Representation of Behaviorally Relevant Sound Frequencies by Auditory Receptors in the Cricket Teleogryllus Oceanicus

The ears of animals are often specialized to listen to sounds that are particularly relevant to their way of life. In the cricket species Teleogryllus oceanicus, two distinct frequency ranges are important. Intraspecific acoustic signals, which are used to attract mates, to induce copulation and in agonistic encounters, are dominated by frequencies in the range 4–5 kHz (Balakrishnan and Pollack, 1996). Aerially feeding echolocating bats produce signals with much higher frequencies, generally above 20 kHz (Simmons et al. 1979), to hunt insects. Crickets, like many other insects, can hear these high-frequency sounds, which trigger evasive behavior (Hoy, 1991; Hoy and Robert, 1996). Both behavioral and neurophysiological studies demonstrate that crickets are particularly sensitive to sounds in these two frequency ranges. For example, the behavioral audiogram describing the frequency-sensitivity of flight phonotaxis of T. oceanicus shows two low-threshold regions, one near 5 kHz and a second near 30 kHz (Moiseff et al. 1978). Similarly, most auditory interneurons in this species show increased sensitivity to one or both of these frequency ranges (Moiseff and Hoy, 1983; Atkins and Pollack, 1986, 1987a,b).

On the basis of recordings from interneurons, Pollack (1994) proposed that sounds in the lower frequency range, 4–5 kHz, are detected by many auditory receptor neurons, while ultrasounds are detected by only a few receptor neurons. In the present work we examine this hypothesis by recording from the auditory nerve.

Teleogryllus oceanicus were raised in a laboratory colony on a diet of Purina Cat Chow with ad libitum access to water. Virgin females, aged at least 9 days after the last molt, were mounted on a wax block ventral side uppermost. The forelegs were fixed with the tibia and tarsus flexed against the femur, similar to their position during flight. However, to provide access to the anterior surface of the femur for electrode implantation, the coxae were rotated so that the legs, which in flight are held against the lateral surfaces of the body, projected laterally nearly at right angles to the longitudinal axis of the cricket Auditory receptor neurons were recorded en masse with Teflon-insulated silver-wire electrodes (o.d. 114 μm). The axons of the receptor neurons travel from their origin in the prothoracic tibia to the central nervous system in branch TB of N5 (Eibl and Huber, 1979). In the femur, branch TB lies on the anterior-dorsal surface of the main leg trachea, which is visible through the intact cuticle. Small holes were made in the cuticle with a pin, using the trachea as a guide, and electrodes were inserted until their tips nearly touched the trachea. To measure conduction velocity, we recorded simultaneously at two sites along the length of the femur separated by approximately 2 mm. The indifferent electrode was placed proximally and ventrally in the femur.

Individual receptors were recorded in N5 at a point close to its entry into the prothoracic ganglion using glass microelectrodes filled with 3 mol l⁻¹ KCl (resistance 20–100 MΩ). The ganglion was supported on a steel plate and bathed in physiological saline (Strausfeld et al. 1983). Experiments were performed at room temperature (20–25°C).

Sound pulses were produced with National Instruments A/D–D/A boards (12-bit resolution; update rate ≥200 kHz) controlled by software written using LabWindows. Unless otherwise noted, pulses had rise and fall times of 0.2 ms and an overall duration of 20.4 ms. Signals were amplified (Amcron) and broadcast either through piezoelectric tweeters (Motorola) or through 4 inch magnetic loudspeakers (Intertan). Stimuli were presented ipsilateral to the recorded nerve. Carrier frequency was either 4.5 kHz or 30 kHz, and intensity was varied (using either a custom-built programmable attenuator from Mike Walsh Electronics or Tucker-Davis Technologies PA4) from 25 dB to 100 dB in steps of 5 dB. Each frequency–intensity combination was delivered 50 times at a repetition rate of 2 s⁻¹. The stimuli were given in separate series of intensity steps, one ranging from 25 to 95 dB in steps of 10 dB, and the other ranging from 30 to 100 dB. In most experiments the curves relating response magnitude and stimulus intensity were smooth when the measurements from the two series were interleaved, indicating that responses and recordings were stable throughout the experiment. Two experiments failed to meet this criterion and were discarded.

