473

Sensory hyperacuity in the jamming avoidance response of weakly electric fish Masashi Kawasaki

Sensory systems often show remarkable sensitivities to The jamming avoidance response small stimulus parameters. Weakly electric; fish are able to The South American weakly electric fish Eigenmannia resolve intensity differences of the order of 0.1% and timing and the African weakly electric fish Cymnanhus perform differences of the order of nanoseconds during an electrical electrolocation by generating constant wave-type electric behavior, the jamming avoidance response. The neuronal organ discharges (EODs) at individually fixed frequencies origin of this extraordinary sensitivity is being studied within (250-600Hz) using the electric organ in their tails. Each the exceptionally well understood central mechanisms of this cycle of an EOD is triggered by coherent action potentials behavior. originating from a coupled oscillator, the pacemaker nucleus in the medulla. An alternating current (AC) electric field is thus established around the body, and its Addresses distortion by objects is detected by electroreceptors on the Department of Biology, University of Virginia, Gilmer Hall, Charlottesville, Virginia 22903, USA; e-mail: [email protected] body surface [5,6] (Figure 1).

Current Opinion in Neurobiology 1997, 7:473-479 When two fish with similar EOD frequencies meet, http://biomednet.com/elelecref~0959438800700473 their electrolocation systems jam each other, impairing

0 Current Biology Ltd ISSN 0959-4388 their ability to electrolocate. To avoid this jamming, they shift their EOD frequencies away from each other in Abbreviations a jamming avoidance response in order to increase the ELL electrosensory lateral line lobe difference in their frequencies [7-91. During the jamming EOD electric organ discharge avoidance response, a fish determines, without trial and error, whether it should increase or decrease its own EOD frequency relative to that of its neighbor by computing the Introduction sign of the difference between its own and its neighbor’s Human psychophysics and animal behavioral studies often EOD frequency: Af=f2-fl,where Af is the frequency reveal the astonishingly high sensitivity of sensory organs difference, and fl and f2 are the fish’s own and its to various stimulus parameters [l]. Examples of this neighbor’s EOD frequency, respectively. include vernier acuity in human vision (5s of arc) [2], interaural time disparity in human audition (6~s) [3], Behavioral experiments have demonstrated that a fish and thermal sensitivity in snakes (O.OOl”C) [4]. These determines the sign of 4 solely from the mixture of behavioral sensitivities, or hyperacuities, often exceed sensory feedback from its own electric organ and its the resolution of individual sensory receptor neurons by neighbor’s EODs, without referring to the pacemaker orders of magnitude and, thus, must result from central nucleus for information about fl [lO-131. In these studies, processing. As hyperacuity often results from many steps the fish’s EODs were silenced by blocking cholinergic of central processing, elucidation of its central mechanisms synapses at the electric organ and were replaced with has been hampered by the absence of ‘transparent’ artificially generated EODs at arbitrary frequencies. As systems, in which the flow of pertinent information all the electroreceptors on the body surface are exposed processing can be tracked within the CNS, from sensory to a mixture of the fish’s own and its neighbor’s EODs, receptors to behavioral output. and no receptor is uniquely stimulated by one of them, information about the sign of 4must be computed from In weakly electric fish, information processing within the a complex mixture of the two stimuli. central electrosensory and electromotor mechanisms for an electric behavior, the jamming avoidance response, Two sensory cues that are embedded in this complex is well understood. Furthermore, this response also mixture have been identified as essential for the cal- demonstrates sensitivities to extremely small stimulus culation of Af-specifically, amplitude modulation and parameters (amplitude fluctuations of less than 0.1% and differential-phase modulation, both of which are necessary time disparities in the range of nanoseconds). Thus, for a fish to perform a jamming avoidance response. As weakly electric fish provide a transparent system for shown in Figure 2, amplitude modulation is a periodic examining high sensitivities expressed at the behavioral change of stimulus intensity that results from the beating level. This review focuses on research that examines the of two signals. Differential-phase modulation represents central mechanisms of the jamming avoidance response in small phase differences at different areas of the body light of hyperacuity. that are created by the different spatial geometry of 474 Sensory systems

(a)

