Experience-dependent modification of ultrasound auditory processing in a cricket escape response

Experience-dependent plasticity is a common characteristic of sensorimotor systems. Such simple systems have provided models for mechanisms of neural plasticity that may underlie learning, pointing to general principles independent of a particular organism. However, there is increasing interest in how mechanisms of plasticity interact within specific neural systems to affect behavior (e.g. Frost et al., 1988; Fitzgerald et al., 1990; Harris-Warrick and Marder, 1991; Vu et al., 1993; Klein, 1995) and how these interactions might be shaped by evolutionary feedback (e.g. Miklos et al., 1994; Tierney, 1996). To address these issues, we must understand the links between neural plasticity and behavioral modulation as well as the consequences for fitness.

We have studied an acoustic startle response (ASR) in the field cricket Teleogryllus oceanicus. A flying cricket exhibits a very rapid turn away from a speaker broadcasting ultrasound. The cricket ASR shows experience-dependent plasticity (habituation and dishabituation) (May and Hoy, 1991), and this has given us an important window for examining properties of auditory discrimination in an insect (Wyttenbach et al., 1996; Wyttenbach and Hoy, 1997). We know the ethological context of the cricket ASR: it is a defense against predatory bats (Moiseff et al., 1978), and it is a type of behavior pattern that appears to be common among night-flying insects (Roeder, 1967; Hoy et al., 1989). We also know a great deal about the cricket auditory system (e.g. Casady and Hoy, 1977; Wohlers and Huber, 1982; Schildberger, 1984; Atkins and Pollack, 1987; Brodfuehrer and Hoy, 1990). The clear ethological context of the behavior pattern, combined with an accessible neural substrate, makes the cricket ASR an excellent system for examining the neural mechanisms of plasticity in conjunction with their functional consequences for the organism. To lay the groundwork for these objectives, in this paper we describe how several parameters of the startle response are modified by ultrasound prepulses in a habituation paradigm.

Two aspects of the ASR, sensitivity and direction, were modified by conditioning with a train of ultrasound prepulses. Prepulses altered the intensity/response relationship for the ASR by reducing response amplitudes and increasing the saturating stimulus intensity, but stimulus threshold was not increased. Prepulses also altered the directional response to subsequent sound pulses broadcast from locations anterior to the cricket. Other experiments examined the stimulus control of ASR habituation. Habituation was only induced by suprathreshold prepulses. When prepulses were broadcast from one side, habituation showed incomplete generalization to sound pulses from the contralateral side. These results indicate that the ASR could be modified in significant ways during a predatory encounter, they provide clues about the underlying neural plasticity in the auditory pathway and they suggest that plasticity in this pathway not only affects behavior but also alters auditory perception of stimulus intensity and direction.

Adult Polynesian field crickets (Teleogryllus oceanicus LeGuillou) were collected from our laboratory colony, tethered by a rod waxed to the dorsal metathorax and flown in a laminar wind stream in an anechoic chamber. Pulses were generated, amplified and calibrated as described previously (Wyttenbach and Hoy, 1997). Stimulating loudspeakers were 0.5 m from the cricket on the left and right sides (−90 ° and +90 ° azimuths, where 0 ° is anterior to the cricket). Ultrasound pulses were 10 ms in duration, 1 ms in rise/fall time and 20 kHz in carrier frequency. In one protocol (response direction), test pulses were delivered from five speakers at 22.5 ° intervals between −45 ° and +45 ° and at a distance of 27 cm, and these test pulses were 40 kHz to match the tuning of those speakers (Wyttenbach and Hoy, 1997). Both 20 kHz and 40 kHz pulses reliably evoke negative phonotaxis (the ASR) (May and Hoy, 1991; Wyttenbach et al., 1996).

We quantified the ASR by using a photodetector to measure the lateral swing of the hindleg contralateral to the source of ultrasound (May and Hoy, 1991; Wyttenbach et al., 1996; J. E. Engel and R. A. Wyttenbach, unpublished observations). To reduce the influence of ‘noise’ in the leg position signal due to vibration from wing beats (typically 35 beats s⁻¹), leg swing amplitudes were measured by comparing the greatest lateral deflections during baseline and response periods. The baseline period began 30 ms before stimulation and lasted 50 ms (for the first prepulse in a bout, the baseline period began at 0 ms and lasted 20 ms). The response period began 50 ms after stimulation and lasted 150 ms.

