An Auditory Interneurone Tuned to the Male Song Frequency In The Duetting Bushcricket Ancistrura Nigrovittata (Orthoptera, Phaneropteridae)

Many bushcricket species produce songs with relatively narrow frequency spectra (Heller, 1988). For several species, experimental evidence shows that temporal cues may be used to discriminate between related species (Heller and von Helversen, 1986; Zimmermann et al. 1989; Stiedl et al. 1991; Jatho et al. 1994; Dobler et al. 1994a), but there is still no evidence to indicate that song spectra might be used for discrimination between different species (Jatho et al. 1994). However, it has been demonstrated that flying bushcricket species may show bat-avoidance behaviour in response to ultrasound outside their species-specific song spectrum (Libersat and Hoy, 1991), as has been well established for crickets (Hoy et al. 1989; Hoy, 1992).

Sex-specific song frequencies have been described for the bushcricket Ancistrura nigrovittata (Phaneropteridae), with males producing a song peaking around 12–16 kHz and females responding with a song of their own peaking around 28–30 kHz (Dobler et al. 1994b). The threshold intensities at which female responses to male song models can be elicited correspond well to the frequency spectrum of the male’s song (Dobler et al. 1994b). Females, additionally, show an intensity-dependence of their response activity at a given frequency, with stimuli at higher intensities being less effective than those approximately 20 dB above threshold. Narrow frequency tuning of female behaviour may have several advantages. (1) It might help to discriminate conspecifics from heterospecifics; for example, from Barbitistes species such as B. ocskayi, which produces a similar temporal pattern, but at a higher song frequency, and which may occur syntopically with A. nigrovittata (Heller, 1988; K.-G. Heller, personal communication). (2) It might help to improve the signal-to-noise ratio considerably, since many of the songs of other Orthoptera peak at around 4–10 kHz or above 20 kHz (e.g. Heller, 1988; Huber, 1992; Meyer, 1994). As a consequence, signal detection at low intensities might be improved and therefore communication distance might be increased. The precise detection of male song components is also necessary, since male song and female response are temporally coupled. Males usually produce a song consisting of a group of eight pulses (rate approximately 35 Hz) followed by a single ‘trigger pulse’ after approximately 350 ms. This temporal pattern of the male song is obviously accurately evaluated by the females. Models of male song with modal temporal parameters (i.e. with 7 ms pulses and 22 ms intervals) are most effective in eliciting female responses. These responses always occur with a fixed latency (approximately 30 ms) following the ‘trigger’ pulse.

Short-latency female responses have been described in several phaneropterid species (Heller and von Helversen, 1986; Robinson et al. 1986; Robinson, 1990). It is usually assumed that this acousto-motor reflex does not involve brain neurones because the relevant motoneurones lie in the mesothoracic ganglion (Robinson, 1990). In crickets, in contrast, short-latency movements during negative phonotaxis most probably involve processing in the brain (Nolen and Hoy, 1984). Moreover, the processing of complex temporal patterns in Orthoptera (and hence probably also in A. nigrovittata) seems to depend on neural networks in the brain (Schildberger, 1984; Bauer and von Helversen, 1987). As a consequence, recognition of the male song and triggering of the female response may depend on different neuronal substrates.

In A. nigrovittata, it is usually the male that performs the phonotactic approach, although motivated females also show phonotaxis. The tuning of male phonotactic behaviour is well adapted to the female song spectrum (Dobler et al. 1994b). At the level of the tympanic nerve, males and females show the same hearing thresholds at the behaviourally relevant frequencies (Dobler et al. 1994b).

Some auditory neurones of bushcrickets have been studied (e.g. Rheinlaender, 1975; Kalmring et al. 1979; Boyan, 1984; Römer, 1987; Römer et al. 1988; Fullard et al. 1989). In a portable apparatus for making extracellular recordings, the local ‘Omega neurone’ was used as a biological microphone in the field (Rheinlaender and Römer, 1986; Römer and Bailey, 1990). A ‘T-fibre’ which can readily be recorded in the neck connective has frequently been used as an indicator of hearing ability or directional responses (e.g. Suga and Katsugi, 1961; Rheinlaender and Römer, 1980; Oldfield and Hill, 1983; Chukanov et al. 1985; Bailey and Römer, 1991) but in only a few studies has such an extracellularly recorded cell also been stained (Kalmring et al. 1979; Rheinlaender et al. 1986; Hardt, 1988). An ascending interneurone has been described the activity pattern of which corresponds to the distance of conspecific songs. This neurone may be involved in the spacing behaviour of males (Römer, 1987). Another ascending neurone shows frequency tuning similar to male behavioural tuning (Hardt, 1988). The present paper describes a neurone (termed AN1, see Discussion) with prominent frequency processing most probably contributing to the release of female song in response to calling males in the duetting bushcricket A. nigrovittata.

