Journal of Comparative Physiology A

, Volume 192, Issue 6, pp 561–572

Walter Heiligenberg: the jamming avoidance response and beyond


    • School of Engineering and ScienceInternational University Bremen
    • Department of NeurosciencesUniversity of California at San Diego
  • T. H. Bullock
    • Department of NeurosciencesUniversity of California at San Diego

DOI: 10.1007/s00359-006-0098-5

Cite this article as:
Zupanc, G.K.H. & Bullock, T.H. J Comp Physiol A (2006) 192: 561. doi:10.1007/s00359-006-0098-5


Walter Heiligenberg (1938–1994) was an exceptionally gifted behavioral physiologist who made enormous contributions to the analysis of behavior and to our understanding of how the brain initiates and controls species-typical behavioral patterns. He was distinguished by his rigorous analytical approach used in both behavioral studies and neuroethological investigations. Among his most significant contributions to neuroethology are a detailed analysis of the computational rules governing the jamming avoidance response in weakly electric fish and the elucidation of the principal neural pathway involved in neural control of this behavior. Based on his work, the jamming avoidance response is perhaps the best-understood vertebrate behavior pattern in terms of the underlying neural substrate. In addition to this pioneering work, Heiligenberg stimulated research in a significant number of other areas of ethology and neuroethology, including: the quantitative assessment of aggressivity in cichlid fish; the ethological analysis of the stimulus–response relationship in the chirping behavior of crickets; the exploration of the neural and endocrine basis of communicatory behavior in weakly electric fish; the study of cellular mechanisms of neuronal plasticity in the adult fish brain; and the phylogenetic analysis of electric fishes using a combination of morphology, electrophysiology, and mitochondrial sequence data.


Cichlid fishAstatotilapia (Haplochromis) burtoniLaw of heterogenous summationElectric fishEigenmannia sp.ElectrolocationParallel processing



Central posterior/prepacemaker nucleus


Frequency of a neighboring fish’s electric organ discharge minus frequency of the fish’s own discharge


Electrosensory lateral line lobe


Electric organ discharge


Nucleus electrosensorius


P-type electroreceptor


Pacemaker nucleus


Sublemniscal prepacemaker nucleus


T-type electroreceptor


Torus semicircularis pars dorsalis

Childhood and first scientific encounters

Walter Heiligenberg (Fig. 1) was born on January 31, 1938 in Berlin, Germany. Forced by the war, his family moved several times and finally settled in Münster (Westphalia) in 1951.
Fig. 1

Walter Heiligenberg in his laboratory at the Scripps Institution of Oceanography. (Photograph by G. K. H. Zupanc)

In 1953, at the age of 15, Walter met Konrad Lorenz, one of the founders of modern ethology and at that time head of a Max Planck research group in Buldern near Münster. Inspired by Lorenz, Walter became interested in zoology, particularly the study of animal behavior. While still at school, he frequently visited Lorenz and conducted his first behavioral observations and experiments on fish and birds kept in his parents’ house.

After graduating from a humanistic gymnasium in 1958, he enrolled at the University of Münster, but soon thereafter transferred to the University of Munich, when Lorenz, jointly with the behavioral physiologist Erich von Holst, established a new research institute solely devoted to the analysis of behavior—the Max Planck Institute for Behavioral Physiology in Seewiesen, 20 km southwest of Munich. In the first few semesters Walter studied primarily zoology and botany, but later focused more and more on mathematics and physics. In the latter subject, he passed his prediploma exam (equivalent to obtaining a B.Sc. degree) in 1962. This keen interest in a stringent analytical approach, combined with the attempt to describe scientific phenomena in mathematical terms, promised to become a characteristic feature of Walter’s scientific approach.