Trials began with a preliminary estimation of threshold for 20 kHz pulses from −90 ° and +90 °, hereafter called the ‘preliminary threshold’. The prepulse protocols used here were adapted from a habituation paradigm (May and Hoy, 1991; Wyttenbach et al., 1996; Wyttenbach and Hoy, 1997) in which a test pulse is given alone (control treatment) or 750 ms after a train of five prepulses with a 750 ms interstimulus interval (prepulse treatment) (Fig. 1A). Five prepulses usually produced an asymptotic level of habituation (Fig. 1A). For these experiments, we varied test pulse intensity, prepulse intensity, test pulse azimuth or prepulse/test pulse lateralization as described below. Stimulus bouts were always separated by 1 min intervals. Results from repeated presentations of a given prepulse treatment were averaged, then normalized by scaling them to the greatest averaged control-treatment test response of that trial. Statistical tests were performed using StatView 5.0 for Macintosh (SAS Institute) except for the sequential Bonferroni correction for multiple comparisons (Rice, 1989).

For intensity/response and prepulse intensity protocols, it was important to distinguish just-suprathreshold responses from ‘noise’ (fluctuations in leg position caused by flight vibration or other sources). To accomplish this, a detection criterion was determined for each trial in these protocols. In intensity/response trials, for which responses were averaged from four presentations of each treatment, we measured sham responses (with no stimulus) 750 ms before the test pulse in control bouts (Fig. 1A) and averaged the four measurements corresponding to each test pulse intensity. This gave the same number of averaged sham responses as averaged control test responses (9–14 per trial). The detection criterion for a trial was the grand mean plus twice the interquartile range of the sham responses. This level ranged from 0.017 to 0.092 of the saturation level (the control-treatment maximal response) for each trial (see, for example, Fig. 1B). The detection criterion was greater than the largest averaged sham response in eight of the ten trials.

For analysis of prepulse intensity trials, it was important to identify suprathreshold responses in individual recordings (not averaged recordings as for the intensity/response protocol). For each trial, five ‘sham responses’ were measured at 750 ms intervals preceding the test pulse for the four control bouts as well as for prepulse intensities 6 dB and 4 dB below the preliminary threshold. This gave a sample of 60 sham responses for a trial, and the detection criterion was defined as the grand mean plus twice the interquartile range of sham responses.

This protocol defined the stimulus intensity/response strength relationship and its modification by prepulses. Prepulse intensity differed between trials because of differing preliminary thresholds, but was held constant within a single trial as test pulse intensities were varied. Prepulses were approximately 20 dB above the preliminary threshold. Test pulses ranged from below the preliminary threshold to approximately 30 dB above it in 4 dB steps (2 dB steps close to threshold). Each test pulse intensity was presented four times in both control and prepulse treatments (two ascending and two descending series), and the four replicates were averaged for analysis.

We calculated three parameters for each trial. Maximal response was the greatest prepulse-conditioned response (normalized to the greatest control response). ‘Maximal response’ is not synonymous with ‘saturation’ because in some trials the prepulse-conditioned response did not appear to reach a plateau within the range of intensities used. To test for habituation, maximal responses from all trials were compared with 1 (this being the normalized control-treatment maximum) using a one-sample two-tailed t-test. Half-maximal stimulus was the pulse intensity that gave half the maximal response for each treatment, interpolated from the two pulse intensities bracketing the 0.5 response level. Stimulus threshold was the pulse intensity sufficient to give a response above the detection criterion, interpolated from the pair of intensities for which only the larger evoked a detectable response. To test for prepulse effects on half-maximal stimulus and stimulus threshold, prepulse and control treatments were compared using paired-comparisons analysis of variance (ANOVA). Test results for the two parameters were then grouped for the Bonferroni correction for multiple comparisons.

To combine the intensity/response curves for plotting, stimulus intensity scales for individual trials were calibrated by defining 0 dB to be the midpoint of the 4 dB interval that spanned the control-treatment stimulus threshold. For example, for the trial shown in Fig. 1B, 60 dB SPL would become 0 dB. This procedure allowed intensity/response curves from all ten trials to be matched at 4 dB intervals and averaged for plotting.