Recordings were digitized (12-bit resolution, sampling rate 10 kHz per channel) using National Instruments A/D–D/A boards and the program SWEEPS (Pollack, 1997), which was also used for data analysis. Statistical tests were performed with SYSTAT or STATISTICA software.

In the Results section, we compare compound action potentials (CAPs) recorded from the auditory nerve in response to 4.5 kHz and 30 kHz sound stimuli. To avoid confounding frequency-dependent and intensity-dependent effects, most comparisons are made at equivalent sound intensities relative to threshold. Clear CAPs, which depend on the synchronous activity of many receptors (see below), are often not apparent until intensity is well above the minimum necessary to elicit auditory responses in the nerve as determined by visual inspection of the recordings. We define threshold not on the basis of CAPS but on the integrated responses of the nerve. We integrated recordings (following full-wave rectification) over two 50 ms time windows, one representing the response to sound, beginning at stimulus onset, and the second, a control measurement, beginning 100 ms later. The response was defined as the difference between the two measurements. Threshold (T) was defined as 5 dB greater than the most intense stimulus for which the mean of all responses (N=27–50) did not differ significantly from zero (P>0.05, t-test). (Responses to some sound pulses were obscured by bursts of motor activity that masked the much smaller auditory potentials. These responses were excluded from the analysis, resulting in lower sample sizes.) Threshold was 55.3±5.8 dB SPL at 4.5 kHz and 69.3±8.4 dB at 30 kHz (mean ± S.D., N=15). These values are 10–15 dB higher than behavioral thresholds and than the thresholds of some auditory interneurons (see, for example, Moiseff et al. 1978; Moiseff and Hoy, 1983; Atkins and Pollack, 1986), probably because of the masking of responses to low-intensity stimuli by ongoing, non-sound-related spiking in the nerve.

Recordings from the whole nerve exhibit periodic compound action potentials (CAPs) (Fig. 1A). Similar periodic waveforms have been recorded in a number of other insects (acridids, Adam, 1977a; tettigoniids, Fullard et al. 1989; Rössler et al. 1990; cicadas, Huber et al. 1980; blattids, Shaw, 1994; Coleoptera, Yager and Spangler, 1995). The origin of the waveform was carefully studied in the locust by Adam (1977b). He found that different receptors have similar minimum latencies and maximum firing rates, and that periodic CAPs reflect the nearly synchronous activity of those receptors that have attained these conditions. Fig. 1 shows that such receptor synchronization explains cricket CAPs. Fig. 1A shows the response of the whole nerve and Fig. 1B shows a simultaneous recording from a single receptor cell recorded at the point where the nerve enters the central nervous system; this trace has been shifted a few milliseconds to the left to align the first spike with the first peak in the CAP. As the response proceeds, the single-unit spike becomes increasingly out of register with the CAPs, suggesting that synchronization among receptors decays with successive spikes. Accordingly, CAPs become progressively smaller throughout the response.

We analyzed in detail only the first CAP of a response, where synchronization among receptors appears to be greatest (see below). Thus, our measurements capture responses of receptors whose first spikes arrive at the recording electrode with similar latency, no matter what their firing rates. Indeed, at lower stimulus intensities, often only a single clear CAP is evident in averaged recordings, indicating either that the contributing receptors produced only one spike or that their firing rates were sufficiently different to preclude synchronization except at the beginning of the response.

Averaged recordings from the auditory nerve are shown in Fig. 2A,B. At low intensity, responses to 4.5 kHz and to 30 kHz are similar in amplitude. As intensity increases, however, the amplitudes of responses to 4.5 kHz increase markedly, while those of responses to 30 kHz increase only slightly. The results of several experiments are summarized in Fig. 2C. At a stimulus intensity 30 dB above threshold, CAPs evoked by 4.5 kHz stimuli are more than four times larger than those evoked by 30 kHz stimuli.

One hypothesis to account for these findings is that more receptor neurons respond to 4.5 kHz stimuli than to 30 kHz stimuli. CAPs increase with intensity as additional receptors are recruited but, because the pool of available receptors is, according to this hypothesis, larger for 4.5 kHz stimuli, the total number recruited will be greater, resulting in larger CAPS. However, CAP amplitudes could also differ because receptors that respond to 4.5 kHz produce larger unit spikes, which would sum to produce larger CAPs, or because they are more tightly synchronized, resulting in more effective summation. We now examine each of these possibilities in turn.

