A Corollary Discharge Mechanism Modulates Central Auditory Processing in Singing Crickets

J.F.A. Poulet, B. Hedwig


Crickets communicate using loud (100 dB SPL) sound signals that could adversely affect their own auditory system. To examine how they cope with this self-generated acoustic stimulation, intracellular recordings were made from auditory afferent neurons and an identified auditory interneuron—the Omega 1 neuron (ON1)—during pharmacologically elicited singing (stridulation). During sonorous stridulation, the auditory afferents and ON1 responded with bursts of spikes to the crickets' own song. When the crickets were stridulating silently, after one wing had been removed, only a few spikes were recorded in the afferents and ON1. Primary afferent depolarizations (PADs) occurred in the terminals of the auditory afferents, and inhibitory postsynaptic potentials (IPSPs) were apparent in ON1. The PADs and IPSPs were composed of many summed, small-amplitude potentials that occurred at a rate of about 230 Hz. The PADs and the IPSPs started during the closing wing movement and peaked in amplitude during the subsequent opening wing movement. As a consequence, during silent stridulation, ON1's response to acoustic stimuli was maximally inhibited during wing opening. Inhibition coincides with the time when ON1 would otherwise be most strongly excited by self-generated sounds in a sonorously stridulating cricket. The PADs and the IPSPs persisted in fictively stridulating crickets whose ventral nerve cord had been isolated from muscles and sense organs. This strongly suggests that the inhibition of the auditory pathway is the result of a corollary discharge from the stridulation motor network. The central inhibition was mimicked by hyperpolarizing current injection into ON1 while it was responding to a 100 dB SPL sound pulse. This suppressed its spiking response to the acoustic stimulus and maintained its response to subsequent, quieter stimuli. The corollary discharge therefore prevents auditory desensitization in stridulating crickets and allows the animals to respond to external acoustic signals during the production of calling song.


A key question in sensory neuroscience is how animals deal with self-generated, or reafferent, sensory information. It has been suggested that a neural signal, an “efference copy” (Holst and Mittelstaedt 1950) or “corollary discharge” (Sperry 1950), from the motor network is used to anticipate and cancel out reafferent responses. In this way animals could prevent reafferent information from desensitizing their own sensory pathway and/or from being confused for external stimuli.

A modulation in the sensitivity to reafferent stimulation by centrally generated neural signals has been identified in a variety of sensory systems, e.g., visual (Zaretsky and Rowell 1979), electroreceptive (Bell 1981, 1982; Bodznick et al. 1999), proprioceptive (Gossard et al. 1991;Sillar and Skorupski 1986; Wolf and Burrows 1995), and mechanoreceptive (Blakemore et al. 1998; El Manira et al. 1996; Murphey and Palka 1974; Sillar and Roberts 1988). A reduction in the responsiveness of auditory neurons in the brain has been recorded in humans (Creutzfeldt et al. 1989,Numminen et al. 1999) and other vertebrates (Kirzinger and Jürgens 1991; McCasland and Konishi 1981; Metzner 1989, 1993;Müller-Preuss and Ploog 1981; Schuller 1979; Suga and Schlegel 1972; Suga and Shimozawa 1974) during vocalization, but the nature and source of the inhibition has never been characterized. Here we take advantage of the simple auditory circuit of a cricket to analyze how this acoustically communicating insect deals with self-generated auditory stimulation.

Male crickets (Gryllus bimaculatus) sing for hours at over 100 dB SPL (Nocke 1972) to attract females. Songs are generated by rhythmically rubbing the forewings together: a form of sound production called stridulation. As crickets' ears are located on the forelegs, they are fully exposed to the self-generated sounds. Many animals reduce the responsiveness of their peripheral auditory system during sound production (Borg and Counter 1989;Hennig et al. 1994; Narins 1992;Suga and Jen 1975), but crickets do not (Poulet and Hedwig 2001). Despite this, behavioral experiments have shown that singing crickets can respond to external sounds (Heiligenberg 1969; Jones and Dambach 1973). To examine how crickets' central auditory system deals with reafferent auditory information, we made intracellular recordings of auditory afferents (Eibl 1978; Michel 1974) and an identified auditory interneuron—the Omega 1 neuron (ON1) (Casaday and Hoy 1977; Popov et al. 1978; Wohlers and Huber 1978)—during stridulation. A summary of part of this work has been published (Poulet and Hedwig 2002).



All experiments were performed on adult male G. bimaculatus selected from a cricket colony maintained on a 12 h light:12 h dark cycle. Prior to dissection they were chilled at 4°C for ≤30 min. They were then fixed in a standing position on a holder that allowed free rotation of the animal (Fig.1 A). To allow for silent stridulation, the left wing of the crickets was removed. When recording fictive stridulation, the cricket was placed upside down in a Plasticene well and its ventral cuticle was removed to expose the abdominal and thoracic ganglia. The thoracic or thoracic and abdominal nerves were cut, except for prothoracic nerve 5, which contains the auditory afferents. Care was taken not to damage the main ventral trachea. To deafen crickets, the forelegs were removed just distal to the coxa. Experiments were done at room temperature (18–22°C).

Fig. 1.

Experimental setup. A: the cricket was held upside down with its wings free to move. A microphone detected sounds produced by the cricket, while an optoelectronic camera recorded wing movement. The stabilizing ring that normally rests on top of the prothoracic ganglion has not been shown to aid clarity. B: a syllable, a whole chirp, and a chirp interval are marked below a typical sound recording of pharmacologically elicited calling song. Arrow 1 indicates the quiet sound produced during opening wing movements.Wing: stridulation wing movements; Sound: microphone recording.

