Who, What, Where? Recognition and Localization of Acoustic Signals by Insects Gerald Pollack

763

Who, what, where? Recognition and localization of acoustic signals by insects Gerald Pollack

Insects, like all hearing animals, must analyze acoustic signals Figure 1 to determine both their content and their location. Neurophysiological experiments, together with behavioral tests, are beginning to reveal the mechanisms underlying these signal-analysis tasks. Work summarized here focusses on two Gryllodinus odicus Uv. issues: first, how insects analyze the temporal structure of a single signal in the presence of other competing signals; and second, how the signal's location is represented by the Pteronemobius heydeni Fish. binaural difference in neural activity.

Addresses Tartarogryllus bucharicus B.-B. Department of Biology, McGill University, 1205 Avenue Dr Penfield, Montreal, Quebec H3A 1B1, Canada; e-mail: [email protected] Tartarogryllus tartarus obscurior Uv.

Current Opinion in Neurobiology 2000, 10:763–767

0959-4388/00/$ — see front matter Tartarogryllus burdigalensis Latr. © 2000 Elsevier Science Ltd. All rights reserved. Introduction The questions posed in the title refer to some of the core Oecanthus pellucens Scop. issues in animal communication. The recipients of a sensory signal need to know who produced the signal, what the signal’s content is, and where the signal originated. Sound, Gryllus campestris L. being both long-range and information-dense, is an excel- lent medium for communication. The information content of the signal is represented by structural features of the Gryllus bimaculatus Degeer sound such as its spectrum and pattern of amplitude modulation. The recipient’s analysis of the message’s content (who? what?) and of its location (where?) must be done in Modicogryllus pallipalpis Farb. real time, often when several signals arrive simultaneously from different sources. These are daunting computational tasks, and understanding how they are performed by ner- Gryllus truncatus Farb. 0.2 s vous systems is a challenge for neurobiology. Current Opinion in Neurobiology

Acoustic signaling is widespread among vertebrates but, The diversity of temporal patterns of cricket songs. The songs among invertebrates, hearing and acoustic communication illustrated are those of several European and Asian species. From [8]. are well developed only in insects, in which they serve behavioral functions such as detection of predators, location of mates, and location of hosts (for reviews, see [1–3]). rare exception [6], lack the ‘melodies’ that characterize Insects offer several advantages as model systems for neu- many vertebrate communication signals. The ears of many roethological studies, including robust behavior, easily insects are capable of spectral analysis, and there are sever- accessible nervous systems, and uniquely identifiable neu- al examples in which a signal’s spectrum may be rons that permit one to frame general questions about the informative (see [2,7] for reviews), but in most instances the neural analysis of signals at the levels both of single nerve main information-carrier in an insect’s song is its temporal cells and of the circuits they form [4,5]. In this review, I structure — in other words, the durations and shapes of will discuss some recent behavioral and neurophysiological sound pulses, the spacing between them, and their organi- experiments carried out on insects that are directed at the zation into higher-order groupings (Figure 1; [8]). problems of signal recognition and localization. Behavioral experiments, using synthetic song models, have outlined the properties of the temporal-pattern filters that Temporal pattern as a carrier of information analyze the signals; however, the neuronal implementation Insect songs generally consist of rhythmic sequences of of these filters is understood only in broad outline. more-or-less similar brief sounds (sound pulses), and, with Temporal-pattern filtering is mainly a central phenomenon; 764 Neurobiology of behaviour

Figure 2 Figure 2 Selective attention to a single signal among many. S1, S2 and S3 represent three singers at different distances from the receiver, R. Because of the differences in distance, their songs arrive at R at different intensities, as illustrated in the oscillograms. The trace labeled S1 S1–3 shows what a microphone at the position of the receiver would detect (i.e. the summation of the three signals). The bottom trace represents the membrane potential of an interneuron, such as those described in [15,16,17•], that receives both fast-acting excitation and slow inhibition. Inhibition has lowered the neuron’s baseline membrane potential below its resting level (resting membrane potential, rmp; S2 represented by the arrow), with the result that only the largest sound- evoked excitatory postsynaptic potentials, produced by the loudest sound pulses, are sufficient to bring the membrane potential above threshold (dotted line). As a result, S1 and S2 are filtered out of the neuron’s spiking response.