Fig. 3 illustrates single-unit recordings from two receptors, one that responded strongly to 4.5 kHz but only weakly to 30 kHz, and one that responded to 30 kHz but not to 4.5 kHz. In this paper, we classify receptors simply as either low- or high-frequency types, according to their strength of response (i.e. number of spikes elicited) to these two frequencies. A more detailed treatment of the physiological characteristics of receptors will be presented elsewhere (Imaizumi and Pollack, 1997, and in preparation).

To measure the amplitudes of single-unit spikes in the whole-nerve recording, we recorded simultaneously from single units and from the whole nerve. We visualized the spike in the whole-nerve recording of the neuron recorded as a single unit by searching for events in the whole-nerve recording that preceded the single-unit spikes with constant latency (Fig. 4A). Because the spikes of different receptors are synchronized at the beginning of responses, the events revealed using this method might include not only the single unit recorded with a microelectrode but also spikes from other, synchronously firing, receptors. We used two procedures to avoid this. When possible, we used the spontaneous activity of receptors to visualize their spikes in the whole-nerve recording. For receptors that were not spontaneously active, we stimulated with 200 ms sound pulses. Periodic CAPs, indicative of synchronization among receptors, are evident only during the initial portion of the response (Fig. 4B). Accordingly, we use the last half of the response to visualize spikes in the whole-nerve recording.

We measured the amplitudes in the whole-nerve recording of spikes of both low- and high-frequency receptors in six preparations. To minimize inter-animal differences in the quality of the whole-nerve recording, we expressed unit spike amplitudes relative to the mean of all of the low-frequency units (N=1–4) recorded in the same preparation. The mean normalized spike amplitude of high-frequency receptors, 1.16±0.79 (mean ± S.D.) did not differ significantly from that of low-frequency receptors [1.0±0.40; 11 low-frequency receptors and eight high-frequency receptors recorded in six preparations; analysis of variance with weighted means (Searle, 1987), F=0.5, d.f.=1, P=0.52].

Axon diameter is an important determinant of spike amplitude in extracellular recordings (Pearson et al. 1970). Other things being equal, spike amplitude is expected to be proportional to the square of axon diameter (Rushton, 1951). Larger axons also conduct spikes more rapidly, with conduction velocity predicted to be proportional to the square root of diameter (Hodgkin, 1954; Pearson et al. 1970). Thus, conduction velocity is an indirect measure of spike amplitude. We measured conduction velocity by making whole-nerve recordings at two sites and measuring the conduction delay between them. The results are shown in Fig. 5. There is no difference in conduction velocities of CAPs evoked by 4.5 kHz and 30 kHz stimuli (Fig. 5A; ANOVA, F=0.058, d.f.=1, P=0.81). There is a slight, but significant (F=4.18, d.f.=3, P<0.01), increase in conduction velocity with increasing intensity, suggesting that receptors with larger-diameter axons may have higher thresholds. For the six preparations in which we could compare conduction velocities of low- and high-frequency receptors, we again found no difference (Fig. 5B, ANOVA with weighted means, F=0.124, d.f.=1, P=0.75).

In this section, we examine the hypothesis that the amplitude difference between CAPs produced by 4.5 kHz and 30 kHz stimuli is due solely to differences in the degree of spike synchronization among receptors. A direct method of studying synchronization would be to record simultaneously from several single receptors, but this is not technically feasible. Instead, we measured the widths of CAPs as an indicator of the degree to which receptors are synchronized. If the smaller amplitudes of CAPs evoked by 30 kHz stimuli were due to poor synchrony among the contributing receptors, then these CAPs would be broader than the larger CAPs elicited by 4.5 kHz. In Fig. 6A, CAPs evoked by both low- and high-frequency stimuli are superimposed. The traces are the same as those in Fig. 2A, T+30, rescaled to allow superposition. The 30 kHz response is not broader than the response evoked by 4.5 kHz. In contrast, successive CAPs within a single response, where receptor synchronization presumably deteriorates with time (Fig. 1), do become broader as they decrease in amplitude (Fig. 6B), supporting the validity of this approach.

We compared the measured relationship between CAP amplitude and breadth with the relationship predicted by a model that produced artificial CAPs from a population of single-unit spikes that were identical in size and shape, but were distributed in time according to a Gaussian distribution. By adjusting the standard deviation of the Gaussian distribution (range 0–1 ms), we adjusted the degree of synchronization among the receptors. Each model CAP was produced by summing, at each point in time, the instantaneous voltages of each of 25 contributing single units (preliminary tests showed that the output of the model was similar with five units), and CAP amplitude and breadth were measured from the average of 50 model CAPs, each produced using the same standard deviation of spike timing.