This protocol examined the relationship between prepulse intensity and habituation with test pulses of constant intensity, in part to determine the smallest prepulse intensity that could induce habituation. Prepulse intensities spanned a range from 7 dB below the preliminary threshold to 9 dB above threshold in 2 dB increments. Weak test pulses were used, 4–6 dB above the preliminary threshold, to provide a sensitive indicator of habituation. Each prepulse intensity was repeated four times, in both ascending and descending series, and there were four no-prepulse control bouts in each trial.

To combine trials for analysis, prepulse intensities were scaled to stimulus threshold for each trial. Detection criteria for this protocol were determined from non-averaged responses, and stimulus threshold for a trial was defined as 1 dB below the intensity at which a detectable response was evoked by the first prepulse in at least two of the four bouts. To calculate the frequency of responses to prepulses at each intensity, we counted the number of responses to all 20 prepulses of each intensity (five prepulses times four replications) and averaged across all trials. Prepulse intensities ranged from −7 dB to 9 dB at 2 dB intervals, which would give 72 values for the eight trials. Of these, 14 values were missing as a result of using 4 dB prepulse intervals in some early trials. These were filled by interpolation before combining results for plotting, but not for statistical analysis.

To measure habituation, mean test response amplitudes in each trial were normalized to the control-treatment test response, and test responses for each prepulse intensity were then averaged across trials. To test the significance of habituation, test pulse responses for the nine prepulse intensities were compared with 1 (the control-treatment value) using two methods: Dunnett’s test for multiple comparisons, and one-sample two-tailed t-tests for each of the nine prepulses intensities with the results grouped for the sequential Bonferroni correction for multiple comparisons.

This protocol characterized the influence of strongly directional (90 °) prepulses upon the lateral direction of responses to weakly directional (near-midline) test pulses. Prepulses 20 dB above preliminary thresholds were delivered from speakers at L90 ° or R90 ° (i.e. −90 ° or +90 °). Test pulses, 10 dB above the average of preliminary thresholds measured for L90 ° and R90 °, were presented from five azimuths evenly spaced between −45 ° and +45 °. To avoid the effects of treatment order, each test azimuth was presented in 18 bouts, grouped into six three-bout blocks corresponding to the six possible sequences of three prepulse treatments (−90 °, +90 ° and none). The 30 blocks in a trial (six blocks times five azimuths) were ordered randomly. Bouts were separated by 1 min as usual, whether within or between blocks. Test pulse responses were scored as ‘rightward’ or ‘leftward’ from photodetector recordings and video monitoring of the cricket. If the response direction could not be interpreted (because of small response size, i.e. habituation), the bout was repeated after 1 min.

The frequency of rightward test responses was plotted as a function of test pulse azimuth. Rightward response frequencies for the three prepulse treatments were compared at each of the five test pulse azimuths using the nonparametric Friedman test with correction for ties (StatView), with trial as the repeated-measures or block variable. Results for all five azimuths were then grouped for the sequential Bonferroni correction for multiple comparisons.

This protocol examined the generalization of habituation across sides. Prepulses from either L90 ° or R90 ° were followed by a test pulse from the same side (ipsilateral pairing) or the opposite side (contralateral pairing). Four combinations of pulse direction (L90 ° or R90 ° prepulses, L90 ° or R90 ° test pulse) were replicated eight times within a trial (the order of presentation being systematically varied). The first prepulse also served as a control-treatment test pulse. Responses for L90 ° and R90 ° test pulse directions were combined and normalized to the averaged control response for that trial, giving two results for each trial (for ipsilateral and contralateral pairings). To test for habituation by each pairing treatment, ipsilateral-paired and contralateral-paired responses were compared with 1 (the control response value) using one-sample two-tailed t-tests, and the two test results were grouped for Bonferroni correction. To test for stimulus generalization across sides, habituation was compared between ipsilateral- and contralateral-pairing treatments using paired-comparisons ANOVA.

A large response to prepulses can briefly leave the prepulse-ipsilateral hindleg in a more medial position, which might have interfered with measurements of the response of that hindleg to a subsequent contralateral test pulse. We used low-intensity prepulses to minimize this effect, and used test pulses of the same intensity so that the first prepulse of each bout could also serve as a control-treatment test pulse. At the start of a trial, pulse intensities from each side were adjusted to give strong habituation in ipsilateral-pairing treatments. This ensured that pulses from each side were able both to induce habituation and to exhibit habituation in ipsilateral pairings, in order to provide a sensitive test for generalization of habituation in contralateral pairings.