The anatomical relationships and relative response latencies of the different types of filter neurons suggest that band-pass selectivity is synthesized by combining inputs S3 from lower-level low- and high-pass filters [9]. Recent work suggests that even receptor neurons may, in some instances, play a role in temporal-pattern filtering. Many insect songs consist of long sequences of rapidly repeated sound pulses. Such stimuli induce sensory habituation — a decline in R responses to successive sound pulses within the sequence. The higher the pulse rate, the more pronounced the habituation; consequently, the receptor neurons act as low-pass temporal filters [10••]. This mechanism can only operate in species with songs that are conducive to habituation — long S1 sequences of rapidly repeated pulses. Such songs are, however, common. Despite these basic low-pass filter S2 properties of the receptor neurons, the general conclusion, that strong temporal-pattern selectivity paralleling the selectivity of behavioral responses emerges only at higher S3 levels of neural circuitry, remains essentially correct.

Providing high-quality input to the temporal- pattern filters S1–3 Insects often sing in aggregations [11–13] in which a listener may be within earshot of several singers. In some species, individual signalers may avoid mutual interfer- ence by singing either in alternation or in synchrony with others [12], but in other species the listener is faced with the problem of teasing a single signal out of apparent Threshold cacophony (the ‘cocktail party’ problem). Crickets and rmp katydids solve this problem with a form of selective attention [14–16,17•]. Some of their low-order auditory Current Opinion in Neurobiology interneurons receive mixed excitatory and inhibitory input. The excitation has a rapid time course, allowing the receptor neurons and low-order interneurons generally neurons to respond crisply and accurately to the pattern of copy the temporal pattern of the stimulus with high fideli- amplitude modulation. The inhibition, which is due in part ty. At higher levels, interneurons are found that respond to a Ca2+-activated current [18] (most probably, a K+ cur- most strongly only to particular stimulus temporal patterns. rent) is more leisurely, building up and, after the stimulus In the cricket Gryllus bimaculatus, for example, the first indi- has terminated, decaying over a period of several seconds. cation of temporal filtering is in brain neurons that are two The inhibition serves to down-regulate the neuron’s sensi- to three synapses removed from receptors. These neurons tivity such that signals of lower intensity are filtered out. show low-, high- or band-pass filter properties. The band- As a result, the temporal pattern of the most intense signal, pass-filter neurons respond best to the same range of corresponding to the loudest (or nearest) singer, remains temporal patterns that best evoke behavioral responses. relatively uncontaminated (Figure 2); thus a ‘clean’ copy of Recognition and localization of acoustic signals by insects Pollack 765

the signal can be passed on to the higher-level filter Figure 3 circuits for further processing.