Electric organ Gymnarchus /

(cl Jamming avoidance response

EOD frequency 2Hz

f2 : neighbor’s EOD frequency +2 Hz Af -2 Hz

1 min 0 1997 Current Opinion m Neurobiology

Electrolocation and jamming avoidance response. (a) figenmannia and Gymnarchus emit EODs from the electric organ in their tail. (b) Distortion of the fish’s electric field in response to an object (dark gray circle) is detected by electroreceptors on the body surface. (c) The top trace depicts a fish’s EOD frequency, displaying a jamming avoidance response, in response to Af (bottom trace), which represents the difference between the fish’s own (ft) and its neighbor’s (f,) EOD frequency.

a fish’s own and its neighbor’s EOD field. These two avoidance responses when the amplitude modulation was parameters vary over time at the same frequency, ]Afi, 0.2% and differential-phase modulation was 1 ps. The with different temporal sequences for Af 0 (see most sensitive fish in the study showed weak but accurate Figure Zc,d,e,f). Various behavioral experiments [lO-131 jamming avoidance responses at an amplitude modulation have predicted that the fish’s CNS must be able to detect of 0.02% and a differential-phase modulation of 90ns the modulation time courses of these two parameters (Figure 3). in order to perform a jamming avoidance response. Despite their independent evolution, both Eigenmannia Because both amplitude and differential-phase modulation and Gymnadus have evolved the same computational are necessary for accurate jamming avoidance responses algorithm for the jamming avoidance response [13]. and because the modulation depths for these parameters co-varied in the above experiments, the detection thresh- old for one of these parameters may even be lower than Hyperacuity in the jamming avoidance estimated. The threshold for one of the parameters while response keeping the other suprathreshold has not been tested. The temporal relation between amplitude and differential- phase modulations depicted in Figure Ze,f determines In the behavioral experiments described above [14,15,16”], whether a fish raises or lowers its EOD frequency during a a fish’s response was observed in -30s periods, during jamming avoidance response. Rose and Heiligenberg [ 141 which time each of the amplitude and differential- and Carr et a/. [15] measured the threshold modulation phase sensitive systems may perform temporal averaging. depths of amplitude and differential phase in Eigenmannia. The jamming avoidance response, however, requires the They reduced the diameter of the circular graphs in temporal structure of amplitude and differential-phase Figure Ze,f until the fish failed to respond to a change modulation, which occurs at a rapid rate (-4 Hz); therefore, in the sense of rotation (i.e. the sign of An by shifting any type of temporal averaging over many seconds, their EOD frequencies in the opposite direction. Even which smears the temporal structure of the signal and when the amplitude modulation was 0.1% and the merely detects the presence of amplitude modulation and differential-phase disparity was 400 ns, the fish still shifted differential-phase modulation, could not be employed. their EOD frequencies in the correct direction. Thus, these behavioral experiments suggest that the fish have an internal representation of the temporal pattern Guo and Kawasaki [16**] have recently shown that of extremely small amplitudes and differential-phase Gymnadus exhibits comparable sensitivities. In their modulations. Spatial averaging, however, appears to play experiments, all the fish performed accurate jamming an important role [14]. Sensory hyperacuity Kawasaki 475

Figure 2

(a) (b) Amplitude modulation __---__

Phase modulation

(c) (l.0

AfO

Signal at A

Signal at B

0 0 LeadL 0 Lag LeadL 0 Lag

Lead Lead Differential phase

0 1997 thnent Opinion in Neurobiology

Amplitude modulation and differential-phase modulation-the essential cues of the jamming avoidance response. (a) A mixture of a fish’s own and a neighbor’s EODs creates a beating signal. Amplitude modulation is the periodic change of the signal envelope. Phase modulation is the periodic change of zero-crossing times. Small vertical ticks mark would-be zero-crossing times of the fish’s own signal alone. (b) While the frequency of both of these modulations equals the absolute frequency difference Id, the magnitudes, or depths, of the modulations are a function of intensity ratios between the two signals. For example, the intensity ratios at body areas A and B are different because the fish’s own EOD establishes a radial electric field (long arrows) because of the internal location of the fish’s own electric organ, whereas the EOD of the fish’s neighbor creates a more-or-less parallel field (short arrows) because of the external location of the neighbor’s electric organ. Thus, the depths of the modulations are different at electroreceptors A and 8. The signal mixture at body areas A and B are plotted for (c) tic 0 and (d) f>O. While traces for A and B both show phase modulation, their depths are different, thus creating differential-phase modulation. (e,f) The temporal patterns of these modulations are different for different signs of Af. The amplitude envelope (top traces) and differential phase (bottom traces) both modulate at the same frequency 14, but their temporal relations are different, as revealed by the opposite senses of rotation in the Lissajous graphs depicted on the top right.