Pharmacological stimulation

The antennae and a piece of head cuticle were removed to reveal the brain for pharmacological stimulation. Glass capillaries, filled with the acetylcholine esterase inhibitor Eserine Salicylate (10−2 mol l−1 in insect saline), were connected to a WPI PV 820 pneumatic Pico Pump and positioned in the anterior protocerebrum of the brain. Having located the auditory neuron for intracellular recording, a 0.3-s pulse of Eserine (about 1 nL) was injected into the anterior protocerebrum to elicit singing (Otto 1978; for details of pharmacological injection techniques, see Wenzel and Hedwig 1999). Injection of acetylcholine and cholinergic agonists into the cricket's anterior protocerebrum activates the stridulatory command neurons, which spur the thoracic central pattern generator into action (Hedwig 2000a).

Acoustic stimulation

Auditory stimuli were generated by the software Cool Edit 2000 running on a Toshiba Laptop (Satellite 4010) and delivered from two piezo-electric speakers through brass tubes with frontal openings of 14 mm which were positioned 20 mm from the posterior tympanic membrane. Sound tubes were used to avoid echoes associated with free field stimulation. All acoustic stimuli had a carrier frequency of 4.5 kHz and rising and falling ramps of 2 ms. Stimuli either had a duration of 8 ms, a period of 15 ms, and an intensity of 75 dB SPL re. 20 μPa root-mean-square (RMS) or, when mimicking cricket calling song, they had a duration of 21 ms, a period of 42 ms, and an intensity of 100 dB SPL re. 20 μPa (RMS). Sound amplitude was calibrated using a Bruel and Kjaer amplifier (Type 2610) and a ½ in. microphone (Type 4191) positioned at the ears of the cricket.

Neuron recordings

For intracellular recording, the holder with the cricket attached was rotated so that the crickets were upside down, and the prothoracic ganglion was exposed (Fig. 1 A). A pair of forceps, which acted as the indifferent electrode, was modified so that one tine could be placed under the ganglion to act as a platform, while the other was a stabilizing ring that could be lowered onto the ganglion. Exposed nervous tissue was bathed in insect saline (ionic composition in mmol l−1: 140 NaCl; 10 KCl; 4 CaCl2; 4 NaHCO3; 6 NaHPO4). Recordings and stainings of auditory neurons were made in the auditory neuropil of the prothoracic ganglion using thick-walled glass micropipettes filled with 5% Lucifer yellow and LiCl at a resistance of 100–150 MΩ. Auditory afferents were recorded close to their terminal branches, whereas ON1 was recorded in its dendritic region. Neurons were held slightly hyperpolarized during an experiment to stabilize the recording. After the experiments, ganglia were dehydrated in an alcohol series and cleared in methyl salicylate, and stained neurons were identified under a UV fluorescence microscope.

Behavioral recordings

Sonorous singing was monitored with a microphone (Audio-Technica, AT853A) positioned 5 cm from the wings. Wing movements were measured with an optoelectronic camera focused onto a reflective disk (2 mm diam, 3M Scotchlite 7610) glued to the right wing (Hedwig 2000b). Motor activity during fictive stridulation was recorded with a suction electrode placed on mesothoracic nerve 3A, which contains motor axons that innervate wing closer and opener muscles (Kutsch and Huber 1970).

Data sampling and analysis

All data were sampled on-line at a rate of 10 kHz per channel transferred directly onto a computer, via a high-speed Data Translation AD board (DT 2821 F8DI). Recordings were analyzed off-line using Neurolab software (Knepper and Hedwig 1997) and Microsoft Excel 2000. Wing movement, mesothoracic nerve 3A activity, or the acoustic stimuli were used as a temporal reference when averaging data (time = 0 ms in the figures). Microphone recordings were full-wave-rectified before averaging. Data values are given as mean ± SE.


Pharmacologically elicited stridulation

Natural calling song consists of a series of chirps, 250 ms long, separated by a 300-ms chirp interval (Doherty 1985). Each chirp is composed of four 21-ms syllables that are produced during the closing wing movement and increase in intensity throughout the chirp (Fig. 1 B). Pharmacological stimulation normally elicited calling song with very similar temporal and acoustic properties to the natural song (Otto 1978; Wenzel and Hedwig 1999). One difference was the appearance of very quiet sounds during wing opening (arrow 1, Fig. 1 B). These were probably a result of the animals singing upside-down. Stridulation started roughly 10 min after injection of Eserine Salicylate into the brain. Sonorous stridulation lasted for 10–15 min, whereas fictive stridulation lasted for 30–45 min. Most intact animals had several pauses in their singing, which lasted for several minutes. Periods of fictive stridulation, however, continued uninterrupted for a few minutes longer.

Activity of auditory neurons during sonorous singing

About 60 auditory afferents project from each ear along the dorsal part of prothoracic nerve 5 and terminate in the auditory neuropile of the prothoracic ganglion (Eibl 1978; Michel 1974). Two mutually inhibitory (Selverston et al. 1985), local, auditory interneurons—the Omega 1 neurons (ON1)—are located within the prothoracic ganglion and have dendritic branches that overlap with the output branches of the auditory afferents (Casaday and Hoy 1977; Popov et al. 1978; Wohlers and Huber 1978). These identified interneurons receive auditory information from the ear ipsilateral to their soma and are most sensitive to a sound frequency of 4.5 kHz, the carrier frequency of male calling song (Casaday and Hoy 1977; Popov et al. 1978; Wohlers and Huber 1978).