The model was run separately for nine preparations, each time using a spike of an appropriate shape for that preparation. Rather than using actual single-unit action potentials (which were available only for a subset of the preparations and which differed slightly from receptor to receptor, even within a single preparation) to determine the unit shape, we used instead the first CAP in the response to a 4.5 kHz, 100 dB stimulus. Fig. 7 shows the spikes of four single units recorded in a single preparation, visualized in the whole-nerve recording with averaging (see Fig. 4), together with the CAP evoked by 4.5 kHz, 100 dB. The shape of the CAP is similar to the shapes of the single-unit spikes, suggesting that this intense stimulus evokes highly synchronized activity among receptors. The half-widths of CAPs (duration at 50% of maximum amplitude; see inset to Fig. 8) in response to 4.5 kHz, 100 dB did not differ significantly from the half-widths of single-unit spikes recorded in the same preparations (15 single units recorded in five preparations; ANOVA with weighted means, F=1.88, d.f.=1, P=0.2). In some preparations, the second positive phase of the CAP failed to return to the baseline before the arrival of the initial positive phase of the second CAP. In these cases, the CAPs were edited with the program GoldWave before use in the model, so that they returned smoothly to the baseline.

Fig. 8 shows that, as expected, when model CAPs decrease in amplitude because of poor synchronization among receptors, they also increase in breadth (Fig. 8, open circles). As a crude test of the validity of the model, we analyzed the first three successive CAPs of responses to 4.5 kHz, 100 dB, which typically decrease progressively in amplitude (Figs 1, 2, 6). We assumed that the population of receptors participating in the response was relatively stable over this short period (approximately 10–12 ms) and, thus, that the decrease in amplitude of successive CAPs was due to decreasing synchronization among receptors rather than to a decrease in the numbers of receptors responding. The points corresponding to these CAPs follow the prediction of the model rather closely (Fig. 8, filled circles). In contrast, CAPs evoked by 30 kHz, 100 dB, though small in amplitude, were considerably less broad than predicted in the model. We conclude that the small amplitudes of these CAPs cannot be accounted for by poor synchronization of receptors.

For responses in which receptors are highly synchronized, i.e. at high stimulus intensity (Fig. 7), the number of receptors contributing to the CAP will be approximated by the ratio of the CAP amplitude to the amplitude of single-unit spikes. We calculated these ratios for responses to 100 dB stimuli for five preparations, using the mean amplitude of all single units recorded per preparation (N=1–5) as the denominator. An average of 14.6±2.6 (mean ± S.D.) receptors contribute to CAPs at 4.5 kHz, 100 dB, whereas only 4.2±2.7 contribute to 30 kHz CAPs (P<0.001, paired t-test).

Our results support the hypothesis that the two frequency ranges known to be behaviorally relevant to T. oceanicus are represented unequally at the level of auditory receptors. Compound action potentials are up to four times larger for stimuli at a frequency of 4.5 kHz than for those at 30 kHz (Fig. 2). This cannot be ascribed to differences in unit spike sizes or in the degree of synchronization of receptors that respond to these two frequencies, leaving different numbers of receptors as the most likely explanation. This conclusion supports Pollack’s (1994) suggestion that central neurons are driven by a relatively large number of low-frequency receptors and by relatively few ultrasound receptors.

We used conduction velocity as an indirect measure of spike amplitude. Both of these parameters reflect longitudinal current flow during propagation of the action potential. The magnitude of the current depends on the change in membrane potential (i.e. the amplitude of the action potential as recorded intracellularly) and on the axial resistance of the axon which, in turn, depends on the resistivity of its axoplasm and on its cross-sectional area (Hodgkin, 1954; Rushton, 1951). Thus, although we found no difference in conduction velocity between low- and high-frequency receptors, this implies equal spike amplitudes only if intracellular action potential amplitude and axoplasmic resistivity are identical. Our conclusions from conduction-velocity measurements are supported by direct measurements of spike amplitude and suggest that these conditions are met.