To examine how habituating prepulses affect the stimulus intensity/response strength relationship, we measured response amplitudes for a series of test pulses over a broad intensity range encompassing threshold and control-treatment saturation (Fig. 2). Stimulus thresholds in the absence of prepulses (control treatment) ranged from 54.1 to 68.1 dB. As expected, the five-prepulse habituation protocol attenuated the test response (Fig. 2). The maximal prepulse-conditioned response was reduced to 78±13.2 % of the control response (P<0.0001, t-test, N=10) (unless stated otherwise, all values are means ± S.D.). Attenuation was greatest for lower-intensity test pulses and less pronounced for more-intense stimuli, a typical characteristic of habituation (Thompson and Spencer, 1966). This had the effect of extending the dynamic range of the intensity/response curve to higher intensity levels (Fig. 2). The stimulus intensity giving half the maximal response increased from 64.7±3.83 dB for controls to 78.0±4.05 dB for prepulse treatments (P<0.001, ANOVA, N=10). In contrast, stimulus threshold did not increase significantly (Fig. 2). Thresholds were 60.5±4.33 dB for the control treatment and 61.0±4.17 dB for the prepulse treatment (P>0.05, ANOVA, N=10).

The flattening of the prepulse-treatment slope, when intensity/response data are combined as in Fig. 2, corresponds to a positive shift in half-maximal stimulus intensity and an unchanging stimulus threshold. A more detailed picture emerged when individual trial recordings were examined. Some recordings showed two distinct slopes in the prepulse-treatment curve: a shallow increase in response amplitude above threshold giving way to a steeper slope at higher intensities before finally flattening at saturation (Fig. 1B).

Because of the comparatively slight effect of prepulses upon stimulus threshold, we examined the relationship between habituation and threshold from another perspective by asking whether subthreshold prepulses could induce habituation. For this protocol, test pulse intensity was held constant at a level just above threshold, where responses are most sensitive to habituation (as shown in Fig. 2). Prepulse intensity was varied over a range that encompassed threshold (Fig. 3A). Fig. 3B shows the mean test response amplitude (open circles) and also the frequency of detectable responses to the five prepulses (filled circles) to indicate the relationship between habituation and prepulse activation of the ASR neural pathway (thresholds do not provide this information because for this protocol they were calculated using responses to the first prepulse only).

Habituation was significant for all prepulse intensities above threshold (Dunnett’s test; for t-tests with Bonferroni correction, the 1 dB point just missed significance). There appeared to be some attenuation of test responses at prepulse intensities up to 3 dB below threshold (Fig. 3B, open circles), at which level the prepulses themselves very rarely evoked detectable responses (Fig. 3B, filled circles). This habituation was not statistically significant according to either of the conservative multiple-comparison tests we applied. In any case, it is clear that thresholds for inducing habituation and for evoking prepulse responses matched to within a few decibels.

The previous experiments examined the sensitivity of the cricket acoustic startle response. The lateral directionality of the ASR gives a window for examining the plasticity of sound localization, a process of discrimination.

The ASR is always directed towards left or right, and when the ultrasound source is at 90 ° the response is always contralateral to the sound source. We tested the direction of responses to test pulses from speakers in a 90 ° arc in front of the cricket (Fig. 4A). Test pulses from L45 ° and R45 ° nearly always evoked contralateral responses (rightward leg swings with test pulses from L45 ° and leftward leg swings with test pulses from R45 °) (Fig. 4B). However, for L22 °, 0 ° or R22 ° test pulse azimuths, rightward and leftward responses were mixed, both in the combined results (Fig. 4B) and in individual trials. This suggests that azimuths between −22 ° and +22 ° are laterally ambiguous to the cricket under these stimulus conditions. We asked how 90 ° prepulses would affect the direction of responses to test pulses in this ambiguous range.