During sonorous singing, auditory afferents were activated by the sound (Fig. 2 A). The afferents responded with bursts of spikes both to the loud syllables produced during wing closing (asterisks in Fig. 2 Ai) and to the quieter sound during wing opening (arrow 1 in Fig. 2 Ai). This is clearly demonstrated by the quantitative evaluation, which shows the peristimulus time (PST) histogram overlaid with the average spike rate (top), the average wing movement (middle), and the sound pattern (bottom) during the sonorous chirps (Fig. 2 Aii). Loud sound pulses occurred in-phase with the closing movements and caused the large maxima in the PST histograms and spike frequency plots (asterisks in Fig.2 Aii) that occurred at the transition from wing closing to opening (dashed line Fig. 2 Aii). The softer syllables, in-phase with the opening movement, caused only weak afferent activity (arrow 1 in Fig. 2 Aii). On average, the afferents reached a maximum spike frequency of 246 ± 24 Hz (n = 6 afferents) during sonorous chirps. During some chirps, with a silent initial wing opening movement, a depolarization of the afferent membrane potential was evident prior to the spikes (arrow 2 in Fig.2 Ai). Since these recordings were obtained close to the terminals of the afferents, such a depolarization could indicate the presence of a primary afferent depolarization (PAD).

Fig. 2.

Activity of auditory neurons during sonorous singing.Ai: auditory afferents responded to the loud (asterisk) and soft (arrow 1) syllables produced during the chirp. A primary afferent depolarization (PAD) preceding the spiking response was sometimes present (arrow 2). The spikes in Ai have been truncated. Aii: the peristimulus time (PST) histogram and the overlaid, averaged spike frequency (top) clearly show that the response of the afferent is in-phase with sound production. A dashed line indicates the timing of the maximum afferent response in relation to the averaged wing movement (middle) and the rectified sound recording (bottom). Bi: the Omega 1 neuron (ON1) spiked in response to the initial quiet sound produced by opening wing movement (arrows 3). Bii: it generated bursts of spikes following each loud syllable (asterisks) with a latency of 15–20 ms, as indicated by the dashed line between the PST histogram and the wing movement. Symbolized crickets above the figures represent 2-winged, 1-winged, or fictive singing with or without ears (see following figures). ON1: intracellular ON1 recording;Afferent: intracellular recording of a primary auditory afferent neuron.

ON1 was also activated during sonorous singing with rhythmic membrane potential oscillations in-phase with the syllables (Fig.2 B). At the beginning of a chirp, the first one to two spikes were elicited by the quiet sound caused by the initial opening wing movement (arrow 3 in Fig. 2 Bi). Further spike activity then occurred only in response to the loud syllables produced during the closing wing movements (asterisk in Fig. 2 Bi). ON1's 15- to 20-ms response delay time meant that these maxima occurred during the subsequent opening wing movements (dashed line in Fig.2 Bii). The response latency of ON1 to the first quiet sound at the start of a chirp is about 20 ms. The response to the following louder syllables has a shorter latency because the cell is already depolarized by the preceding quieter sound that moves the membrane potential nearer to spiking threshold. The quantitative evaluation of ON1's response demonstrated that the average ON1 maximum spike frequency to self-generated syllables was 176 ± 28 Hz (n = 10 crickets). Immediately after ON1's response during the chirp, it was slightly hyperpolarized and during the chirp interval the membrane potential gradually repolarized. These experiments suggest that the afferents and ON1 respond to reafferent sound. To test this hypothesis, recordings were made in crickets with one wing removed so that they stridulated silently.

Responses of auditory neurons during silent stridulation

During silent stridulation, the afferents only occasionally generated action potentials, indicating that their high-frequency spiking during sonorous stridulation was indeed a response to the cricket's own song. With the major source of excitation removed, PADs were now evident in all of the 13 auditory afferents recorded (Fig. 3, A,B, and C). During silent stridulation, PADs were recorded both in afferents which responded best to 4.5 kHz (n = 12 crickets) (Fig. 3, A andC) and in an afferent that responded best to 10–30 kHz (Fig. 3 B). In four of these crickets, the average duration of the opening to closing wing movement cycle underlying the production of syllables was 36.0 ± 1.8 ms. The PADs started 4.2 ± 0.4 ms after the start of wing closing and reached a maximum at 19.3 ± 1.4 ms, just after the transition between wing closing and opening at 17. 8 ± 1.0 ms (n = 4 crickets) (dashed lines in Fig. 3, Aii, Bii, Cii). This corresponds to the phase of maximum response to reafferent sound. On average, in six crickets, the PADs increased in size from 2.6 ± 0.5 mV (measured from chirp interval baseline membrane potential to peak amplitude), at the start of the chirp (arrow 1 in Fig.3 B), to 3.3 ± 0.7 mV at the end. Although wing movement often decreased in amplitude during singing, the timing and amplitude of the PADs were not affected, indicating that they did not depend on the intensity of sensory feedback associated with wing movements (Fig. 3, B and C).

Fig. 3.

Responses of auditory afferents during silent 1-winged stridulation with (Ai) normal and (Bi,Ci) small wing movements. PADs were present in low- (Ai, Ci) and high-frequency (Bi) auditory afferents. They started just after the start of the closing wing movement and reached a maximum at the beginning of the consecutive wing opening. Spikes were occasionally generated during silent chirps. On average, the PADs increased in size during the chirp; arrow 1 (B) shows an example of a small PAD at the start of a chirp. PADs were compound potentials resulting from the summation of many small amplitude potentials. Arrow 2 (Ai) indicates a PAD where the small-amplitude depolarizing potentials were obvious. Aii,Bii, Cii: superpositions of the afferent recording (top traces), triggered by the onset of the wing movement (bottom traces), demonstrate the time course of the PADs in relation to the average wing movement. The peak PAD amplitude occurred just after the transition between wing closing and opening, as indicated by the dashed lines.