Most of our measurements of conduction velocity were made from whole-nerve recordings, using the negative peak of the CAP to indicate its arrival time at each recording electrode. This is a rather poorly defined aggregate measure of the conduction velocities of all of the receptors that contribute to the CAP. Conduction velocities were similar when measured for single units (Fig. 5B), suggesting that any errors associated with measurements from CAPs are minor. More importantly, both sorts of measurements indicated that conduction velocities are similar for low- and high-frequency receptors.

Pollack (1994) found that the response latency of the interneuron ON1 is longer for low-frequency stimuli than for ultrasonic stimuli and hypothesized that this might be due to more rapid spike conduction in high-frequency receptors. These receptors drive short-latency responses that are believed to mediate escape from predators (Hoy, 1991), and selection for high conduction velocity would not be unexpected. We can now reject this hypothesis; the explanation for the latency difference of ON1 must be sought elsewhere. Experiments toward this end are in progress (Faulkes and Pollack, 1997).

We measured the relationship between the amplitude and breadth of CAPs to determine whether differences in spike synchronization might explain the amplitude difference of CAPs in response to low and high frequencies. Ultrasound-elicited CAPs were not broader than those evoked by 4.5 kHz (Fig. 7), nor did their relative amplitude and breadth match the predictions of a model of the effect of synchronization on CAP amplitude (Fig. 8). We did find that successive CAPs within a single response (to 4.5 kHz), which might be expected to show increasingly poor synchronization, did match the prediction of the model (Fig. 8). This gives us some confidence that the model we used is an appropriate one.

Our measurements include only receptors whose spikes arrive at the recording electrode nearly synchronously. Receptors that respond to a stimulus, but whose latencies or conduction velocities differ from the ‘norm’, will be excluded. The low-frequency receptor of Fig. 3, for example, responded to 30 kHz stimuli, but with such long latency that its spike would not contribute to the first CAP of the response. The latency of this receptor to 30 kHz is approximately 10 ms longer than the latency of another receptor(s?) in the same preparation, visible as small deflections of the trace. CAPs are only 1–2 ms in duration (Fig. 6), so receptors will be excluded from the first CAP of a response even with much smaller latency shifts than that shown in Fig. 3. Single-unit recordings show that, whereas low-frequency receptors only rarely respond to ultrasound, most high-frequency receptors respond to 4–5 kHz stimuli if intensity is high enough, although usually with longer latency than to the frequency to which they are most sensitive (Imaizumi and Pollack, 1997, and in preparation). By comparing CAP amplitudes and single-unit amplitudes, we estimate that approximately 15 receptors contribute to the first CAP of the response evoked by 4.5 kHz, 100 dB. If the total number of receptors in the ear of T. oceanicus approximates that in the closely related species T. commodus, which has approximately 70 receptors (Young and Ball, 1974), then only approximately 20% of the receptor population is included in the first CAP evoked by this stimulus. This probably seriously underestimates the total number of receptors that respond to this stimulus and may more closely reflect the number of receptors that are tuned to frequencies near 4.5 kHz, which would be expected to have the shortest latencies. A similar argument applies to the few receptors that contribute to CAPS elicited by 30 kHz, 100 dB stimuli.

Previous studies in crickets (summarized in Pollack, 1994), using single-unit recording techniques, also found that many receptors were tuned to the conspecific song frequency. In the single study that investigated sensitivity to ultrasound (Hutchings and Lewis, 1981), 36% of the recorded receptors were tuned to 4–5 kHz and at least 10% were tuned to ultrasound. That the agreement between these estimates (which are based on recordings from 42 receptors) and ours is imperfect is perhaps not surprising, given the differences in technique.

The representation of sound frequency at the receptor level has been studied in other orthopteran insects. In tettigoniids, the frequency-sensitivity of receptors is distributed more or less evenly throughout the hearing range (e.g. Römer, 1983; Kalmring et al. 1990), although the occurrence of several receptors with similar tuning (to the dominant frequency of the calling song) has been reported (Oldfield, 1984; H. Stölting and Stumpner, personal communication). The broad representation of sound frequency in tettigoniid ears corresponds to the rather broad-band acoustic signals of these insects (Kalmring et al. 1990). In acridids, receptors fall into four groups on the basis of both anatomy and physiology (Gray, 1960; Michelsen, 1971; Römer, 1976; Halex et al. 1988). Two of these groups, which share similar tuning to low frequencies, account for more than half the receptors in the ear (physiological types 1 and 2, anatomical groups a and c: Gray, 1960; Römer, 1976; Halex et al. 1988). Little is known about acoustic behavior in the species in which receptors have been studied most thoroughly (Locusta migratoria, Schistocerca gregaria), so the functional significance of this striking over-representation of a limited portion of the hearing range is not known. Experiments with another species, Chorthippus biguttulus, demonstrate that acridids do analyze the frequency content of acoustic signals. Males perform phonotaxis only in response to the low-frequency components of the broad-band signal of the female, while females answer male songs reliably only when both low- and high-frequency components of the song are present (von Helversen and von Helversen, 1997).