Prepulses from ±90 ° altered the distribution of response directions for near-midline test pulses between −22 ° and +22 ° azimuth (P<0.05, Friedman test with Bonferroni correction, N=9). This is the same azimuth range over which response direction is unpredictable in the absence of prepulses (Fig. 4B). Prepulses from L90 ° increased the likelihood of rightward responses to test pulses, and prepulses from R90 ° increased the likelihood of leftward responses. The 90 ° prepulses themselves always evoked contralateral responses, which means that the test responses were primed to occur in the same direction as the preceding prepulse responses. Interestingly, a tendency to turn in the direction of the most recent prior response has also been noted in a different paradigm (Wyttenbach and Hoy, 1993) in which the directionally ambiguous test stimulus was a pair of pulses delivered simultaneously from L90 ° and R90 °.

The priming of response direction suggested that habituating prepulses from 90 ° affected the prepulse-ipsilateral and -contralateral sides of the ASR pathway to different degrees. Such asymmetrical modification could have led to ‘priming’ by altering the relative activation of the two sides by a subsequent near-midline test pulse enough to change the direction of the test response in some proportion of tests. To test the lateral symmetry of habituation, we compared the degree of habituation induced when prepulses were ipsilateral versus contralateral to 90 ° test pulses (‘ipsilateral pairing’ and ‘contralateral pairing’; Fig. 5A).

Pulse intensities were selected for the capability to display habituation in the ipsilateral pairing treatment (see Materials and methods). Indeed, ipsilateral prepulses induced strong habituation, attenuating test responses to 28±6.3 % of control levels (P<0.001, t-test, N=9) (Fig. 5B). Contralateral prepulses also induced habituation, attenuating test responses to 53±13.6 % (P<0.001, t-test, N=9). However, the difference in habituation by the two treatments was significant (P=0.0001, ANOVA, N=9). This indicates that habituation generalizes partially but not completely across sides and is strongest on the prepulse-ipsilateral side.

We have shown that ultrasound prepulses cause short-term changes in sensitivity and response direction in the cricket acoustic startle response (ASR). Lability of this kind could alter the effectiveness of the ASR as an escape response. Experience-dependent modification of an acoustically evoked behavior may also point to malleability of auditory perception in the cricket. In addition, these experiments offer insights into the organization of neural plasticity in the cricket auditory pathway. Our results provide a basis for further examination of plasticity in a sensorimotor system from both ethological and neurophysiological perspectives.

Several studies have described habituation to predator-associated signals in prey animals of various taxa. While habituation may sometimes benefit predators (Arnold, 1971; Fleishman, 1986; Callahan, 1993; Smale et al., 1995), it may also benefit prey animals by reducing false alarms or increasing the ability to distinguish real threats from innocuous stimuli (Shalter, 1978, 1984; Magurran and Girling, 1986; Urfi et al., 1996). The cricket ASR should be a particularly good system for examining the functional significance of behavioral plasticity in a predator-avoidance response, because both the sensory target (bats) and the behavioral goal (turning) are clearly defined.

Prepulse modification of the ASR could very well affect the ability of the cricket to escape predation and, because of the implications for fitness, it is reasonable to consider scenarios in which this lability could be adaptive. Prepulses reduced the slope, or gain, of the intensity/response relationship, but the maximal response was only moderately reduced (Fig. 2). A cricket might therefore continue to respond strongly to a bat remaining in pursuit, while responding less strongly to weak stimuli that do not indicate a continuing threat. Prepulses also enhanced the tendency to respond in the same direction as previous responses (Fig. 4). This (together with diminished response amplitude) might benefit a cricket by reducing the likelihood of interrupting a turn-in-progress with a new turn in the opposite direction.

How informative are such scenarios? The prepulse paradigm in these experiments has ethological relevance because echolocating bats emit repeated cries, and the leg swing is an appropriate behavioral indicator because it contributes to turning in flight (May and Hoy, 1990). However, sensory signals in real predatory encounters would be more complex (Shalter, 1984). In encounters with bats, ultrasound pulse intensity, pulse direction and interpulse interval could all vary from moment to moment. Echolocation pulses can also be frequency-modulated, although there is no indication that crickets discriminate between ultrasound frequencies (Wyttenbach et al., 1996). More complex stimuli could modify the ASR in ways not seen here. For instance, we have previously shown that a change in pulse direction can partially dishabituate the response (May and Hoy, 1991; Wyttenbach and Hoy, 1997). The present work describes two types of experience-dependent modification that might occur during predatory encounters. Further experiments, in which pulse intensity, direction and timing are varied systematically with reference to the events in actual predatory encounters, will give a better picture of the ethological consequences of this lability.