ON1 only occasionally spiked during silent stridulation. Inhibitory postsynaptic potentials (IPSPs) occurred in all 21 ON1s recorded that started and built up in-phase with the closing wing movements of each syllable (Fig. 4 Ai). The average duration of a wing movement cycle underlying a syllable was 38.1 ± 1.5 ms (n = 5 crickets). IPSPs started 5.0 ± 0.2 ms after the start of wing closing and reached a maximum at 23.3 ± 1.4 ms, just after the transition between wing closing and opening at 19.0 ± 0.7 ms (dashed lines in Fig. 4,Aii and Bii). The average mean amplitude of the IPSPs increased from −2.3 ± 0.3 mV at the first syllable to −5.3 ± 0.5 mV at the last (average from 8 crickets, measured from chirp-interval baseline membrane potential to peak amplitude). Following the last IPSP, the membrane potential gradually repolarized to the resting level. Similar to the PADs, the amplitude and timing of the IPSPs was independent of the amplitude of wing movements (Fig.4 B).

Fig. 4.

Activity of ON1 during silent stridulation. Inhibitory postsynaptic potentials (IPSPs) of similar amplitude occurred in ON1 during normal-amplitude (Ai) and small-amplitude (Bi) wing movements. The IPSPs started after the start of the opening movement of each syllable and reached a maximum at the beginning of the consecutive wing closing. An IPSP is highlighted (arrow 1, Ai) where the small-amplitude hyperpolarizing potentials comprising the IPSP are evident. Aii,Bii: superimposed recordings of ON1 together with the averaged wing movement and sound demonstrate the timing of the IPSPs in relation to the wing movement. The peak amplitude of the IPSP occurs just after the transition between wing closing and opening (dashed lines in Aii, Bii). The spikes have been truncated in the superpositions. A and Bare recordings from the same neuron. C: repetition rate of small potentials. A typical PAD and IPSP are shown and an example of how the timing of the potentials was measured is marked below the recordings with dashed lines and an arrow. The average maximum frequency of the small potentials comprising the PADs and IPSPs was very similar in the PAD (228 ± 29 Hz, n = 5) and the IPSPs (232 ± 22 Hz, n = 6).

The PADs and IPSPs were composed of many small-amplitude potentials (Fig. 4 C). The PADs were built up by the summation of small depolarizations (arrow 2 in Fig. 3 Ai, arrow 1 in Fig.4 Ai), while the IPSPs were composed of summated hyperpolarizing potentials (arrow 1 Fig. 4, Ai andC). The average maximum frequency with which these small potentials occurred were very similar in both the PADs (228 ± 29 Hz, n = 5 crickets) and the IPSPs (232 ± 22 Hz,n = 6 crickets), implying that they could originate from the same source.

Thus the responses of ON1 in sonorously singing crickets must be mixture of excitation from reafferent sound and additional inhibition. The inhibitory synaptic inputs may have two sources: reafferent feedback from nonauditory sense organs or the network of neurons generating the stridulation motor pattern. To examine this further, we recorded the responses of the neurons in fictively stridulating crickets, devoid of sensory feedback and efferent muscle control.

Activity of auditory neurons during fictive stridulation

During fictive stridulation, the afferents and ON1 rarely produced spikes (Fig. 5). PADs were present in low-frequency afferents (n = 17) and in a high-frequency afferent. The data collected from six low-frequency afferents were analyzed further. Although wing closer and opener motor activity is not as clear an indicator for the timing of stridulation as wing movement, the PADs seemed to have a similar timing and amplitude to those seen in intact animals (Fig. 5 A). Just as in intact animals, the PADs were composed of smaller amplitude depolarizing potentials that occurred in-phase with the syllables. The PADs thus increased in amplitude from 1.9 ± 0.3 mV at the start of the fictive chirp to 3.5 ± 0.1 mV at the end (n = 6 crickets).

Fig. 5.

Activity of a primary auditory afferent and ON1 during fictive stridulation. Ai: PADs were recorded in auditory afferents during the chirps indicated by thoracic motor activity, with the same amplitude and timing as in silently stridulating crickets. A chirp is marked below the extracellular nerve recording.Bi: IPSPs were present in ON1 during fictive chirps with the same timing and amplitude as in the silently stridulating crickets.Aii, Bii: superimposed traces of neuron recordings (top) show the timing of the PADs and IPSPs in relation to the averaged, rectified mesothoracic nerve 3A recording (bottom). Meso Nv 3A: extracellular nerve recording with bursts of wing opener and closer motor neuron activity.

IPSPs were present in ON1 that were also composed of many smaller amplitude hyperpolarizing potentials. They occurred in-phase with the syllables of the fictive chirps, which were indicated by the motor activity. The maximum amplitude of the IPSPs increased from −2.4 ± 0.6 mV in-phase with the first syllable of the fictive chirp to −5.1 ± 0.7 mV at the last (n = 5 crickets) (Fig.5 Bii). As during silent singing, following the last IPSP of a chirp, the membrane potential gradually repolarized back to a resting level. To prove that the IPSPs and PADs do not originate in the auditory pathway, in some of the fictively singing crickets the front legs, which contain the ears and other sensory systems, were removed. During fictive stridulation, with the ears removed, both PADs (n = 2 crickets) and IPSPs (n = 4 crickets) persisted, with the same overall temporal structure and amplitude as in the intact animal (Fig.6, A and B). We conclude that the PADs in the afferents and IPSPs in ON1 were not the result of any reafferent feedback but were generated within the CNS of the crickets and therefore represent an efferent input from a corollary discharge generated by the singing motor network.

Fig. 6.

Activity of auditory neurons during fictive stridulation after the ears were removed by amputating the forelegs. Ai: PADs andBi: IPSPs persisted in fictively stridulating crickets even after removing the ears. In the afferent shown in this figure the motor activity at the start of the chirp was hard to identify; therefore, the last burst of activity in the chirp was used as a temporal reference point. Aii, Bii: the superpositions of the neuronal recordings (top) indicate that the PADs and IPSPs had a very similar time course to those occurring in the silently stridulating cricket.

