pulses of different duration. It is not known whether male phonotaxis in A. nigrovittata depends on the duration of the female reply, although in another phaneropterid bushcricket, Leptophyes punctatissima, the probability of male phonotaxis is reduced if artificial female responses are longer than naturally occurring responses. However, in this species, the carrier frequency of the male and female songs is identical, and male song structure is simple compared with that in A. nigrovittata. Therefore, song duration may be a decisive cue for discriminating between males and females in L. punctatissima (Hardt, 1988; Zimmermann et al., 1989).

The AN5 neurone shows greater directionality than auditory receptors. It receives clear and effective inhibition from the ear contralateral to the axon and main prothoracic arborizations ipsilateral to the soma. Surprisingly, some soma-ipsilateral excitation was nevertheless detected following lesion of the soma-contralateral ear. This excitation probably reduces the effect of inhibition at low intensities and seems, therefore, to be counterproductive for directional coding. Its origin may lie in the few arborizations that are regularly found crossing the midline back to the ipsilateral side of the prothoracic ganglion. Soma-ipsilateral excitation has response latencies comparable with those of soma-contralateral excitation. Interestingly, in grasshoppers, some excitation was also detected in most, even highly directional, thoracic auditory interneurones (e.g. Kalmring, 1975; Marquart, 1984; A. Stumpner, unpublished observations). To date, it is not clear what the functional advantage of such excitation might be.

One might argue that sensitivity to ultrasound makes AN5 a good candidate for a bat-detector neurone (even though A. nigrovittata is unable to fly, there are bats that catch sitting prey). This view is supported by the gross morphology of AN5, which makes it conceivable that it has evolved from a cercal interneurone that functions in defence behaviour. There are several reasons, however, why AN5 is unlikely to be involved in predator avoidance against bats. First, AN5 has a very small axon outside the brain (diameter less than 1 μm; low conduction velocity; Stumpner, 1999). This would be unusual for a neurone eliciting defensive responses. The bat-detector neurone AN2 in crickets, for example, has a thicker axon than their AN1 neurone, which elicits positive phonotaxis (e.g. Wohlers and Huber 1985). Even more prominent, giant cercal interneurones initiating wind-induced escape responses are the largest in the connectives of crickets and cockroaches and are at least 20 times thicker than the axon of AN5 (e.g. Mendenhall and Murphey, 1974). Second, AN5 is a strongly habituating neurone, which shows a dramatic reduction of spike numbers to approximately 30 % at a repetition rate of 10 Hz, which is a quite normal value for many bat calls during search flights. Fast repetition rates of 100 Hz, which occur in the final buzzes of bats about to catch a prey, would probably evoke just an initial response by AN5. The same is true for longer calls, such as those of rhinolophid bats, potentially the most dangerous group of bats for non-flying bushcrickets such as A. nigrovittata. Third, since AN5 has a lower sensitivity than auditory receptors above 20 kHz, it does not appear to be a simple high-pass neurone for carrier frequency, responding to all ultrasonic frequencies and being as sensitive as the tympanic nerve (Dobler et al., 1994; Stumpner, 1996a). This, however, is what would be expected for a bat-detecting neurone since bats call with frequencies between 20 and 120 kHz.

In conclusion, AN5 is an interneurone that, in spite of its unusual abdominal origin and gross morphology, seems to function as a typical auditory interneurone (Stumpner, 1999). It is better tuned to the female response signal than any other known interneurone of A. nigrovittata with an ascending axon to the brain. The AN5 interneurone may, therefore, be the ‘counterpart’ to AN1, which is tuned to the male song frequency and may be responsible for the recognition of female song (Stumpner, 1997). Both neurones occur with fundamentally identical properties in males and females. This means that both sexes possess the frequency filters to detect the song of either sex. The AN5 interneurone was also found in related bushcrickets of the genus Barbitistes (A. Stumpner, unpublished observation). In these species, however, male and female songs have similar carrier frequency. The AN5 interneurone in these species would respond to the female reply as it does in A. nigrovittata but, in addition, it would respond to the male song. This would probably make it less well-suited to detect the short female reply in a duet of these species in comparison with that of A. nigrovittata.

I would like to express my gratitude to Norbert Elsner, Kirsten Jacobs, Reinhard Lakes-Harlan, Friederike Lang and Heiko Stölting for many discussions and suggestions. Bettina Jahn, Sabine Meyer, Heiko Stölting and Katrin Wehler helped with the sectioning. Kirsten Jacobs and the editors of JEB helped a lot to improve the English. Supported by the DFG Stu 189/1-1,2.