The ethological phase

His Ph.D. thesis, carried out under the guidance of Konrad Lorenz and the sensory physiologist Hansjochem Autrum, was entitled “Ursachen für das Auftreten von Instinktbewegungen bei einem Fische (Pelmatochromis subocellatus kribensis BOUL., Cichlidae)” [“Proximate causes of instinctive behavior in a fish (Pelmatochromis subocellatus kribensis BOUL., Cichlidae)”]. In this investigation, Walter performed a quantitative analysis of the effect of motivational factors on the occurrence of various social behavioral patterns.

After obtaining his Ph.D. in 1963, Walter remained in Seewiesen, but left the group of Lorenz to join the more analytically oriented department of Horst Mittelstaedt. During the following years, he continued to analyze motivational processes in animals, particularly in cichlids and crickets. He developed techniques to register behavior patterns under controlled experimental conditions and to describe the temporal structure of these behaviors by mathematical means. Contrary to the then dominant view, he showed that the presentation of a behavioral stimulus may not only have an immediate releasing (“phasic”) effect but also a long-lasting motivational (“tonic”) effect, and that these effects are just the two end-points of a spectrum of possibilities that might occur.

Walter also obtained clear experimental evidence that different features of a stimulation regime lead to an independent behavioral stimulation in the receiver. This phenomenon, known as the “law of heterogeneous summation” (Seitz 1940), had been described by other investigators before, but Walter, together with his student Daisy Leong, was one of the first to succeed in its quantitative demonstration (Leong 1969) (Fig. 2).
Fig. 2

Quantitative demonstration of the law of heterogenous demonstration using Burton’s mouthbrooder, Astatotilapia (Haplochromis) burtoni. In males of this East African cichlid fish, two features of their coloration are crucial in eliciting aggressive responses from other males: a black eye bar that runs from the posterior end of the mouth to the eye; and bright orange spots in the pectoral region as well as on the dorsal, caudal, and anal fins. To test the effect of these two stimulus features quantitatively, an adult male was placed with a group of young fish in an aquarium and stimulated with one of three dummies presented behind a glass partition. Dummy (1) had a black eye-bar, but no orange pectoral spots. Presentation of this dummy increased the average bite rate toward the young fish by 2.81 bites/min. Dummy (2) had orange spots in the pectoral region as well as on the dorsal, caudal, and anal fins, but lacked the black eye-bar. Its presentation lowered the number of attacks by 1.77 bites/min. Dummy (3) combined both features by having the black eye-bar and the orange spots. This dummy elicited an average increase in the attack rate by 1.12 bites/min. The law of heterogenous summation predicts that when Dummy (1) and Dummy (2) are combined, an increase in attack rate by 1.04 bites/min [= 2.81 bites/min + (−1.77 bites/min)] should result. Thus, the predicted value (1.04 bites/min) and the observed value (1.12 bites/min) are in remarkable agreement. As this study also demonstrates, the approach used by Walter Heiligenberg has the remarkable feature that it does not result in a “use up” or exhaustion of the aggressive drive, so that many tests can be made with little variation. (After Leong 1969)

In the course of his studies, Walter accumulated evidence incompatible with the then popular “psychohydraulic model of motivation” of Konrad Lorenz (Lorenz 1963). This model proposed that animals and humans possess an aggressive drive which builds up spontaneously and is discharged by an aggressive action. However, by studying males of a territorial cichlid fish, Walter showed that no spontaneous build-up of aggressive motivation occurs during isolation, as predicted by the model. Instead, the motivation to attack adult conspecifics needs to be continually replenished by specific external stimuli. Repeated presentation of such stimuli has even a long-lasting incremental effect on the level of aggressive motivation (Fig. 3). On the other hand, when the fish display aggressive behavior, no cathartic effect is observed, as predicted by Lorenz’s popular model.
Fig. 3