Responses of auditory neurons to acoustic stimulation during stridulation

To examine the effect of the efferent synaptic input on auditory processing in the afferents and ON1, trains of 4.5 kHz, 75 dB SPL acoustic stimuli with a short period (15 ms) and duration (7 ms) were played to silently singing crickets (Fig.7). This stimulus paradigm revealed the time course of the auditory response during stridulation, as sound stimuli were present at all phases of stridulation (Fig. 7,Aiii and Bii). In the afferents spikes elicited by the sound were superimposed on the PADs caused by the corollary discharge (asterisk in Fig. 7 Ai). The PADs did not have an effect on the frequency of spiking, which was the same in chirps and the chirp interval (Fig. 7 Aiii) (n = 5 crickets). In two recordings, however, spikes riding on top of the PADs were reduced in amplitude by approximately 3 mV (Fig. 7, Aiand Aii).

Fig. 7.

Responses of an afferent and ON1 to sound during silent stridulation.Ai: the spiking response of an auditory afferent to a train of sound pulses (4.5 kHz, 75 dB SPL, 7-ms duration, 15-ms interval) is superimposed on PADs during the chirps (asterisk,Ai). Aii: the afferent spike height decreased if it coincided with a PAD. A short section of the original recording that has been expanded to show this more clearly.Aiii: quantitative analysis shows the PST histogram with the superimposed spike frequency (top), the averaged wing movement (middle), and the distribution of the sound stimuli (bottom). Neither the PST histogram nor the average spike frequency show any modulation of spiking activity during the chirp. Bi: when presented with the same sound stimuli, ON1 responded to sound during the chirp intervals, but its response was inhibited during the chirps. Excitatory postsynaptic potentials (EPSPs) were recorded during the chirp (asterisk,Bi); occasionally spikes were recorded on top of the larger EPSPs during the transition between closing and opening wing movement (arrow 1, Bii). Bii: the sound stimuli were played constantly throughout the chirp and chirp interval (bottom). Both the PST histogram and the average spike frequency of ON1's response (top) show a clear reduction during the chirp. The start of the inhibition is marked by arrow 2 and the end by arrow 3. Acoustic stimuli: sound pulses.

ON1 spiked consistently to the continuous sound stimuli while the cricket was at rest. During silent stridulation, ON1 was activated by the acoustic stimuli in the chirp intervals, but its response was strongly inhibited during the silent chirps (Fig. 7 Bi) (n = 21 crickets). Inhibition started at the first closing wing movement of a chirp and occurred in the rhythm of the syllable pattern, as seen before (Fig. 4). Excitatory postsynaptic potentials (EPSPs) were elicited by the sound pulses during the chirp (asterisk in Fig. 7 Bi), and occasionally spikes were triggered by the larger amplitude EPSPs at the transition from wing opening to closing (arrow 1 in Fig. 7 Bii) when the inhibition slightly decreased (Fig. 4). The inhibition followed the structure of the IPSPs in ON1: it had a rapid onset at the start of the chirp (arrow 2, Fig. 7 Bii) and at the end of the chirp its effect gradually waned (arrow 3, Fig. 7 Bii, compare with Figs. 5 Bii and 6 Bii).

Afferent and ON1 recordings also showed these inhibitory effects on auditory processing in fictively stridulating crickets: afferents received PADs during fictive chirps but constantly responded to sound stimuli throughout fictive stridulation (Fig.8 A) (n = 10 crickets), whereas ON1 was rhythmically inhibited during fictive chirps (Fig. 8 B) (n = 6 crickets). In one fictively stridulating cricket, the spikes on top of the PADs were reduced in amplitude by approximately 3 mV (Fig. 8 Ai). These experiments clearly demonstrate that the corollary discharge has a major influence on auditory processing during stridulation at two levels. It presynaptically inhibits the auditory afferents, thereby reducing the amount of excitation forwarded to the central auditory pathway, and additionally it inhibits auditory interneurons in-phase with wing closing movement.

Fig. 8.

Auditory responses to acoustic stimulation during fictive singing.Ai: auditory afferents spiked in response to a sequence of sound pulses. The afferent spike amplitude was reduced if it coincided with a PAD. Aii: there was no modulation, however, in the spiking activity as demonstrated by the PST histogram, with overlaid average spike rate (top), and the PST histogram of the sound pulses (bottom).Bi: when presented with the same sequence of sound stimuli, ON1 was activated during the chirp intervals and at rest, but its response to sound was inhibited during the chirps.Bii: analysis shows the PST histogram and overlaid average spike rate (top) of ON1's response to acoustic stimuli (bottom) are reduced during the fictive chirp, as indicated by the rectified motor activity (middle).

The impact of the inhibition on sound patterns similar to cricket song

To examine the efficacy of the corollary discharge, we presented five silently stridulating crickets with acoustic stimuli that mimicked the species-specific calling song (4.5 kHz, 100 dB SPL, 21-ms duration, 42-ms interval) (Fig. 9). Sound pulses presented at rest and during the chirp intervals evoked a depolarization and a burst of six spikes in ON1 with an average maximum spike frequency of 376 ± 50 Hz. During the chirps, the compound EPSPs elicited by the sound stimuli were superimposed with small deflections, indicating the efferent IPSPs (arrows, Fig.9 A). Alongside the presynaptic inhibition, these IPSPs reduced the amplitude and duration of the sound-evoked EPSPs. During the chirps, these depolarizations triggered only one to two action potentials/stimulus with a spike rate of only 123 ± 29 Hz (n = 5 crickets) (responses to gray stimuli in Fig.9 A). The response during the chirps, as reflected by the discharge rate and the number of spikes per stimulus, was only 33% of the response during the chirp intervals, which was significantly lower (2-tailed paired t-test: P < 0.002,t = 7.71, df = 4) (Fig. 9 B).

Fig. 9.