This same mechanism explains the remarkable ability of 1.0 t=0s these insects to discriminate reliably between simultaneous t=10s signals originating on the two sides [14,15,17•,19], a situa- 0.8 Ipsilateral tion that they presumably encounter frequently in the field. Each of the stimuli, when presented by itself, is detectable Contralateral by both ears, and so one might suppose that the simultane- 0.6 ous presence of many stimuli would result in their Relative response obscuring each other’s temporal pattern. However, sound 0.4 intensity at the two ears will differ, depending on the direction of the sound source [20]. This auditory directionality, together with the slow inhibition described above, enables 0 10 20 30 (s) neurons on each side of the nervous system to ‘listen’ selec- Current Opinion in Neurobiology tively to the sound played on their side. The result is that distinct copies of the two songs can coexist in the nervous Intensity-dependent habituation may reverse the sign of the binaural system, one on each side. Presumably, these can then be difference in response strength. The inset shows a recording from the analyzed by recognizer circuits, also situated separately on auditory nerve of a cricket to a high-repetition-rate, high-intensity sequence of sound pulses (bottom trace). The top trace shows each side of the nervous system, with the behavioral out- responses immediately after onset of stimulation, and the second trace come (preference for one or the other of the songs) after 10 s of stimulation have elapsed. Responses to successive sound determined by comparing the outputs of the recognizers; pulses are quantified in the graph by integrating the nerve recording the side receiving the ‘best’ signal pattern wins [14,21]. over an appropriate time window surrounding each sound pulse. The points show responses of a single nerve to stimuli played from the ipsilateral and contralateral sides; this approximates the responses of The code for sound direction the two ears to a single stimulus played from one side. Initially, the In most instances of communication, simply recognizing ipsilateral response is greater than the contralateral response, as a the signal is not enough — it must also be localized (e.g. so result of auditory directionality. However, because stimulus intensity is greater ipsilaterally, the ipsilateral response experiences greater that the recipient of a mate-attraction signal can approach habituation and, after approximately eight seconds of stimulation, it falls the sender). Sound localization in insects depends on com- below the contralateral response. Modified from [10••]. parison of sound intensity at the two ears (see [20] for a review of the physical mechanisms responsible for gener- ating binaural intensity differences). approximately 1 s, is truncated to one quarter of its normal length [24], thus severely limiting the grasshopper’s abili- Binaural cues ty to use time-averaging of neural activity to increase How is binaural intensity difference represented by audi- acuity. How, then, is binaural intensity difference encod- tory neurons? The intensity level of the stimulus affects ed? This difference is, of course, accompanied by a two aspects of neural responses: the response strength, that binaural difference in latency, which may be as much as is the number and/or rate of action potentials, increases 5 ms [22]. Are the turns driven by differences in response with increasing intensity; and the response latency strength (spike counts of receptor neurons) or in latency? decreases with increasing intensity. Binaural differences in Ronacher and Krahe [25•] approached this question by either or both of these response properties might serve as measuring the response variability of single receptor neu- the code for localization (note that here we are speaking of rons and using these data to model the probability with a binaural difference in the timing of neural activity, which which bilateral pairs of receptor neurons would correctly may amount to several milliseconds [22], not the time-dif- ‘judge’ the sign of small binaural intensity differences. ference in arrival of sound at the ears which, even for a They found that latency was more variable than spike relatively large insect such as a field cricket, is at most count and, as a result, the model’s performance was much about 30 µs). Several recent studies, focusing on sound poorer (error rate up to 4.6 times as high) if the determi- localization by grasshoppers and crickets, have addressed nation of sound direction was based on latency the code for sound direction. comparison rather than comparison of spike counts. However, even the spike-count difference was unable to Male grasshoppers turn towards the side from which a account for the performance of the grasshopper if only a female’s song is broadcast with exceptional acuity and single pair of receptors was considered; performance of reliability [23]; a binaural intensity difference of only the model comparable to that of the animal was obtained 1.5 dB, corresponding to a deviation of the source from only by pooling the spike counts from 6–13 pairs of recep- the midline by only a few degrees, is sufficient to evoke tors. Presumably, if the decision were based on binaural turns in the correct direction with >80% reliability. This latency comparison, input from an even larger number of performance is even more remarkable given that it is receptors would have to be pooled; however, Ronacher essentially unaffected when the natural song, which lasts and Krahe did not examine this. 766 Neurobiology of behaviour