The investigations were performed using both sexes of the bushcricket Ancistrura nigrovittata (Brunner von Wattenwyl) (Phaneropteridae, Tettigonoidea) from Greece. Animals were captured in the field as nymphs or young adults, or laboratory-reared F₁ generations were used. Most adult animals were separated according to sex; however, rearing conditions had no obvious influence on the response properties of the interneurone.

Preparation and processing of ganglia

An experimental animal was briefly anaesthetized with CO₂ and fixed ventral side up to a plastic holder using a wax–resin mixture. The forelegs were fixed in a normal (inverse) standing position with their tarsi attached to wires, while the coxae were fixed to the body wall. The prothoracic ganglion was exposed and stabilized with a Ni–Cr spoon from below and a steel ring from above. The exposed tissue was covered with Fielden’s saline (Fielden, 1960). Intracellular or quasi-intracellular (McIlwain and Creutzfeld, 1967) recordings were made using thick-walled borosilicate glass capillaries filled with Lucifer Yellow (4–5 % in 0.5 mol l⁻¹ LiCl, Sigma) or with Neurobiotin (5 % in 1 mol l⁻¹ potassium acetate or KCl, Vector). The resistances ranged between 80 and 160 MΩ. The recordings were amplified by a d.c. amplifier, continuously controlled on screen and stored on a digital recorder (Sony). In some experiments, a leg was cut completely in the middle of the femur to disrupt the auditory input through its nerves to the central nervous system. During such cuts, the interneuronal recording was permanently monitored with headphones and on the oscilloscope. Interneurones and auditory receptors were recorded in the ganglion, close to the midline. After physiological characterization, the cell was stained iontophoretically with a constant negative current of 0.5–3 nA (Lucifer Yellow) or by injection of positive current with frequent switching to negative current for stabilization of the recording (Neurobiotin). Details of the treatment of ganglia are given elsewhere (Stumpner et al. 1995; Stumpner, 1996). For histology, ganglia were embedded in epoxy resin (Agar 100), cut into parasagittal sections of 20 μm and (in the case of Neurobiotin stainings) counterstained with Methylene Blue. The outline of the auditory neuropile in the first section lateral to the midline was fitted to a standardized section to obtain a standardized projection area. The orientation in the brain is given as it compares with the developing or dissected nervous system: anterior (frontal) in situ is termed ventral, and posterior is termed dorsal. The results presented in this paper for females are based on 14 Lucifer Yellow and Neurobiotin fills, on six recordings with unequivocal physiological identification of AN1 without staining and on three stained auditory receptors. Data for AN1 in males come from five stained cells and one physiologically identified cell.

The mounted bushcricket was placed in an anechoic chamber with two dynamic speakers (Dynaudio DF 21) to 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 from 3 to 50 kHz and intensities from 30 to 90 dB SPL (sound pressure level). For directional tests, 100 ms stimuli (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. Frequencies used for this program were 16 kHz (male song frequency) and 28 kHz (female song frequency). In tests with model calling songs of different frequencies, a behaviourally effective model song (eight 7 ms pulses, 22 ms intervals, the eighth pulse followed after a 350 ms interval by a single pulse, the next pulse group after another 350 ms interval) was presented at intensities ranging from 30 to 90 dB SPL, each intensity repeated three times. The frequencies used for this test were 8, 16 and 28 kHz. In tests with simplified model songs (no trigger syllable), the duration of either pulses or intervals was varied (see Fig. 5B and Dobler et al. 1994a). The stimuli were built from white noise (2–42 kHz), the intensity used was 70 dB SPL and each intensity was repeated five times. Calibration was achieved using a Bruel & Kjaer amplifier (2610) and Bruel & Kjaer microphones (1/2 or 1/4 inch). Repeated measurements gave an accuracy of ±2 dB.