Effectiveness of centrally generated inhibition during silent stridulation. A: a typical ON1 response to 100 dB SPL sound pulses during a silent chirp. ON1 responded with about 6 spikes to stimuli in the chirp interval (in black), but with only 2 spikes during the chirps (in gray). Arrows indicate small IPSPs generated by the corollary discharge superimposed with the depolarization elicited by the sound pulses. B: the mean maximum spike frequency during the chirp interval (black) or the chirp (gray); data based on averages of 150 sound pulses presented to 5 silently stridulating crickets.

Responses to the 100 dB SPL stimuli varied in amplitude depending on the phase of occurrence within the silent chirp, e.g., compare response to the first stimulus in the chirp to the response to the second (Fig.9 A). To examine the exact time course of the response modulation throughout the silent chirp, we plotted the maximum spike frequency of each response (represented as a dot) against the averaged wing movement and then calculated the average maximum spike frequency during the chirp (solid line, Fig.10 A). ON1 responded with regular high-frequency bursts of spikes (in this example, at about 280 Hz) during the chirp interval. The response of ON1 to the acoustic stimuli was minimal (in this example, about 10 Hz) during the opening wing movements (gray bars, Fig. 10; compare with Fig. 4 A). A weak spiking response to the acoustic stimuli could occur during the closing movement when the inhibition decreased (compare with Figs.4 A and 7 B). Thus the effect of the efferent inhibition is maximum in-phase with wing opening.

Fig. 10.

Timing of the centrally generated inhibition in ON1. A: each dot represents the maximum spike frequency of ON1 made in response to a 100 dB SPL acoustic stimulus played during silent singing. The black line represents the average spike frequency. B: the wing movement of a different cricket during a period of sonorous singing was aligned with the wing movement in (A) to compare the period of inhibition with the average spike frequency and PST histogram in (B). Gray bars show that the timing of the maximum response reduction during silent singing coincides with the maximum response of ON1 to self-generated sound.

We then compared the timing of the inhibition (Fig. 10 A) with ON1's response to reafferent sound (Fig. 10 B). Sound is generated during the closing movement, but due to ON1's latency of 15–20 ms its response occurred in-phase with the following wing opening movement (Fig. 10 B). If the wing movements from a sonorously and a silently singing cricket were aligned, the phase of maximum inhibition coincides with the phase in which ON1 is maximally driven by reafferent sound stimulation (gray bars in Fig. 10). Thus the inhibition, mediated by the corollary discharge, is precisely timed to reduce reafferent auditory stimulation.

The functional role of the corollary discharge

In female crickets, intense acoustic stimulation causes ON1 to spike but it also causes a gradual activity-dependent hyperpolarization that reduces the neuron's sensitivity to sound (Pollack 1988). This inhibition is thought to be due to activation of Ca2+-dependent K+ channels by the spikes (Sobel and Tank 1994). The duration and amplitude of the hyperpolarization is dependent on the duration and amplitude of acoustic stimulation (Pollack 1988). If a similar mechanism is present in males, self-generated acoustic stimulation during singing should gradually hyperpolarize ON1 and desensitize it to external sounds. We tested whether such a desensitization is present and whether it is prevented by the corollary discharge. Reducing ON1's spiking response to loud chirps should increase its response to subsequent, quieter sounds.

We examined the responses of ON1 to sound pulses that mimicked the cricket's own song and to quieter test stimuli. A control series of test sound pulses (80 dB SPL) each elicited a burst of spikes with an average maximum frequency of 203 ± 24 Hz (n = 8 crickets) (Fig. 11, A andD). The sound of normal singing was then mimicked with 100 dB SPL chirps, which were presented immediately before a series of test stimuli. The response to these stimuli was significantly weaker than the response to the control series with an average maximum spike frequency of 30 ± 11 Hz (2-tailed paired t-test:P < 0.001, t = 8.70, df = 7) (Fig. 11, A, B, and D). The reduction persisted for the whole of the artificial chirp interval but was greatest just after the 100 dB SPL chirp (Fig. 11 B).

Fig. 11.

Inhibition of ON1's spiking response during loud acoustic stimulation prevents desensitization. A: in a control experiment, ON1 produced a burst of spikes at 203 Hz in response to 80 dB SPL pulses. B: ON1's response to 80 dB SPL sound pulses was greatly reduced if they were preceded by a 100 dB SPL chirp.C: if spiking was prevented by hyperpolarizing current injection during a response to the 100 dB SPL chirps, the spiking response to the subsequent 80 dB SPL pulses was 143 Hz.D: the pooled, averaged maximum spike frequency of 8 crickets to 150 consecutive presentations of each stimulus paradigm. ON1 responses to (left) the control 80 dB SPL stimuli; (middle) 80 dB SPL stimuli preceded by a 100 dB SPL chirp; (right) 80 dB SPL stimuli preceded by hyperpolarizing current injection during the 100 dB SPL chirp.Current: Current injected into the cell.

We then mimicked the effects of the inhibition from the corollary discharge by injecting hyperpolarizing current into ON1 to prevent it from spiking during the 100 dB SPL chirp. The average maximum response to the subsequent 80 dB SPL test stimuli was now a burst of spikes at 143 ± 32 Hz (n = 8 crickets) that was significantly higher than the response to the test pulses after the 100 dB SPL chirps without hyperpolarizing current injection (2-tailed paired t-test: P < 0.004, t= −4.21, df = 7) (Fig. 11, A, C, andD). ON1's response following current injection did not reach the same amplitude as to the test stimuli because afferent adaptation occurs at high sound intensities (Givois and Pollack 2000; Nocke 1972). The effect was still present after deafening the ear contralateral to the soma of ON1 and was therefore independent from inhibition from the contralateral ON1. In three one-eared crickets, the average spike frequency to 80 dB SPL stimuli was 263 Hz. When the stimuli were preceded by a 100 dB SPL chirp, the response to the same stimuli was reduced to 23 Hz. If the neuron was hyperpolarized during the chirp, the response increased to 168 Hz. If a burst of spikes was generated by a depolarizing current injection in a completely deafened cricket, the graded hyperpolarization following spiking persisted. Therefore a reduction in the spiking response of ON1 to intense sounds helps to maintain the neuron's sensitivity to quieter, subsequent sounds. In this way the corollary discharge will prevent auditory desensitization during stridulation.