Habituation affects localization cues depends mainly on response-strength difference. This has Givois and Pollack [10••] studied the encoding of localiza- interesting implications for our understanding of the forces tion cues by auditory receptors of crickets, focusing on the driving the evolution of song structure and song prefer- ways in which binaural differences are affected by the ences. In crickets, songs are produced only by males, and temporal pattern of the stimulus. The song of the species females analyze them not only to determine the species- they studied, Teleogryllus oceanicus, consists of long-lasting identity of the singer, but also to evaluate his genetic (of the order of minutes) trains of rapidly occurring sound quality. Several studies suggest that females prefer songs pulses. As described above, auditory receptors habituate that are more energy-demanding to produce (i.e. which when challenged with such stimuli. The response decre- have longer-duration sound pulses [29], longer-length ment increases with increasing stimulus intensity, and this sequences of pulses [30], or higher pulse rates [31]), all of has important consequences for sound localization. which presumably reflect on the singer’s health, his ability Because intensity is greater at the ear ipsilateral to the to acquire resources, and his overall genetic quality. These sound source, habituation is more pronounced in receptor female preferences, in turn, exert selective pressure favor- neurons serving that ear. As a result, the binaural differ- ing the evolution of longer pulses, higher pulse rates, and ence in response strength decreases as habituation builds longer-lasting pulse sequences. However, the song fea- up. Under extreme conditions, the difference may decline tures that best demonstrate male quality are the same to zero, or even reverse in sign (Figure 3). Clearly, this is features that most effectively produce habituation. This is potentially problematic for sound localization. The cellular thus a potentially paradoxical situation; if songs were to basis for habituation has not yet been studied in insect evolve so as to maximally demonstrate male quality, the auditory receptors but, on the basis of findings in other males singing them would be difficult to find. It is of little arthropod sensory systems, it seems likely that the causes use for a female to know that a prize male is out there are fundamental features of nerve cell biology, for example somewhere, but to be unable to locate him. It seems like- Na+-channel inactivation and Na+/K+-ATPase activity ly, instead, that the need for females to locate the singer [26,27]. Habituation may thus be an unavoidable conse- exerts a counteracting selective pressure, which limits the quence of the intense firing activity that typifies these duration rate and sequence-length of sound pulses. sensory neurons (firing rates of 300–500 Hz are not uncom- mon; K Imaizumi, G Pollack, unpublished data). Conclusions and prospects A strong advantage of insects as model systems for the Binaural latency comparison may provide a way around the study of signal analysis is that clear behavioral functions localization problem imposed by habituation. Like response can be ascribed to relatively few identified neurons. In strength, response latency also changes for rapidly repeated crickets, for example, three low-level interneurons on each sound pulses, becoming longer for successive sound pulses. side of the nervous system have been shown, through However, in contrast to the change in response strength, the direct manipulation of their activity, to be important play- change in latency is independent of intensity [10••]. Binaural ers in acoustic orientation responses [4,5]. This latency difference thus remains constant, and is immune to experimentally favorable situation can now be exploited in the effects of habituation. Interestingly, latency comparison two ways. First, most of what we know of these neurons is the cue for sound direction in a recent robotic model of comes from laboratory experiments performed under pris- cricket phonotaxis [28]. It remains to be seen, however, tine and, therefore, entirely unrealistic acoustical whether crickets actually use latency difference as the main conditions. We need to know what challenges the neurons cue for direction. The findings of Ronacher and Krahe [25•] face under natural conditions in which sound fields are cor- on grasshoppers, summarized above, suggest that binaural rupted by vegetation [21,32] and extraneous noises may latency difference may be encumbered by its own shortcom- interfere with the signal of interest. This is crucial not only ing, namely high variability. It is thus important to learn to ensure that we are asking sensible questions in our lab- which of the two possible cues, latency difference or oratory experiments, but also to find out whether, and to response-strength difference, dominates in directional deci- what extent, the basic properties of the neurons are affect- sions. Behavioral experiments currently underway in our ed by factors such as background noise (which is known to laboratory are directed at this question. We are exploiting the affect responses in other systems [33]). With rare, but fact that intensity differentially affects the habituation of important, exceptions [11,16], the behavior of auditory response strength and of response latency to uncouple the neurons under field conditions has been essentially binaural differences in these two potential cues; that is, we neglected. Second, our understanding of the neural pro- establish conditions under which the response of one ear is cessing of acoustical signals rests mainly on traditional both stronger and later than that of the other ear. By observ- analytical techniques, such as counting spikes and measuring the direction of behavioral responses under these ing latencies. Evidence in other systems, however, shows conditions, we hope to learn whether crickets attend mainly that information may be encoded more subtly in the to one or the other of these potential cues for direction. detailed timing of action potentials [34]. This conclusion stems from the application of analytical procedures derived Our preliminary results suggest that, despite the con- from systems analysis and information theory. Several lab- founding effects of habituation, sound localization oratories are now beginning to apply these techniques to Recognition and localization of acoustic signals by insects Pollack 767