Digitized data were evaluated using the NEUROLAB program (Hedwig and Knepper, 1992). The threshold value for a tuning curve was defined as the intensity of sound of the respective frequency that elicited one spike above the spontaneous activity level in AN1 in three out of five stimuli. This criterion also allowed calculation of the threshold values at frequencies that did not cause a dramatic change in the total number of spikes compared with spontaneous firing. At frequencies that elicited strong responses in AN1, nearly identical threshold values were obtained using different criteria (e.g. 1 spike instead of 0.6 spikes above spontaneous activity). Thresholds of inhibitory postsynaptic potentials (IPSPs) that occurred below the threshold for excitation (‘subthreshold IPSPs’) were calculated as follows. The first intensity of a standard test series (see above) at which at least four out of five stimuli evoked an IPSP was determined for each frequency, and will be called the ‘first’ intensity in the following. The averaged amplitude of the IPSP at that intensity was calculated for any given frequency. The maximum averaged IPSP amplitude of all frequencies at the respective ‘first’ intensity was then determined. Threshold was defined as the ‘first’ intensity of a given frequency that evoked an averaged IPSP amplitude of less than 50 % of the maximum averaged IPSP amplitude. At those frequencies at which the IPSP amplitude was 50 % or more of the maximum IPSP amplitude, the threshold was defined to be 5 dB lower than the ‘first’ intensity.

The behavioural data shown in Fig. 3 are either modified from Dobler et al. (1994b) or taken from the same data set, but instead of showing the means, as in Dobler et al. (1994b), the response curves of individuals are given here. For details of the procedures, see Dobler et al. (1994a,b).

The neurone described in this paper will be termed AN1 (ascending neurone 1; see Discussion). It has a soma in an anterior dorsal region of the prothoracic ganglion, close to the entry of the neck connectives (Fig. 1A,B). The primary neurite runs in a medial ventral direction and crosses the midline next to the posterior dorsal end of the anterior ring tract, close to or within dorsal commissure III (DCIII, after Lakes and Schikorski, 1990). There are no dendrites of soma-ipsilateral origin. In the soma-contralateral hemiganglion, the neurite thickens and gives off 4–6 major branches with an anterior medial ventral orientation (Fig. 1). One or two further branches may have a more lateral and posterior orientation. The origin and shape of the main branches may vary (Fig. 1C). An ascending axon extending into the soma-contralateral neck connective branches off caudo-laterally at the origin of the larger dendrites. The fine dendrites generally have a smooth appearance and project into the medio-ventral association centre (auditory neuropile) close to the midline (Fig. 1D; see also Römer et al. 1988, termed anterior intermediate sensory neuropile by them). The dendritic endings are most dense in the medio-ventral portion of the auditory neuropile, a region into which receptors with best frequencies (=lowest thresholds) between 12 and 20 kHz project (Fig. 1D). Branches in more dorsal portions of the auditory neuropile, which are the projection areas of receptor cells with best frequencies above 30 kHz (A. Stumpner, unpublished observations), were found occasionally at much lower density. Several dendrites may cross the midline and terminate in the soma-ipsilateral neuropile. The ascending axon passes through the suboesophageal ganglion without branching (Fig. 1A,B). It enters the brain in the dorsal (in situ, posterior) medial region of the circumoesophageal connective and runs straight towards the protocerebrum. All collaterals of this neurone found in the brain lie in the protocerebrum. They extend into ventral (in situ, anterior) portions of the brain, usually originating from two major collaterals, which may differ in their relative thickness. The more caudal branch runs close to the posterior end of the protocerebrum, near a large trachea. Collaterals from the anterior branch form a characteristic circle in the ventral protocerebrum lateral to the mushroom bodies. The branches in the brain have a beaded appearance and are considerably thicker than the fine dendrites in the prothoracic ganglion.

A. nigrovittata has a hearing range from approximately 2 kHz into the ultrasound, with maximum sensitivity around 20–25 kHz at 30 dB SPL (Dobler et al. 1994b). The responses of AN1 are much more frequency-selective than is the whole ear. It is most sensitive at 16–20 kHz, with a steep rise in threshold between 20 and 30 kHz (Fig. 2A). While the threshold values of the tuning curve between 3 and 16 kHz do not vary much between individuals (roll-off 15–25 dB per octave), interindividual variability between 20 and 30 kHz is greater (roll-offs between 20 and 100 dB per octave; Fig. 2A). In all cases, however, at frequencies between 4 and 10 kHz and between 24 and 50 kHz (the highest frequency tested), a prominent inhibition below the excitatory threshold can be seen (Fig. 2B,D). This inhibition is revealed by IPSPs which may dominate the responses of AN1 over more than 20dB, e.g. between 30 and 40kHz. The shapes of the IPSPs differ for low and high frequencies, those at high frequencies falling faster and rising more slowly than those at low frequencies (see averaged IPSPs in Fig. 2E). In both frequency ranges, IPSPs typically show small ‘bumps’ which might indicate the existence of simultaneous excitatory inputs. Between 10 and approximately 20kHz, there are no subthreshold IPSPs. The first stimulus-related activity in this frequency range is excitation (Fig. 2B,D). The thresholds for excitation of AN1 are equal to or lower than the behavioural thresholds for eliciting responses in females (stippled plots for ‘narrow’ and ‘broad’ tuning in Fig. 2B; after Dobler et al. 1994b; S. Dobler, unpublished results). AN1 is consistently more sensitive than female response thresholds around 16–20kHz. This behaviour, like that of AN1, shows a higher variability in the high-frequency range.