Motivational effect of social stimuli. A male of Burton’s mouthbrooder, Astatotilapia (Haplochromis) burtoni, was kept for ten consecutive days with young fish in an aquarium without a dummy presented, and its baseline level of aggression, characterized by the bite rate exhibited toward the young fish, was measured. Then, starting with day 11, a male fish dummy (inset) was presented every 15 min for 30 s over 8 h every day for 10 days (indicated by the bar underlying the abscissa). Upon completion of the stimulation, the attacks exhibited toward the young fish continued to be measured for another 20 days. The data points are the mean of six experiments. The line indicates the theoretical build-up and decay of the attack rate. The results of this experiment show that the presentation of the male dummy not only has an immediate releasing effect upon the male’s attack rate, but it also leads to long-term changes in the motivation of the male to attack the young fish. (After Zupanc 2004, based on data by Heiligenberg and Kramer 1972)

Based on the results of his ethological studies, Walter developed a general stochastic model of motivation. According to this concept, motivational actions are governed by excitable random processes in the central nervous system. Walter characterized such processes in terms of the stimuli they respond to, and their decay rates after stimulation. Today, the results of his behavioral work can be found in many textbooks on animal behavior, and they are likely to remain of major influence on ethologists for many years to come.

The neuroethological phase

Despite the elegance of his studies in the 1960s and early 1970s, Walter was well aware of the limitations of a purely behavioral analysis. Since his ultimate goal was to understand the physiological mechanisms that underlie behavior in an organism, he had to acquire additional expertise in neurobiology. The foundation for this step was laid in 1969 when he asked Ted Bullock in a letter whether a position would be available for him at the Scripps Institution of Oceanography of the University of California at San Diego in La Jolla, where Bullock headed the Neurobiology Unit. The move from Seewiesen to La Jolla, and thus the transition from ethology to neuroethology, materialized in 1972 when Heiligenberg joined the laboratory of Bullock as a postdoctoral fellow. This marked the beginning of the work that made him a world leader in neuroethology—the analysis of the neural basis of the behavior of weakly electric fish.

One year later, he was appointed to the faculty of UCSD and established his own research group. Another 3 years later, in 1976, he was promoted to the rank of full professor. In the same year, Ted Bullock introduced him to the habitat of electric fish in the Amazon. Over the following 17 years, ten more expeditions—to Surinam, Brazil, Panama, and Venezuela—followed. They all reflected another important aspect of Walter’s research philosophy: the attempt to relate the results of physiological experiments conducted under the rigor of the laboratory to observations of the behavior of the animal in its natural environment.

Walter’s fame as a world leader in neuroethology is primarily based on his work on the so-called “jamming avoidance response” of certain weakly electric fish. This behavior was first described by Akira Watanabe and Kimihisa Takeda (1963) in the glass knifefish Eigenmannia sp. This gymnotiform fish continuously generates, by means of an electric organ, a sinusoidal electric signal around its body. Although the discharges vary among individual fish within a species-specific frequency range of approximately 250–600 Hz, they are usually extremely constant in a given individual even over hours and days. One function of the electric organ discharges (EODs) is to mediate location of objects in the closer vicinity of the fish and to analyze their electric properties. This ability, commonly referred to as “electrolocation”, is based on the fact that objects that differ in their electrical impedance from the surrounding water cause distortions of the fish’s electric field. These distortions are sensed by electroreceptors located on the body surface and analyzed by specialized brain structures.

Despite the high degree of regularity of the EODs, frequency shifts may occur when two fish with similar frequency meet. Then, each fish shifts its own EOD frequency away from the neighbor’s. This action is referred to as the jamming avoidance response (Fig. 4). It represents a complete social behavioral pattern that can be readily evoked in an experiment by stimulating the fish with an external sinusoidal electric signal in the frequency range of its own EOD.
Fig. 4