The best-studied example of peripheral over-representation of behaviorally relevant frequencies occurs in CF-FM bats, whose echolocation cries include a long constant-frequency (CF) tone. The region of the basilar membrane that is maximally excited by echoes of this CF frequency is expanded (Bruns, 1976), forming an ‘acoustic fovea’ (Suga and Jen, 1976). Receptors that are tuned to the CF range are disproportionately numerous and are also more sharply tuned than receptors for other frequencies (Suga et al. 1976). Over-representation of the CF range persists through central levels of auditory processing (Suga and Jen, 1976; Schuller and Pollak, 1979).

In crickets, sensitivity to 4–5 kHz and to ultrasound is apparent at the level of interneurons. Of six auditory interneurons that are known or, on the basis of anatomy, are likely to receive direct input from auditory receptors (Wohlers and Huber, 1985; Atkins and Pollack, 1987a; Hennig, 1988), three are primarily sensitive to 4–5 kHz (AN1, DN1, TN1; Wohlers and Huber, 1982; Atkins and Pollack, 1987b), another is mainly sensitive to ultrasound (AN2; Moiseff and Hoy, 1983), one shows dual sensitivity to both 4–5 kHz and to ultrasound (ON1; Atkins and Pollack, 1986) and the sixth shows broad-band tuning (ON2; Wohlers and Huber, 1982). Auditory information is carried to the brain, where behavioral decisions are made (Pollack and Hoy, 1981; Schildberger, 1984; Schildberger and Hörner, 1988), mainly by two first-order ascending interneurons, one (AN1) tuned to 4–5 kHz and the other (AN2) to higher frequencies. Among second- and higher-order interneurons, sensitivity to either or both frequency ranges appears to be relatively common (Atkins and Pollack, 1987a,b; Boyan, 1980; Brodfuehrer and Hoy, 1990; Schildberger, 1984). It is evident that the numerical under-representation of ultrasound receptors does not prevent them from having strong influences on central neurons and on behavior (Pollack, 1994). In this respect, crickets are similar to moths, in which detection and evasion of echolocating bats can be accomplished with as few as a single receptor neuron per ear (Surlykke, 1984).

Even in the absence of marked central over-representation of low sound frequencies, an abundance of low-frequency receptors might enhance the processing of intraspecific signals. First, differences in threshold among receptors would be expected to increase the dynamic range of their summed input beyond that of any single receptor, perhaps accounting for the observed wide (>50 dB) dynamic range of phonotactic responses to cricket song (Pollack and Plourde, 1982). Second, central convergence of the activity of a number of receptors might allow more precise encoding of the temporal features of the stimulus (Carr et al. 1986), which are known to be important in triggering behavioral responses to cricket song (Pollack and Hoy, 1979; Thorson et al. 1982; Pollack and El-Feghaly, 1993). Synchronous activity of groups of receptors might also provide more accurate cues for sound direction. In acridids, interaural latency differences of receptors, generated by the acoustic directionality of the auditory system and the intensity-dependence of response latency, can serve as a cue for determining sound direction (Rheinlaender and Morchen, 1979; von Helversen and Rheinlaender, 1988), and preliminary experiments indicate that this may be true for crickets as well (D. von Helversen and G. S. Pollack, unpublished observations). Determining interaural latency difference from the responses of single receptors is problematic because of the variability of latency both within and between receptors. The concerted activity of a number of receptors, however, could provide a well defined time marker for response latency in each ear (Krahe and Ronacher, 1993). Moreover, because responses to single sound pulses consist of trains of regularly spaced CAPs (Figs 1, 2), central neurons might be able to make multiple comparisons of interaural latency, once for each CAP in the response.

This work was supported by grants from NSERC and the Whitehall Foundation.