AN1 is found in males and females. Fig. 2C shows that the tuning of AN1 in males is very similar to that in females, with slightly lower thresholds at 20 and 24kHz.

Fig. 2D,E demonstrates the occurrence of frequency-dependent IPSPs in AN1 with soma-contralateral stimulation (the more sensitive side of AN1, see below). Lesion experiments were performed by cutting the soma-ipsilateral front leg, and thus its tympanic nerve, to reveal the source of this inhibition. As Fig. 2F demonstrates, the IPSPs, e.g. at 8 kHz, 50 dB SPL and at 38 kHz, 50 dB SPL, remain fully developed even when only the soma-contralateral ear is functioning. The thresholds for excitation and inhibition at various frequencies below and above the best frequency of AN1 (cf. Fig. 2B) increase slightly after the cut (on average by less than 5 dB) as seen in Fig. 2G.

AN1 shows phasic–tonic response patterns which depend strongly on frequency (Fig. 2D,F, see also Fig. 4E). At most frequencies, AN1 shows optimum-type intensity/response functions (Fig. 3). The interindividual variability is large (Fig. 3A), but at each frequency tested between 5 and 24 kHz there is an intermediate intensity which elicits maximal responses (Fig. 3C). The strongest responses are elicited between 8 and 16 kHz (measured in 4 kHz increments), with the maximum lying at 12 or 16 kHz in 10 out of 12 cases. At higher intensities, inhibition, often visible as IPSPs (e.g. at 5 kHz, 90 dB SPL in Fig. 2D), reduces the number of spikes. Stimuli at 20 and 24 kHz still evoke optimum-type response curves, with a much reduced peak response however. At frequencies of 24 kHz or higher, the responses remain weak and are dominated by combined inhibition and excitation (Fig. 2D). The responses of AN1 tested with model songs of varying intensities are similar to those with standard stimuli (data not shown). The responses of AN1 find a parallel in the female response probabilities to male model songs of different frequencies. As seen in Fig. 3B,D, females respond most frequently to frequencies between 12 and 16 kHz and show an optimum-type response function at those frequencies tested at sufficiently high intensities. The variability is larger at high frequencies (e.g. 20 kHz; Fig. 3B) than at lower frequencies (e.g. 12 kHz).

The responses of AN1 to stimuli from the left or the right side are not identical. As shown in Fig. 4A, at 16 kHz the threshold is lower for soma-contralateral stimuli and the maximum response elicited by soma-ipsilateral stimuli is only approximately 60 % of the maximum response elicited by soma-contralateral stimuli. Moreover, the response curve for soma-ipsilateral stimulation has a more saturation-like shape, while an optimum-type curve is found for soma-contralateral stimulation (see also Fig. 3). As a consequence, from threshold up to 55 or 65 dB SPL, soma-contralateral stimulation evokes 50–90 % more spikes than soma-ipsilateral stimulation, which corresponds to a maximum left–right difference of approximately 30 dB. From 70 dB SPL upwards, stimuli from both sides are similarly effective or soma-ipsilateral stimuli may even be more effective. At 28 kHz, the differences between soma-ipsilateral and soma-contralateral stimulation are much smaller and only occasionally exceed 20 % (Fig. 4B). Response latencies at 16 kHz also differ between ipsilateral and contralateral stimulation, but only up to 65 dB SPL (Fig. 4C). The shortest latencies measured for AN1 range between 13.3 and 16.6 ms and are therefore approximately 3 ms longer than the shortest latencies of receptors (10.7–13.4 ms). In receptors and in AN1, latencies are nearly unchanged from 40 dB SPL upwards when stimulation is from the more sensitive side.