Emergence of behavioral accuracies in brain jamming avoidance response. While behavioral sensitivity nuclei to amplitude modulation is largely accomplished at the In Eigenmannia, the chain of neuronal structures in- first brain station, acuity to phase information improves volved in the jamming avoidance response has been progressively along the neuronal chain. well characterized physiologically and anatomically [17,18] (Figure 4). Neuronal acuity to amplitude and phase have Amplitude information is sampled by P-type electrorecep- been measured in the different nuclei involved in the tors, which encode stimulus amplitude by the probability 476 Sensory systems

Fiaure 3

(a) W Differential-phase modulation: 2.8ps Amplitude modulation: 0.58% _A0 ._._Q -Da

Differential-phase modulation: 92.6 ns I m ’ ’ ’ “,‘I m - ’ ’ “,‘I ’ v 0.1 1 .o 10 Mean differential-phase modulation @s.)

0.01 0.1 10 1 min Mean amplitude modulation (%) , 0 1887 Currenl Opinion in Nsurcbiilogy

Jamming avoidance response of Gymnarchus in hyperacuity conditions. (a) The sense of rotation (Af0) in the circular Lissajous graphs in Figure 2 was switched every 30s (dotted vertical lines). The diameter of the circular graph (i.e. depth of modulation) was made progressively smaller to determine the behavioral thresholds for the modulations. The frequency traces are not averaged; depicted here are the results of a single trial consisting of eight 30 s segments. This particular individual showed weak but correct jamming avoidance responses (JARS) at amplitude modulation of 0.0192% and differential-phase modulation of 92.6 ns. (b) Collective data from eight individuals. Different symbols represent different individuals. Unconnected symbols represent the percentage of correct responses in the eight sessions; connected symbols represent the magnitude of the JAR. Adapted from [16*‘1.

of firing of action potentials. Although the responses of anatomical convergence of afferent fibers to ELL neurons P-type electroreceptor afferents to a very small amplitude [ZZ]. Descending pathways from the midbrain nucleus modulation have not been recorded, Scheich et a/. [19] back to the ELL appear to have no effect on sensitivity have reported that a step increase of stimulus amplitude by enhancement in ELL neurons [23,24]. a factor of 4% induced an additional single action potential in a primary afferent fiber that was firing at -SOspikes/s. These ELL neurons project to the torus semicircularis Taking noise and fluctuations in firing probability into (torus hereafter) in the midbrain [25,26]. When Rose and consideration, representation of amplitude modulation at a Heiligenberg [27] examined the threshold for amplitude level of 0.1% in an individual P-type afferent fiber is weak. modulation in amplitude-sensitive neurons in the torus, they found that the median was 2%, and the most sensitive The primary afferent fibers from P-type electroreceptors of 28 neurons showed a threshold response to 0.1% of project to the basilar pyramidal cells via direct excitatory amplitude modulation. Thus, there appear to be small connections and to nonbasilar pyramidal cells via indirect improvements of amplitude sensitivity in this nucleus. inhibitory connections in the electrosensory lateral line lobe (ELL) in the medulla [ZO]. Shumway [Zl] measured Whereas amplitude sensitivity (as demonstrated at the the threshold of both types of pyramidal cells in the behavioral level) is achieved largely at earlier stages ELL. The mean threshold of neurons in the most in the neuronal chain, sensitivity to phase is improved sensitive subdivision of the ELL (lateral map) was -3%, progressively, involving many nuclei (as described below). and the most sensitive neuron’s threshold was 0.3%. In a closely related gymnotiform fish, Apteronotus, which Phase, or zero-crossing time of sensory signals, is sampled also perform jamming avoidance responses, both basilar by T-type electroreceptors, which fire one action potential and nonbasilar pyramidal cells gave good responses to at each zero-crossing of the signal [19]. Afferents from amplitude modulation of 0.3%, which was the smallest T-type electroreceptors terminate on the spherical cells amplitude modulation tested [Z?]. Sensitivity to amplitude in the medulla, which in turn project to giant cells in modulation at this first sensory processing stage seems to the torus. As these cells are phase-locked- that is, they be remarkably improved over that of primary afferents. respond with one action potential to a particular phase of This large increase of sensitivity requires (statistically) a one cycle of sensory stimuli- the phase of the sensory lbfold larger sample size. This is probably achieved by signal is represented by action potential times of giant Sensory hyperacuity Kawasaki 477