Crickets sing so loudly that reafferent sound could be confused with external sound and/or desensitize the cricket's own auditory system. One solution to this problem could be to modulate the biophysical sensitivity of the ear during sound production. However, the tympanic membrane of the cricket remains fully responsive during stridulation (Poulet and Hedwig 2001). Therefore we examined how their central auditory system copes with the intense reafferent stimulation.

Inputs to auditory afferents and ON1 during singing

The auditory afferents and ON1 spiked in response to sound production during sonorous chirps (Fig. 2). Spiking was greatly reduced in both types of neuron in silently stridulating crickets (Figs. 3 and4); therefore, during sonorous stridulation, the response of the neurons was a reaction to reafferent auditory stimulation. Excitation from reafferent sound stimulation, however, was not the only type of input to these neurons. During silent chirps, PADs became evident in the afferents and IPSPs in ON1 (Figs. 3 and 4).

These inhibitory inputs could have had two sources: reafferent nonauditory feedback generated during stridulation or a centrally generated pattern of neuronal activity. As both the PADs and the IPSPs were presented during fictive chirps with all forms of reafferent feedback removed, the major source of the inhibition must be a central one (Figs. 5 and 6). Vibrating the foreleg of the cricket inhibits ON1 (Wiese 1981; unpublished observations). In the intact animal, therefore, vibrations conducted from the wing via the exoskeleton to the foreleg may result in reafferent inhibition superimposed on the central inhibition.

Both the PADs and the IPSPs were present during the chirps and occurred at the same phase of the syllable cycle. Both were composed of many, summed, smaller potentials that were generated at regular intervals (every 4–5 ms, 230 Hz). This suggests that they may have a common source. Furthermore, the unitary nature of the small potentials suggests that they are the result of input from a single neuron. The neuron(s) causing these inputs is unlikely to be the descending command neuron(s) for stridulation as it spikes during the chirp at about 35 Hz with little modulation of its firing rates (Hedwig 2000a). Interneurons participating in stridulatory motor pattern generation in crickets and grasshoppers reach phasic spike rates of ≥200 Hz (e.g., Hedwig 1992a,b; Hennig 1990). The most likely candidate neuron(s) mediating the inhibitory corollary discharge is therefore those contained within the so-far uncharacterized stridulatory motor network in the thoracic ganglia. The amplitude of the PADs and IPSPs remained the same during low-amplitude wing movement (Figs. 3 and 4). This implies that the neuron(s) mediating the corollary discharge is involved in central rhythm generation rather than in the finer control of the stridulatory wing movement (for a distinction between these types, see Hedwig 1992a,b).

Neural signals generated within motor networks that affect the processing of reafferent information were termed efference copies byHolst and Mittelstaedt (1950) or corollary discharges bySperry (1950). According to Bell (1984), all neural discharges from the networks of neurons forwarding motor activity to a sensory pathway should be termed corollary discharges, since an efference copy depends on the properties of reafferent information. One well-studied example of a corollary discharge occurs in the knollenorgan electroreceptive pathway, which is involved with electrocommunication between mormyrid electric fish (Bell 1989; Bell and Grant 1989). To prevent reafferent stimulation from interfering with the detection of external signals, this pathway is inhibited by a corollary discharge during the electric organ discharge, which is independent from the intensity of reafferent stimulation. The inhibition in the cricket auditory pathway is very similar to that seen in the electric fish, as it is also the result of a corollary discharge that is independent from the production of reafferent feedback (Figs. 3, 4, 5, and 6). Furthermore, it is stable, in that it continued for as long as we could record from a singing cricket (up to 1 h). Inhibition by an invariant corollary discharge is a sensible strategy for crickets to adopt as they may sing continuously for many hours with a very stereotyped, regular sound pattern, so the afferent response to self-generated sound is also unlikely to change.

Effects of central inputs in the cricket and other systems

The responses to acoustic stimulation during silent and fictive singing indicate that both the PADs in the afferents and the hyperpolarizations in ON1 are inhibitory. PADs in many other systems are inhibitory synaptic inputs with a reversal potential more positive than the resting membrane potential (Clarac and Cattaert 1996). If the auditory nerve was intact (Figs. 3 and 5), spikes were occasionally recorded in the afferents even when the cricket was silent. These spikes originated in the periphery and were not generated by the PADs, as spikes were never recorded in afferents once the ears had been removed (Fig. 6 A). In many systems, PADs are thought to be mediated by GABA-ergic inputs to the afferent terminals (Burrows and Laurent 1993; Cattaert et al. 1992; Eccles et al. 1963). Consistent with this observation, GABA-immunoreactive boutons have been identified on the terminals of cricket auditory afferents (Hardt and Watson 1999). PADs can cause a reduction in the spike amplitude invading the afferent terminal (Burrows and Matheson 1994; Cattaert et al. 1992), as seen in this study (Figs. 7 and 8). This is in agreement with the shunting effect of PADs on action potentials due to an increased conductivity of the afferent terminal's membrane (Clarac and Cattaert 1996). By reducing the afferent spike height, the PADs could reduce the amount of excitatory neurotransmitter released from the afferent synapses during spiking, thus reducing the amplitude of the EPSP evoked in postsynaptic cells (Burrows and Matheson 1994; Cattaert et al. 1992).