The mean interspike intervals for AN1 (in the range 1–10 ms considered) are approximately 5 ms, and this value is considerably larger for soma-ipsilateral stimuli. The shortest intervals were encountered in bursts at the onset of stimuli and were between 2.3 and 3.0 ms (mean 2.7±0.3 ms, S.D., N=8 females).

The obvious asymmetry of responses to soma-ipsilateral and soma-contralateral stimulation indicates the existence of soma-ipsilateral inhibition. In order to demonstrate directly that this reduced response is due to soma-ipsilateral inhibition, lesion experiments were performed by cutting the soma-contralateral front leg and thus severing its tympanic nerve. This cut eliminates all spiking responses while stimulus-coupled inhibitions, seen as IPSPs, become evident (Fig. 4E). The thresholds of these IPSPs are similar to those encountered with soma-contralateral stimulation in the intact animal, but the IPSPs are smaller in amplitude. This inhibition obviously causes the depression of the soma-ipsilateral responses at higher intensities (as seen in Fig. 4A). To test this assumption, the intensity-dependence of the responses to left and right stimuli was tested before and after cutting the soma-ipsilateral leg (Fig. 4F). The results confirm this assumption. The responses to soma-ipsilateral stimuli are dramatically changed after the cut, becoming similar to the responses to soma-contralateral stimuli. There still is some difference between the responses of AN1 to right and left stimuli, which may be explained by the fact that, with soma-contralateral stimuli, all AN1s respond best to 50 dB SPL, while with soma-ipsilateral stimuli the response maximum is at 60 or 70 dB SPL.

Auditory receptors with best frequencies around 16 kHz show side-dependent differences in their responses at 16 kHz (Fig. 4D). Responses to contralateral stimuli have an approximately 20 dB higher threshold than responses to ipsilateral stimulation. Therefore, the relative response differences of receptors to left and right stimulation at 16 kHz, 30–50 dB SPL are not much smaller than those found in AN1. Furthermore, for receptors and AN1, stimuli of 70 dB SPL or greater are similarly effective, whether presented to the left or the right of the animal. The reasons for this equalisation, however, differ between the receptors and AN1. In the receptors, it is response saturation that equalises the responses to left and right stimuli at high intensities, whereas in AN1 it is the combined effect of inhibitions originating in the soma-contralateral and in the soma-ipsilateral ear.

Female A. nigrovittata evaluate not only the frequency content of a model song but also its temporal pattern (see Introduction and Dobler et al. 1994a). If one compares the spike trains which AN1 produces in response to song models of various frequencies, it appears that the spike patterns in response to modal song frequencies (12–16 kHz) show a tighter coupling to the stimulus pattern than those at other frequencies (Fig. 5A). This dependence of AN1 responses might add to the frequency-dependent differences in threshold and response magnitude described above. It also raises the question of whether AN1, in addition to its frequency-selectivity, might contribute to the temporal processing of song models. To deal with this question, song models with either pulses or pauses between pulses of varied duration were presented with white noise, 70 dB SPL, as in the behavioural tests. The parameters not varied were kept at the behaviourally most effective value: 7 ms pulse duration and 22 ms pause duration. Pulses between 3 and 15 ms and pauses between 5 and 50 ms elicited at least half-maximal behavioural responses. The number of spikes produced by AN1 decreases with increasing pause duration (Fig. 5B, nos 1–3), but the interindividual variability is relatively high. In contrast, the number of spikes remains more or less constant with increasing pulse duration (Fig. 5B, nos 2, 4–6), and interindividual variability is less pronounced. The histograms in Fig. 5B demonstrate that a temporal patterning of spikes may occur with, in some cases, a clear correlation to the stimulus pattern. However, there are no distinct differences between behaviourally effective (stimulus no. 2) and behaviourally ineffective (e.g. stimulus no. 4) song models. The responses to single pulses within a stimulus are variable (see peri-stimulus-time histograms in Fig. 5B), and the spike train encodes the stimulus pattern less precisely than, for example, do auditory receptors.

This paper presents the first complete morphological identification of an ascending auditory neurone, termed AN1, in a phaneropterid bushcricket. As judged from the smooth appearance of its prothoracic branches, the input sites lie in the soma-contralateral prothoracic hemiganglion. This is also the side with the main excitatory and frequency-dependent inhibitory input, while the soma-ipsilateral side seems to provide only inhibitory (‘directional’) inputs to AN1. The beaded appearance of branches in the dorsal and frontal protocerebrum indicates that major output sites can be found here.