insect auditory systems, and the next few years will 17. Römer H, Krusch M: A gain-control mechanism for processing of • chorus sounds in the afferent auditory pathway of the bushcricket undoubtedly see new insights into the sorts of information Tettigonia viridissima (Orthoptera; Tettigoniidae). J Comp Physiol that individual neurons, and the animals that they serve, A 2000, 186:181-191. This paper builds on earlier work [14–16] to examine how insect neurons are extract from the sounds that surround them. able to focus on a single signal among many. Importantly, in this paper, the acoustical conditions during the neurophysiological experiments are carefully References and recommended reading matched to those that the animals encounter in the field, so that it is clear that the neurophysiological findings have true impact in the lives of the animals. Papers of particular interest, published within the annual period of review, have been highlighted as: 18. Sobel EC, Tank DW: In vivo Ca2+ dynamics in a cricket auditory neuron: an example of chemical computation. Science 1994, • of special interest 263:823-826. •• of outstanding interest 19. Schul J, Helversen D von, Weber T: Selective phonotaxis in 1. Hoy RR: The evolution of hearing in insects as an adaptation to Tettigonia cantans and T. viridissima in song recognition and predation from bats. The Evolutionary Biology of Hearing In . Edited by discrimination. J Comp Physiol A 1998, 182:687-694. Webster DB, Fay RR, Popper AN. New York: Springer; 1992:115-129. 20. Michelsen A: Biophysics of sound localization in insects. In 2. Pollack GS: Neural processing of acoustic signals. In Comparative Comparative Hearing: Insects. Edited by Hoy RR, Popper AN, Hearing: Insects. Edited by Hoy RR, Popper AN, Fay RR. New York: Fay RR. New York: Springer; 1998:18-62. Springer; 1998:139-196. 21. Stabel J, Wendler G: Cricket phonotaxis: localization depends on 3. Zuk M, Kolluru GR: Exploitation of sexual signals by predators and recognition of the calling song pattern. J Comp Physiol A 1989, parasitoids. Quart Rev Biol 1998, 73:415-438. 165:165-177. 4. Schildberger K, Hörner M: The function of auditory neurons in cricket phonotaxis. I. Influence of hyperpolarization of identified neurons 22. Mörchen A, Rheinlaender J, Schwartzkopff J: Latency shift in insect auditory nerve fibers. 65 on sound localization. J Comp Physiol A 1988, 163:621-631. Naturwissenschaften 1978, :656-657. Acoustic communication and orientation in 5. Nolen TG, Hoy RR: Initiation of behavior by single neurons: the 23. Helversen D von: grasshoppers. role of behavioral context. Science 1984, 226:992-994. In Orientation and Communication in Arthropods. Edited by Lehrer M. Basel: Birkhäuser; 1997:301-341. 6. Gogala M: Songs of four cicada species from Thailand. Bioacoustics 1995, 6:101-116. 24. Ronacher B, Krahe R: Song recognition in the grasshopper Chorthippus biguttulus is not impaired by shortening song 7. Pollack GS, Imaizumi K: Neural analysis of sound frequency in signals: implications for neuronal encoding. J Comp Physiol A insects. Bioessays 1999, 21:295-303. 1998, 183:729-735. 