During silent and fictive chirps, ON1's response to acoustic stimuli is inhibited in-phase with the IPSPs generated by the corollary discharge (Fig. 7). Hyperpolarizing ON1 inhibits its response to acoustic stimulation (Fig. 11). During sonorous chirps, ON1 receives a simultaneous drive of excitation from reafferent sound stimulation and inhibition from the centrally generated IPSPs.

The strength of the centrally generated inhibition was tested by playing sounds, which mimicked the calling song, to silently stridulating crickets. ON1's response during silent chirps was reduced to 33% of the response at rest and during chirp intervals (Fig. 9). The response at rest was also much higher than the response to self-generated sounds, confirming the response reduction during stridulation. The response during the chirps, however, was slightly lower than the response to self-generated sounds. This may be due to differences in the stimulus design compared with natural sounds.

The PADs and IPSPs occur throughout the chirp in-phase with the syllables: they start during each closing wing movement and reach a peak during the consecutive opening wing movement (Figs. 3 and 4). We therefore examined the responses to sound pulses played during silent chirps. The largest response reduction coincides with the opening wing movement, which is the time when ON1 would respond to self-generated sounds (Fig. 11). As the inhibition is generated within the nervous system, it can be precisely timed to certain phases of motor activity.

Studies made in stridulating grasshoppers demonstrated a reduction of the spiking response to acoustic stimuli during sound production (Hedwig 1986; Wolf and Helversen 1986). This reduction is, however, due to a desynchronization of receptor activity in response to the self-generated sound and not the result of neural inhibition (Hedwig 1990; Hedwig and Meyer 1994). A reduction in auditory neural responses during sound production has also been recorded in vertebrates (Creutzfeldt et al. 1989; Kirzinger and Jürgens 1991;McCasland and Konishi 1981; Metzner 1989,1993; Müller-Preuss and Ploog 1981;Numminen et al. 1999; Schuller 1979;Suga and Schlegel 1972; Suga and Shimozawa 1974). In these studies, however, the CNS has always been intact or semi-intact; therefore, inhibitory influences from nonauditory reafferent information during sound production cannot be completely ruled out. As these experiments used extracellular recordings, a reduction in spiking during vocalization could be the result of a reduction in excitatory input, rather than neural inhibition; therefore, a complete characterization of any inhibitory mechanisms involved is lacking.

Centrally generated inhibitory mechanisms that affect proprioceptive and mechanoreceptive pathways have been characterized during other motor behaviors, such as walking (Gossard et al. 1991;Murphey and Palka 1974; Sillar and Skorupski 1986; Wolf and Burrows 1995) and swimming (El Manira et al. 1996; Sillar and Roberts 1988). These types of sensory information have a direct impact on the ongoing motor activity and contribute to the maintenance of an adaptive motor output (Pearson 1993). We show that the cricket's stridulatory motor activity modulates the responsiveness of a sensory pathway that is not involved in regulating the ongoing motor activity. Inhibition of proprioceptive and mechanoreceptive pathways acts presynaptically at the afferent terminals (El Manira et al. 1996; Gossard et al. 1991; Murphey and Palka 1974; Sillar and Skorupski 1986;Wolf and Burrows 1995), but has also been observed at the level of a sensory interneuron (Sillar and Roberts 1988). Our study highlights the finding that a centrally generated neural mechanism, which controls sensory responsiveness, can act in parallel both at the afferent and at the interneuron level.

Biological function of the corollary discharge

Spiking in ON1 causes a graded hyperpolarization and a reduction of its response to acoustic stimuli. This mechanism is thought to allow female crickets to selectively attend to the louder of two singing males (Pollack 1988; Römer and Krusch 2000). Just as in females, we show that presentation of loud stimuli to male crickets desensitizes ON1 and prevents it from responding to subsequent, quieter stimuli (Fig. 11 B). This mechanism may allow males to listen selectively to the louder, and therefore nearer, of two conspecific males, which may pose a threat to their mating success with nearby females. This hyperpolarization is present if ON1 is made to spike with depolarizing current injection, which adds further weight to the proposal that this mechanism is an inherent property of ON1 (Sobel and Tank 1994).

Despite the desensitizing effect of loud sounds on the responsiveness of ON1, behavioral studies have shown that singing male crickets respond to acoustic stimuli that occur in the chirp intervals (Heiligenberg 1969; Jones and Dambach 1973). We show that the corollary discharge inhibition during the chirps will prevent desensitization in ON1 and allow the cricket to hear quieter, subsequent sounds in the chirp intervals (Fig.11 C). In a group of stridulating male G. bimaculatus, where individuals are spaced apart by 2 m on average, the ability to hear during singing will be an advantage. It would allow males to defend their territory from rival singing males, to maintain a fixed distance from each other, and to hear noisy predators (Simmons 1988).

Auditory information ascends from the prothoracic ganglion, where these recordings were performed, to the brain via at least two identified interneurons. Experiments are underway to examine whether auditory sensitivity is also modulated in these neurons during stridulation. A reduction in auditory responses has been recorded in walking (Schildberger et al. 1988; Staudacher and Schildberger 1998) and flying (Tomioka and Yamaguchi 1984) crickets; these reductions were suspected to be the result of central and peripheral effects. We aim to identify the neuron(s) causing the inhibition during singing and examine whether they are responsible for similar effects in other behaviors.


We thank T. Matheson and D. Parker for constructive comments on the manuscript, J. Rodford for help with Fig. 1 A, and S. Ellis and G. Harris for technical assistance.

This work was supported by a Biotechnology and Biological Sciences Research Council studentship and grants from the Wellcome Trust and the Royal Society.


  • Address for reprint requests: J. Poulet, Department of Zoology, University of Cambridge, Cambridge, CB2 3EJ, UK (E-mail:jfap2{at}cam.ac.uk).


View Abstract