AN1 of A. nigrovittata is most sensitive to frequencies around 16–20 kHz. The excitation may originate directly from auditory receptors, since AN1 and receptors most sensitive to 16 kHz show extensive overlap in the ventral portion of the auditory neuropile. The shortest latencies to the first action potential in AN1 and receptors differ by 2–3 ms, which would be in agreement with a direct connection (Hennig, 1988; Römer et al. 1988). AN1 shows optimum-type intensity/response functions and receives marked subthreshold inhibition below 10 kHz and above 20–24 kHz, apparently shaping its narrow tuning curve. Low-frequency and high-frequency inhibitory inputs evoke IPSPs of different shapes, which therefore seem to be caused either by different presynaptic elements or by one presynaptic element exhibiting different responses at different frequencies. These elements need yet to be identified.

Does this information about AN1 suffice to identify possible homologues in other ensiferans? Ascending neurones have been described morphologically in five bushcricket species: Mygalopsis marki (Römer, 1987), Tettigonia viridissima (Marquart, 1984; Römer et al. 1988; Schul, 1997), T. cantans (Hardt, 1988), Leptophyes punctatissima (Hardt, 1988) and

Decticus albifrons (Nebeling, 1994). The only complete staining (reconstruction after vibratome sectioning and anti-Lucifer immunohistochemistry) of an ascending neurone in a bushcricket is for D. albifrons (Nebeling, 1994). All these neurones have similar gross morphology in the prothoracic ganglion.

The ascending neurone described in L. punctatissima is tuned to the conspecific song frequency (40 kHz). The dendritic branches of this neurone project primarily into the posterior dorsal portion of the auditory neuropile, where 40 kHz receptors terminate (Hardt, 1988). Therefore, the projections of the neurone described in L. punctatissima and those of AN1 of A. nigrovittata are morphologically and functionally different. Also, no frequency-dependent inhibition was found in the ascending neurone of L. punctatissima. Therefore, it seems unlikely that these two neurones are homologous.

In other tettigoniid species, two ascending neurones have been described (M. marki, Römer, 1987; T. viridissima, Römer et al. 1988; Schul, 1997). In both species, one of the two neurones shows response properties (e.g. frequency-dependent inhibition, optimum-type intensity/response functions) which compare well with those of AN1. The neurone is termed AN1 in T. viridissima and is an unnamed neurone in Figs 6D, 7 and 8D of Römer (1987) in M. marki. The ascending neurone described for T. cantans (Hardt, 1988) is probably homologous to this neurone in T. viridissima. These neurones have (nearly) exclusively soma-contralateral dendrites, and these dendrites terminate mainly in the medial ventral portion of the auditory neuropile, as in AN1 in A. nigrovittata. There is little doubt that AN1 of A. nigrovittata is homologous to AN1 of T. viridissima as well as to the unnamed neurones of T. cantans and in M. marki.

Crickets, like bushcrickets, have their ears in their forelegs, and it is generally believed that the ears and therefore the auditory system as a whole descend from a common progenitor, although this has recently been called into question (Gwynne, 1995). Nevertheless, there is a striking similarity of auditory interneurones at the prothoracic level which led to the identification of homologous elements such as Omega neurone 1 (Wohlers and Huber, 1978; Römer, 1985) and AN1 (Schildberger, 1994). For AN1, the principal organization at the prothoracic level is very similar in crickets and bushcrickets, despite variations in the organization of the auditory neuropiles (Wohlers and Huber, 1985; Römer et al. 1988). In both groups, AN1 seems to have direct connections with afferents (Hennig, 1988; Römer et al. 1988), and in both groups AN1 is more sensitive to low frequencies (e.g. Schildberger et al. 1986; Hennig, 1988; Stout et al. 1988; Römer, 1987; Schul, 1997; this study) and receives frequency-dependent inhibition, at least in the high-frequency range (e.g. Schildberger, 1984; Stumpner et al. 1995; Römer, 1987; Schul, 1997), as well as soma-ipsilateral inhibition (Horseman and Huber, 1994; Stumpner et al. 1995, this study). Brain projections of AN1 in Gryllus bimaculatus show a quite similar pattern and location to those described here for A. nigrovittata. A similar pattern, however, is also described for AN2 of G. bimaculatus and therefore does not help to support the potential homology to AN1 and AN2 in crickets. All in all, the homology of AN1 in crickets and AN1 in A. nigrovittata seems to be a reasonable hypothesis according to the data available.