/ Pacemaker nucleus Nucleus electrosensorius Prepacemaker Sublemniscal \ nucleus prepacemaker Primary afferents from nucleus electroreceptors

0 1997 Current Opinm in Neurobiology

Brain nuclei involved in the jamming avoidance response of Eigenmannia. The physiological properties of some of these neurons have been reported: electric organ/behavior [14,15,16’*1; primary afferents from electroreceptors 1191; sublemniscal prepacemaker nucleus 1321; prepacemaker nucleus [36]; nucleus electrosensorius [31]; torus semicircularis 127-301; electrosensory lateral line lobe (ELL) [21,22,27-301.

neurons in the torus. Carr eta/. [ 151 measured the accuracy -0.2% and differential-phase modulation is -1 ps [31]. of action potentials in these phase-locked neurons and In the one of two final nuclei for jamming avoidance showed that it progressively improved due to anatomical response processing [32], the prepacemaker nucleus convergence. While the T-type afferents show average jit- [33-3.51, individual neurons are almost as accurate as the ter of -30 ps (range of 10-100 s), jitter of midbrain neurons behavioral output [36], culminating in the improvement was -11 ps, on average (range of 4-32 s). Phase differences of the accuracy for small modulations along the neuronal between giant cells and inputs from spherical cells are chain. detected by small cells in the torus, which project to other differential-phase-sensitive neurons [Z&29]. When Mechanisms of differential-phase comparison Rose and Heiligenberg [30] examined the threshold for Although we know that small cells in the torus extract differential-phase modulation in differential-phase-sensi- differential-phase information in Eigenmannia [29,37], little tive neurons in the torus, they found that the threshold is known about the cellular mechanisms of differential- median was lops, and the most sensitive of 37 neurons phase detection. How do small cells detect time disparities showed a threshold response to 6ps of differential-phase of microseconds from phase-locked action potentials of modulation. These differential-phase-sensitive neurons longer duration! The small size of the neurons in encode the time course of differential-phase modulation the torus has precluded routine intracellular recordings by their timing of action potential bursts. Thus, temporal so that postsynaptic potential could be studied. In a encoding is not constrained by the timing accuracy of context of comparative studies [38*,39], Kawasaki and individual action potentials, allowing higher-order neurons Guo [40**] discovered a phase comparison circuitry in the to further improve accuracy. medullary structure of Gymnanh, in which neurons are also sensitive to differential-phase modulation in the range Neurons in the torus that are sensitive to the sign of of microseconds and are relatively large (12-15 pm). 4 have convergent projections from amplitude-sensitive neurons and differential-phase-sensitive neurons [27], but In my laboratory, we are using in oivo whole-cell recordings the scarcity of such neurons and the difficulty of recording to measure postsynaptic potentials in neurons that receive from them prevent examination of their sensory threshold phase-locked inputs and are sensitive to differential phase. [31]. Neurons in the subsequent processing area, the We have also described the complex response dynamics nucleus electrosensorius, have been recorded under hyper- of these neurons [40**], suggesting that differential-phase acuity conditions. Neurons in this nucleus discriminate comparison is performed not only by simple coincidence the sign of Af, even when amplitude modulation is detection but also by a complex adaptation mechanism. In 478 Sensory systems

a recent paper [16”], we showed that phase accuracy of These two limitations are attributable to difficulties primary afferent fibers is somewhat better in Gymnarzhs of standard intracellular recording and labeling in the (-6~s) than in Eigenmannia (-3Ops), reflecting the fact small neurons of the fish brain. Recently, however, in that differential-phase computation is performed one step &JO whole-cell recording techniques [Sl’], which allow earlier in Gymnadus. long-term intracellular recording of postsynaptic potentials and reliable tract tracing, have been applied successfully in Temporal hyperacuity in other systems electrosensory systems and hold much promise for further Mormyrid electric fish use temporal information in the study. Future studies of the detailed mechanisms of the range of microseconds for species and individual recog- jamming avoidance response should continue to provide nition [41]. The neuronal mechanisms for temporal fruitful ground for understanding not only hyperacuity but analysis in these species show interesting parallels with also the neural substrate of behavior in general. Eigenmannia and Gymnarchs [39,42-114]. Remarkable sen- sitivity (-5 nV/cm) to low-frequency electrosensory signals Acknowledgements has also been reported in elasmobranch fish [45], but Work reported from the author’s laboratory was supported by National nothing is known about the central mechanisms for this Insritutes of Health grants RZ9MH48115-05 and K02MH01256-01. I thank Cameron McLaughlin for editing my English and Yasuko Kawasaki for hyperacuity. preparation of the figures.