Perception is dynamic. Continual modification by experience is crucial both for calibration in a changing environment (e.g. perceptual constancy) and for detecting salient features of that environment (e.g. selective attention). Behavioral experiments in animals can contribute to our understanding of perception, especially when ‘perception’ is broadly defined as the attribution of properties to an object based upon sensory energy. Psychophysics examines perception and sensation through the measurement of behavioral responses (which may include verbal responses in humans). We have previously applied psychoacoustic methods to the cricket ASR to demonstrate categorical perception and a precedence effect, and to measure the acuity of sound localization (Wyttenbach and Hoy, 1993, 1997; Wyttenbach et al., 1996).

Our present behavioral results may provide a window into the lability of auditory perception in the cricket. The prepulse-induced alteration of the intensity/response curve (Fig. 2) could reflect a change in the perceived loudness of suprathreshold pulses, which might even improve the ability of the cricket to judge the distance and velocity of a bat whose cries fell upon the dynamic slope of the altered curve. Strongly directional prepulses could modify the perceived location of subsequent near-midline pulses, so that the shift of the distribution of response directions (Fig. 4B) would, in effect, be an index of perceptual bias.

What might our results imply about processes of perception? An important consideration is the stage of processing at which plasticity occurs. It may be difficult to draw a sharp line between sensory processing and motor systems in a cricket, but if habituation of the cricket ASR were mediated at a neuromuscular junction (for example), this would surely not be thought of as affecting perception. A process that modifies perception should alter behavioral choices that depend upon the properties of an object (such as the decision to turn left or right based upon the apparent location of a predator). This could be tested behaviorally if paradigms could be developed to reveal the cricket’s evaluation of such properties (e.g. Wyttenbach et al., 1996; Wyttenbach and Hoy, 1997) and especially if such evaluation could be assayed using multiple independent paradigms. A process that modifies perception should also be mediated earlier in the neural pathway than behavioral choices that depend upon the properties of an object. This leads to a consideration of what our results can say about the neural substrates of prepulse modification of the cricket ASR.

From our behavioral results, we can draw inferences about the relationship between neural plasticity and other parameters of the ASR response, which can be compared with what is known of the physical pathway, allowing hypotheses to be generated that should be testable using neurophysiological approaches. There are two principal sets of conclusions. First, habituation is mediated later than stimulus threshold and, therefore, the neural substrate of habituation is probably in the brain. Second, priming of response direction must be due to a process of plasticity that is laterally asymmetrical. This cannot be the same mechanism that mediates habituation, because habituation would have to be strongest on the prepulse-contralateral side, which is the opposite to what was found. Priming of response direction points to a second process of plasticity in the pathway, possibly facilitation.

Habituation is mediated later than stimulus threshold. If habituation were mediated before threshold, the neural signal arriving at the threshold ‘gate’ would be attenuated, so stimulus threshold would increase. Conversely, if habituation were mediated after threshold, stimulus threshold would not change but attenuation later in the pathway would reduce response amplitudes (Pilz and Schnitzler, 1996). Our observations support the latter alternative (Fig. 2). Furthermore, if habituation were mediated before threshold, then the neural mechanism underlying habituation could be activated by subthreshold prepulses. However, we found that the smallest prepulse intensities that induced habituation were very close to stimulus thresholds (Fig. 3).

The neural substrate of habituation is probably in the brain because the brain is where threshold appears to be mediated. The primary carriers of ultrasound-induced signals from the prothoracic ganglion to the brain in the ASR pathway are the ascending fibers of the paired Int-1 interneurons (Casady and Hoy, 1977; Moiseff and Hoy, 1983; Nolen and Hoy, 1984). Rapid spiking of Int-1 is both necessary and sufficient to induce the ASR during flight (Nolen and Hoy, 1984). However, the Int-1 fibers spike less rapidly in response to ultrasound stimuli well below behavioral threshold. Hence, the threshold ‘gate’ probably lies in the brain (or in the subesophageal ganglion, where Int-1 also has terminal branches). Habituation could formally be mediated even later in the ASR pathway, after the brain. This might be revealed by further experiments to compare habituation of different motor elements of the ASR response.