What are the proposed functions of potentially homologous neurones to AN1 in the different bushcricket species? In M. marki, this neurone could account for the spacing response of males, which use intensity as an important cue for spacing (Römer and Bailey, 1986). In T. viridissima, AN1 is excited by the low-frequency component of the broad-banded song (Schul, 1997) and, in combination with other neurones, could serve as an indicator of the distance to a potential mate (see Latimer and Sippel, 1987).

In crickets, AN1 is important for determining positive phonotaxis (Atkins et al. 1984; Schildberger and Hörner, 1988) and apparently carries information about frequency, song pattern and sound direction in its tonic responses (Schildberger, 1984; Stout et al. 1988; Stabel et al. 1989; Horseman and Huber, 1994).

In A. nigrovittata females, the potential contributions of AN1 to three different aspects of acoustic communication behaviour should be considered.

Female A. nigrovittata respond to models that represent the temporal pattern of a male song and are presented at certain frequencies. These frequencies represent only a fraction of the hearing range of this species (Dobler et al. 1994b). The male song peaks at approximately 15 kHz, while the female responses can be elicited with lowest intensities and highest efficacy at 12 and 16 kHz (Dobler et al. 1994b; see also Figs 2B, 3B,D). A comparison of the responses of AN1 with behavioural data allows the conclusion that AN1 may be part of the frequency-recognizing network in female A. nigrovittata. The thresholds for excitation of the AN1 neurone encompass the thresholds for eliciting behavioural responses in females (Fig. 2B). Female response thresholds vary markedly with stimuli above 12 kHz, but little below 12 kHz. They vary between individuals, but sometimes also within one individual when tested for longer periods (Dobler et al. 1994b; S. Dobler, unpublished results). The thresholds of AN1 vary much more between individuals in response to stimuli above 20 kHz than to those below. Excitation of AN1 is of similar sensitivity to behavioural thresholds at 8, 12 and 28 kHz, but up to 10 dB more sensitive than even ‘broad’ female responses between 16 and 24 kHz. At intensities above threshold, the similarities become even more obvious (Fig. 3). Both the response frequency and neuronal activity of females are greatest for frequencies of 12 or 16 kHz, but much smaller for frequencies of 20 or 24 kHz, despite AN1 being most sensitive at 20 kHz. Both behaviour and AN1 show optimum-type responses to increasing stimulus intensities, with intensities 10–20 dB above threshold being the most effective. This supports the hypothesis of Dobler et al. (1994b) that it is a property of auditory interneurones and not of receptors which brings about this intensity-dependence of behaviour (receptors in A. nigrovittata show saturating intensity/response curves at all frequencies, as has been found in other bushcrickets, Fig. 4D; A. Stumpner, unpublished observations). Since the tuning of auditory receptors becomes broader with increasing intensity (see Stumpner, 1996), the intensity-dependent inhibition of AN1, which is still present when the soma-ipsilateral leg is cut, is most probably evoked by the same interneurones, which cause frequency-dependent inhibition most clearly seen as subthreshold IPSPs. This suggests that the intensity-dependence of the behaviour is an effect of the frequency-dependent inhibition by central processing.

Twelve different auditory interneurones have been identified in the prothoracic ganglion of A. nigrovittata (A. Stumpner, unpublished results). Among these are six neurones with an axon ascending to the brain. These neurones have broad tuning similar to the whole hearing range or are most sensitive in the ultrasound. Most of them respond most strongly at the highest intensities tested, and none of them responds best to 12 or 16 kHz. AN1, in contrast, shows a frequency-dependent excitation that matches the frequency-dependence of female behaviour very well. Some differences are found, however, between AN1 and behaviour in the high-frequency range, where interindividual variability is largest. The most interesting finding in this respect is the variability of female behaviour: one individual can change its response threshold from the narrow tuning curve shown in Fig. 2B to the broad tuning curve shown in the same figure by shifting thresholds in the high-frequency range. This indicates that a high-frequency neurone might modulate frequency processing, having an inhibiting effect on the female response and reducing the excitatory input of AN1 into the recognizing system at 16 kHz or above. Differences between neuronal and behavioural tuning may also be explained by the reduced responses of AN1 to stimulus frequencies of 20 and 24 kHz, compared with those to 8 and 12 kHz, due to efficient inhibition.