Two specialists in audition, echolocating bats and the barn References and recommended reading owl, are of particular interest in examining hyperacuity. Papers of particular interest, published within the annual period of review, Simmons et a/. [46] reported that the threshold for echo have been highlighted as: delay detection is as small as 12 ns in an echolocating bat. . of special interest They propose that neurons in the inferior colliculus deal l 0 of outstanding interest with the very small temporal signals by time expansion in which nanosecond temporal codes are expressed by 1. Autrum H: Performance limits of sensory organs. lnterdisciplin multiple unit potentials of millisecond range [47,48’]. Sci Rev 1966, 13:27-39. Barn owls can localize a sound source in elevation using 2. Westheimer G: Visual hyperacuity. Prog Sensory Physioll961, 1 :l -30. interaural intensity differences and in azimuth using 3. Tobias JV: Curious binaural phenomena In Foundations of interaural time differences; in behavioral experiments, Modern Auditory Theory, vol 2. Edited by Tobias JV. New York: Knudsen, Blasdel and Konishi (491 determined that the Academic Press; 1972:465-466. accuracy of localizing sound is approximately 2” both in 4. De Cock Buning T: Threshold of infrared sensitive tectal neurons in Python retiw/atus, Boa wnstrictor and Agkistrvdon elevation and azimuth. These values correspond roughly rhodostcme. J Comp Physioll983, 151:461-467. to interaural intensity difference of 5% and interaural time 5. Bastian J: EIectrolocatIon. Behavior, anatomy, and physlology. differences of -5 ps [SO]. In Electroreception. Edited by Bullock TH, Heiligenberg W. New York: Wiley & Sons; 1966:577-612. Condusions 6. Lissmann HW, Machin KE: The mechanism of object location in Gymnarchus nilotiws and similar fish. J Exp Biol 1956, 35:451- The study of the electrosensory system of weakly electric 466. fish shows how sensory acuity expressed in behavior 7. Watanabe A, Takeda K: The change of discharge frequency by is shaped along a neuronal chain of processing areas. AC. stimulus in a weakly electric fish. J Exp Biol 1963, 40:57- 66. While amplitude sensitivity is largely accomplished by 6. Heiiigenberg W: Principles of Electrolocation and Jamming one processing area (the ELL), sensitivity to small phase Avoidance in E/e&c Fish -A Neuroethological Approach, vol I. differences emerges from successive processing by many New York: Springer-Vet-lag; 1977. brain areas. This difference reflects perhaps a difference 9. Bullock TH, Hamstra RH, Scheich H: The jamming avoidance in the complexity of neuronal processing for these two response of high frequency electric fish. I. General features. J Comp Physioll972, 77~1-22. parameters. Persistence of the electrical behavior and 10. Heiiigenberg W, Baker C, Matsubara J: The jamming avoidance accessibility of all involved brain nuclei in neurophysio- response in Eigenmennia revisited: the structure of a neuronal logical preparations of electric fish have greatly facilitated democracy. J Comp Physioll976, 127:267-266. a systems-level analysis of behavioral hyperacuity, which 11. Heiligenberg W, Bastian J: The control of Eigenmennids pacemaker by distributed evaluation of electroreceptive could not be accomplished by studying individual steps of afferences. J Comp Physioll960, 136:113-l 33. neuronal processing in reduced preparations. 12. Bastian J, Heiligenberg W: Neural correlates of the jamming avoidance response in Eigenmannia. J Comp Physioll960, Our understanding of the neuronal mechanisms for 136:135-l 52. hyperacuity is limited, however, in two ways. First, we 13. Kawasaki M: Independently evolved jamming avoidance responses employ identical computational algorithms: a have not expIained how signals of small magnitude behavioral study of the African electric fish, Gymnanhus are processed within each component neuron, because nilotiws. J Comp Physiol 1993, 173:9-22. postsynaptic potentials for subthreshold signals have 14. Rose GJ, Heiiigenberg W: Temporal hyperacuity in the electric sense of fish. Nature 1965, 316:176-l 60. been difficult to record. Second, more detailed studies 15. Carr CE. Heiligenberg W, Rose GJ: A time-comparison circuit are needed to determine the anatomy of neuronal in the electric fish midbrain. I. Behavior and physiology. connections, particularly those beyond the midbrain. J Neurosci 1966, 6:107-l 19. Sensory hyperacuity Kawasaki 479