Priming of response direction must be due to a process of plasticity that is laterally asymmetrical. The direction of an ASR response is presumably determined by the relative activity of the right and left sides of the neural pathway. (Here, ‘side’ means the portion of the circuit that is activated most strongly by stimuli from one half of auditory space.) A pulse from near the midline should reach both ears with similar strength, arrival time and phase. At some point during processing, slight differences in signal intensity must be enhanced so that the activity of the right and left sides differs enough to drive a unidirectional response. That point of lateral ‘contrast enhancement’ is effectively a directional comparator. The unpredictability of response direction for near-midline pulses (Fig. 4) suggests that the two sides of the pathway are activated so similarly that slight changes in relative signal strength (i.e. ‘noise’) can alter the balance of signals reaching the comparator. A system with these characteristics could be very sensitive to priming by prepulse-induced plasticity with even a modest degree of lateral asymmetry.

The plasticity underlying priming cannot be the same process that mediates habituation. Priming increased the likelihood of near-midline pulses evoking responses in the prepulse-contralateral direction (Fig. 4), which is the same response direction elicited by the prepulses themselves. This must result either from strengthening of the side of the pathway that prepulses activate most strongly (the prepulse-ipsilateral side, which mediates contralaterally directed responses) or from weakening of the side of the pathway that prepulses activate least strongly (the prepulse-contralateral side). However, we found that habituation was actually greatest on the prepulse-ipsilateral side of the pathway (Fig. 5). Since this is inconsistent with the direction of priming that we observed, habituation (or at least the component of habituation that is laterally asymmetrical) must be mediated after the directional comparator or in parallel with it.

Priming of response direction therefore points to a second process of plasticity in the ASR pathway. This could be a process of attenuation (distinct from habituation) that affects the prepulse-contralateral side of the circuit. It seems more parsimonious, however, to suggest that facilitation occurs in the prepulse-ipsilateral side of the pathway. In either case, this process of plasticity could be much weaker than habituation in absolute terms, yet still influence response direction through asymmetrical modification of the inputs to a directional comparator. This could explain why we did not observe sensitization of the response or relative strengthening of the prepulse-ipsilateral side of the pathway in intensity/response or stimulus generalization experiments (Figs 2, 5).

The behavioral results argue strongly that habituation and priming of response direction are mediated by two distinct mechanisms of prepulse-induced plasticity. This may indicate that stimulus amplitude is processed independently from direction (or at least left/right lateralization) in the cricket auditory system. There is evidence that sound localization can involve physical and neural processes at several points in the auditory pathway, including the ears (Michelsen et al., 1994), the prothoracic ganglion (Schildberger and Hörner, 1988; Horseman and Huber, 1994) and the brain (Wyttenbach and Hoy, 1993, 1997). It is clear that the neural plasticity that mediates priming of response direction would have to precede the final stage at which response direction is determined (the hypothetical ‘directional comparator’), but it is not clear where that stage is. Identifying the neural substrate of priming will give us more information about directional processing in the auditory pathway. If priming is due to facilitation, then this facilitation (and hence the directional comparator as well) may be located in the brain because prepulses did not lower stimulus thresholds (Fig. 2). An alternative, which has not been excluded, is that priming of response direction arises from proprioceptive feedback or some other effect of prepulse responses, rather than from physiological modification of the ASR pathway as posited here.

The organization of the central nervous system in crickets suggests an immediate neurophysiological approach to localizing the physiological substrates that underlie habituation and priming. Auditory signals ascending to the brain pass through the neck connectives, where the central nervous system is divided both longitudinally (thoracic ganglion versus brain/subesophageal ganglion) and laterally (left versus right cervical connectives). Recordings from Int-1 fibers in the cervical connectives should allow testing of several hypotheses: that habituation occurs in the head rather than in the thorax, that near-midline pulses induce similar ascending activity in the left and right connectives, and that 90 ° prepulses alter relative activity levels on the left and right sides. Together with additional behavioral experiments, this would provide a foundation for intracellular recordings of identified auditory neurons with the aim of locating and characterizing mechanisms of plasticity that are relevant to functionally significant behavioral modification in a predator-avoidance response.

We thank Paul Faure, Andrew Mason and Robert Wyttenbach for helpful comments. This work was supported by NIH grant R01-DC00103 to R.R.H.