The prominent IPSPs recorded in AN1 at low and high frequencies below the excitatory threshold demonstrate frequency-specific processing at the prothoracic level. This processing adds to the frequency-selectivity of the auditory receptors. There are receptors, most sensitive to 12–16 kHz, that demonstrate frequency tuning that matches behaviour (A. Stumpner, unpublished results). The projection area of the dendrites of AN1 in the auditory neuropile and the projections of auditory receptors tuned to frequencies between 12 and 20 kHz (Fig. 1D) make it probable that AN1 receives excitation from more receptors than just those tuned specifically to 12–16 kHz. However, this can only be tested if the frequency-dependent inhibition can be eliminated without also eliminating the excitatory input. This might be possible by pharmacologically blocking transmitter receptors or Cl⁻ channels, e.g. by application of picrotoxin (Marder and Pauperdin-Tritsch, 1978), and measuring excitatory tuning of AN1 prior to and following application.

The responses of AN1 indicate that it is not likely to be involved in extracting or weighting specific features of the temporal pattern of male song by the females. AN1 responds phasic–tonically to stimuli of white noise, 70 dB SPL, the parameters used for the behavioural tests. It does not discriminate obviously or reliably between temporal patterns which have different efficacies in releasing behavioural responses (Fig. 5). Nevertheless, its responses usually exhibit coupling to the temporal pattern of a stimulus (Fig. 5), and this coupled activity may contain some information about the pulse period and, less probably, about the pulse duration. At least some additional information about the song pattern must be transported to the brain by other neurones. However, the male song is composed of a pulsed syllable, which is followed after approximately 350 ms by a single pulse that triggers the female response. This single pulse, in behavioural tests, showed the same tuning as the pulsed syllable (Dobler et al. 1994b). AN1 responds to this single pulse (Fig. 5A) and its shortest latencies are approximately 15 ms shorter than the shortest female behavioural responses. Therefore, AN1 might be the relevant neurone to trigger the response. In this case, however, the information would have to loop through the brain before reaching the forewing motoneurones in the mesothoracic ganglion. This would mean that only very few synapses in the brain could be involved.

The responses of AN1 depend on the direction of the sound. It is excited by sound from the soma-contralateral ear, which is the side where the vast majority of dendritic branches and the ascending axon are found. In contrast, AN1 is inhibited by sound from the soma-ipsilateral ear. This inhibition is sufficient to reduce the activity of AN1 and to evoke asymmetric response curves for ipsilateral and contralateral stimulation. However, in the range of strongest AN1 responses at the male song frequency (40 and 50 dB SPL), the response difference between stimulation from either side is not much stronger than that found in auditory receptors with a best frequency similar to that of AN1. Therefore, this direction-dependent inhibition obviously becomes effective only at higher intensities. At these intensities, the inhibition prevents the left AN1 from becoming more responsive to stimuli from the right than does the right AN1 (and vice versa). This may be important in preventing the nervous system from having to cope with contradictory information, which would be present without such inhibition (see Fig. 4F). In the communication system of A. nigrovittata, with the male singing and the female responding, it is usually the male that performs phonotaxis (which is elicited by the 28 kHz signal of the female, so that male phonotaxis cannot rely on the activity of AN1). For response singing, females do not have to determine the direction of male songs. Females kept as virgins for longer, however, start to perform phonotaxis on their own if a male does not approach when they respond (A. Stumpner, unpublished observations). AN1 presents directional information below 70 dB SPL and may contribute to the orientation behaviour of the females.

AN1 was found in female and male A. nigrovittata with similar response properties (Fig. 2B,C). This means that AN1 cannot be the neurone responsible for recognizing the song of the conspecific partner in both sexes, since such a neurone in the male must be tuned to frequencies around 28 kHz (female song). Is there a probable function for AN1 in males? When two males are singing together, they interact and produce songs comprising only the pulsed syllable without the single trigger pulse. Therefore, some kind of agonistic behaviour may exist among males, and this behaviour may be tuned to the male song frequency just like the female response behaviour is. AN1 would be an appropriate candidate to serve this function in male A. nigrovittata.

My thanks are due to Professor N. Elsner, who supported this study from the beginning and helped in many ways. I thank Dr R. Lakes-Harlan for countless discussions on hearing in bushcrickets, for technical advice and for criticising the manuscript. Dr S. Dobler put the behavioural data in Fig. 3 at my disposal. Her careful behavioural studies and the discussions with her, Dr K.-G. Heller, Dr D. von Helversen and Professor O. von Helversen were necessary prerequisites for initiating these studies. Discussions with Dr J. Schul gave helpful insights in the auditory system of Tettigonia. He and Dr M. Hardt made valuable comments on the manuscript. H. Stölting and M. Schink helped with sectioning. J. Henley and D. Zacharias helped a lot to improve the English. Two anonymous referees made numerous helpful suggestions. Supported by the DFG Stu 189/1-1.