16. Guo Y. Kawasaki M: Repreoentetlon of accurate temooral 35. Kawasaki M: Comperetlve studies on the motor control . . information in the electrosensory system of the Af&an electric mechanisms for electro-communicetfon in gymnotiform fishes, fish, Gymnarchus niloticus. J Neurosci 1997, 17:1761-l 768. J Comp Physioll993, 173:726-728. The acuity of the phase processing system in Gymnarc!tus was measured 36. Kawasaki M, Rose GJ, Heiligenberg W: Temporel hypemcuity in usina hiahlv orecise stimulus and recordinn methods. The iitter of ohase- single neurons of electric fish. Nature 1988, 336:173-l 76. looked ~ou&s was -6~s. The stability 07 the electric organ discharges was also measured and their jitter was less than 1 ps. This stability of electric 37. Heiligenberg W, Rose G: phase end amplitude computetfons organ discharge, however, is not required for hyperacuity. in the midbrein of an electric fish: intracellular studies of neurons partlclpetlng In the jamming evoldance response of 1 7. Heiligenberg W: Neural Ners in Electric Fish. Cambridge, Eigenmannia. J Neurosci 1965,5:515-531. Massachusetts: MIT Press; 1991. 38. Kawasaki M: Comparative analysis of the jamming avoidance 18. Heiligenberg W: Jamming avoidance responses: model . response in Africen and South Amerlcen wave-type flshen Biol systems for neuroethology. In Elecfroreception. Edited by Bull 1996, 191 :103-108. Bullock TH, Heiligenberg W. New York: Wiley & Sons; 1986:613- Two phylogenetically unrelated groups of fishes perform electrolocation. Not 649. only do they share this basic function, but both fishes perform jamming 19. Scheich H, Bullock TH, Hamstra RH: Coding properties avoidance responses of identical form. Moreover, they share rather complex of two classes of afferent nerve fibers: high-frequency computational algorithms. Different brain areas, however, are used to per- electroreceptors in the electric fish, Eigenmacnia. form a certain step of neuronal computation. Comparative implications are J Neurophysiol 1973, 36:39-60. discussed. 39. Hopkins C: Convergent designs for electrogenesls end 20. Maler L, Sas E, Rogers J: The cytology of the posterior lateral electroreception. Curr Opin Neurobiol 1995, 5:769-777. line lobe of high frequency weekly electric fish (Gymnotidee). Dendritic differentiation and synaptic specificity in a simple 40. Kawasaki M, Guo Y-X: Neuronal circuitry for comperison of cortex. J Comp Neural 1981, 195:67-139. . . timing in the electrosensory lateral line lobe of an Afrlcen wave-type electric fish, Gymnanzhus niloticus. J Neurosci 1996, 21. Shumway CA: Multiple electrosensory maps in the medulla of 16:300-391. weakly electric gymnotlform fish. I. Physiological differences. Intracellular recording and labeling were performed in a neural circuit that J Neurosci 1989, g:4368-4399. detects ohase differences in the microsecond raoae. Phase-locked neurons 22. Bastion J: Electrolocation. II. The effects of moving objects spread &on terminals in both halves of a medulltistructure, the electrosen- and other electrical stimuli on the activities of two categories sory lateral line lobe. Neurons sensitive to phase differences are found in the of posterior lateral line lobe cells in Apteronotus a/bifrons. inner cellular layer of this structure. These neurons respond to changes in J Comp Physiol 1961, 144:481-494. smell phase differences, but ignore steady-state differences in phase.

23. Bastian J, Bratton B: Descending control of electroreception. 41. Hopkins CD, Bass AH: Temporal coding of species recognition I. Properties of nucleus praeeminentialis neurons projecting signals in an electric fish. Science 19131, 21265-87. indirectly to the electrosensory lateral line lobe. J Neurosci 42. Friedman MA, Hopkins CS: Neural substrate for species 1990, IO:1 226-l 240. recognition in the time-coding elecbosensory pathway of mormyrid electric fish J Neurosci 1997. in press. 24. Bastian J: Gain control in the electrosensory system mediated by descending inputs to the electrosenoory lateral line lobe. 43. Amagai S, Friedman MA, Hopkins CD: Time-coding in the J Neurosci 1906, 6:553-562. midbrain of mormyrid electric fish. I. physiology end anatomy of cells in the nucleus exterolateralis pars anterior (ELa). 25. Carr CE, Maler L, Heiligenberg W, Sas E: Laminar organization J Comp Physiol IA] 1997, in press. of the afferent and efferent systems of the torus semicircularis of Gymnotlform fish: morphological substrates for parallel 44. Amagai S: Time-coding in the midbrain of mormyrid electric processing in the electrosensory system. J Comp Neural 1961, fish. II. Stimulus selectivity in the nucleus exterolaterelis pars 203:649-670. posterior (ELp). J Comp Physiol fAj 1997, in press.