8. Popov AV, Shuvalov VF, Svetogorskaya ID, Markovich AM: Acoustic 25. Ronacher B, Krahe R: Temporal integration vs. parallel processing: behavior and auditory systems in insects. Abh Rheinisch-Westfäl • coping with the variability of neuronal messages in directional Akad Wiss 1974, 53:281-306. hearing of insects. Eur J Neurosci 2000, 12:2147-2156. This paper builds on the findings in [24] to examine the code for sound 9. Schildberger K: Temporal selectivity of identified auditory neurons direction. The authors use data on the reliability of receptor responses in in the cricket brain. J Comp Physiol A 1984, 155:171-185. order to model directional decisions. They show that spike count is less vari- 10. Givois V, Pollack GS: Sensory habituation of auditory receptor able than response latency, suggesting that the former might be a better cue •• neurons: implications for sound localization. J Exp Biol 2000, for sound intensity. The variability of spike count is such that the animal’s 203:2529-2537. behavioral performance can be accounted for only by pooling input from This paper demonstrates that the temporal pattern of an acoustical signal 6–13 receptor neurons. may affect its ease of localization. Signals consisting of long-lasting trains of 26. French AS: Ouabain selectively affects the slow component of rapidly repeating sound pulses cause the responses of receptor neurons to sensory adaptation in an insect mechanoreceptor. Brain Res decline (sensory habituation). The effect is more pronounced at higher inten- 1989, 504:112-114. sities, and so habituation is stronger for receptors serving the ipsilateral ear. As a result, the binaural difference in response strength declines as habitu- 27. French AS: Two components of rapid sensory adaptation in a ation builds up, compromising its utility as a cue for sound direction. The cockroach mechanoreceptor neuron. J Neurophysiol 1989, response latency of receptor neurons for successive sound pulses increases 62:768-777. for stimuli that induce habituation, but this effect is independent of intensity. As a result, binaural latency difference is unaffected, and may be a more reli- 28. Webb B, Scutt T: A simple latency-dependent spiking-neuron able cue for sound direction. model of cricket phonotaxis. Biol Cybern 2000, 82:247-269. 11. Römer H, Bailey WJ: Insect hearing in the field: II. Male spacing 29. Shaw KL, Herlihy DP: Acoustical preference functions and song behavior and correlated acoustic cues in the bushcricket variability in the Hawaiian cricket Laupala cerasina. Proc R Soc Mygalopsis marki. J Comp Physiol A 1986, 159:627-638. Lond B Biol Sci 2000, 267:577-584. 12. Greenfield MD: Synchronous and alternating choruses in insects 30. Hedrick AV: Female preference for male calling bout duration in a and anurans: common mechanisms and diverse functions. Amer field cricket. Behav Ecol Sociobiol 1986, 19:73-77. Zool 1994, 34:605-615. 31. Simmons LW: The calling song of the field cricket, Gryllus 13. Hill PSM: Lekking in Gryllotalpa major, the prairie mole cricket bimaculatus (De Geer): constraints on transmission and its role in (Insecta: Gryllotalpidae). Ethology 1999, 105:531-545. intermale competition and female choice. Anim Behav 1988, 36:380-394. 14. Pollack GS: Discrimination of calling song models by the cricket, Teleogryllus oceanicus: the influence of sound direction on neural 32. Gilbert F, Elsner N: Directional hearing of a grasshopper in the encoding of the stimulus temporal pattern and on phonotactic field. J Exp Biol 2000, 203:983-993. behavior. J Comp Physiol A 1986, 158:549-561. 33. Epping WJM: Influence of adaptation on neural sensitivity to 15. Pollack GS: Selective attention in an insect auditory neuron. temporal characteristics of sound in the dorsal medullary nucleus J Neurosci 1988, 8:2635-2639. and torus semicircularis of the grassfrog. Hearing Res 1990, 45:1-14. 16. Römer H: Environmental and biological constraints for the evolution of long-range signaling and hearing in acoustic insects. 34. Rieke F, Warland D, de Ruytor von Stevenink R, Bialek W: Spikes: Philos Trans R Soc Lond B Biol Sci 1993, 340:179-185. Exploring the Neural Code. Cambridge, MA: MIT press; 1997:395.