26. Shumway CA: Multiple electrosensory maps in the medulla of 45. Kalmijn A: Electric and magnetic field detection in weakly electric gymnotiform fish. II. Anatomical differences. elasmobranch fishes. Science 1982, 218:916-918. J Neurosci 1989, 9:4400-4415. 46. Simmons J, Ferragamo M, Moss C, Stevenson S, Altes R: 27. Rose GJ, Heiligenberg W: Neural coding of difference Discrimination of jittered sonar echoes by the echolocating frequencies in the midbrain of the electric fish Eigenmannia: bat Eptesicus fuscus: the shape of terget images in reading the sense of rotation in en amplitude-phese plane. echolocation. J Comp Physiol IA1 1990, 167:569-616. J Comp Physiol 1966, 159:613-624. 47. Simmons J, Saillant P, Ferragamo M, Haresign T, Dear S, Fritz J, 28. Bastian J, Heiligenberg W: Phase-sensitive midbrain neurons McMullen T: Auditory computations for biosonar target imaging In Eigenmannia: neural correlates of the jamming avoidance in bats. In Auditory Computation. Springer Handbook of Auditory response. Science 1980, 209:826-831. Research Edited by Hawkins H, McMullen T, Popper A, Fay R. New York: Springer-Verlag; 1995:401-468. 29. Carr CE. Maler L. Tavlor B: A time-comoarison circuit in the , 48. Simmons J, Dear S, Ferragamo M, Haresign T, Fritz J: electric fish midbrain. II. Functional morphology. J Neurosci . Representation of perceptual dimensions of insect prey during 1986, 6:1372-l 363. terminal pursuit by echolocating bats. Biol Bull 1996, 191 :I 09- 30. Rose GJ, Heiligenberg W: Limits of phase and amplitude 121. sensitivity in the torus semlcirculsris of Eigenmannia. J Comp On the basis of multiunit recording data in the inferior colliculus of a bat, Physiol IA1 1986, 15g:al3-822. Eptesicus, the authors have developed a new theory that bats’ hyperacuity is endowed by time expansion. Latency of multiple unit response peaks to 31. Keller CH: Stimulus discrimination In the diencephalon of pulse-echo pairs changes by l OO-200~s as echo delay changes by about Eigenmannia: the emergence and sharpening of a sensory 5 us. filter. J Comp Physioll996, t62:747-757. 49. Knudsen El, Blasdel GG, Konishi M: Sound localization by the 32. Metzner W: The jamming avoidance response in Eigenmannia barn owl (Tyto a/be). J Comp Physiol 1979, 133:13-21. is controlled by two separate motor pathways. J Neurosci i 993, 13:i 862-i 878. 50. Brainard MS, Knudsen El, Esterty SD: Neural derivation of sound source location: resolution of spatial ambiguities in binaural 33. Kawasaki M, Maler L, Rose GJ, Heiligenberg W: Anatomical cues. J Acoust Sot Am 1992, 91 :1015-l 027 and functlonel organization of the prepacemaker nucleus in gymnotiform electric fish: the accommodation of two 51. Rose G, Fortune E: New techniques for making whole-cell . behaviors in one nucleus. J Comp Neural 1988, 276:113-l 31. recordings from CNS neurons in ho. Neurosci Res 1996, 26:89-94. 34. Kawasaki M, Heiligenberg W: Different classes of glutamate A whole-cell recording technique using patch-clamp pipettes is refined for receptors and GABA medlate distinct modulations In the firing in viva preparation of electric fish. This method permits long-term recording pattern of a neuronal oscllletor, the medullary pacemaker of of synaptic potentials as well as action potentials in preparations showing gymnotlform electric flsh. J Neurosci 1990, 10:3896-3904. electrical behavior.