Jamming avoidance response of the glass knifefish Eigenmannia sp. In the experiment, the fish was stimulated with an extrinsic sinusoidal signal 1.6 Hz above its baseline frequency. This signal imitates the electric organ discharge of a neighboring fish. Immediately after the beginning of the stimulation (indicated by the black bar), the fish shifted its discharge frequency away from the stimulus frequency. Upon reaching a discharge frequency approximately 3.5 Hz below the stimulus frequency, the fish made several attempts to return to baseline frequency by raising its frequency (indicated by arrows). However, as soon as the difference between the fish frequency and the stimulus frequency became smaller than approximately 3 Hz, the fish lowered its frequency again. Upon termination of the stimulation, the fish started to return to its baseline discharge frequency. Each data point represents the mean discharge frequency of 100 instantaneous, cycle-by-cycle, frequency measurements. (From Zupanc and Bullock 2005)

Together with his postdoctoral fellow Henning Scheich, Bullock had carried out a thorough characterization of the jamming avoidance response and made a computer model that accurately predicted the dynamics of the input–output relationship for a range of different stimuli (Bullock et al. 1972). These stimuli were defined frequency differences, dF (=frequency of the neighbor’s signal minus frequency of the fish’s signal), that were set and maintained by a feedback computer. The model suggested by Bullock and co-workers included several key parameters of the stimulus but lacked a measure of the assumed function, namely the accuracy of detecting objects. It was Walter Heiligenberg’s merit to overcome, within a year of his arrival in La Jolla, this deficit and to demonstrate that the jamming avoidance response and the electrolocation capability of Eigenmannia sp. are closely related to each other. By forcing the fish to avoid clashes with objects swung in a sinusoidal manner, he showed that contamination by signal frequencies in the range of the fish’s own frequency result in detrimental effects on the electrolocation system. To avoid such jamming, the fish shift their frequency so as to hold the difference in the discharge frequencies of the two fish at a level of dF of approximately 20 Hz.

In order to perform the jamming avoidance response, the fish must determine whether its own frequency is lower or higher than the frequency of the interfering signal. Contrary to expectation, the fish achieves this task not by using the frequency of the pacemaker nucleus in the medulla oblongata, whose oscillatory activity drives the electric organ discharge, as an internal reference. Instead, the decision of the fish in which direction to shift its frequency is exclusively based on the analysis of afferent information contained in the beat pattern that results from the superimposition of the two jamming EOD signals—its own and the neighbor’s. Specifically, the mixing of the two signals causes modulations at beat frequency of both amplitude and phase of the superimposed signal in reference to the pure fish signal. It was Walter Heiligenberg’s merit to have shown that analysis of the pattern of the amplitude and phase modulation enables the fish to determine which of the two signals has the higher and which the lower frequency, and thus to shift its own frequency in the “correct” direction. One direction is released when phase lag accompanies amplitude increase, and the other when phase lead coincides with a rise in amplitude.

Based on the elucidation of the computational rules that govern the jamming avoidance response, Walter and his group performed, over a period of 20 years, a comprehensive analysis of the neural substrate underlying this behavior, tracing the pathway from the sensors all the way to the effector organ through more than 14 orders of neurons (Fig. 5). They showed that at lower levels of electrosensory processing—in the electrosensory lateral line lobe (ELL) of the hindbrain—amplitude and phase information are analyzed in separate neuronal pathways, a mode of action which has become known as “parallel processing”. Convergence of these two pathways occurs at higher levels of neuronal processing—in the torus semicircularis (TSd) of the midbrain. However, although amplitude and phase information converge at this level, and certain toral neurons are the first in the neuronal hierarchy that are able to recognize the sign of the frequency difference between the fish signal and the signal of the neighboring fish, the sign selectivity of these neurons is ambiguous, as it depends on the orientation of the jamming signal, which is determined by the orientation of the neighbor relative to the fish. Cells that encode the sign of the frequency difference between the signal of the fish and the discharges of the neighbor unambiguously, i.e. independently of the orientation of the jamming stimulus, are found at the next higher level of electrosensory processing, namely in the nucleus electrosensorius (nE). At an even higher level of the neuronal hierarchy, in the central posterior/prepacemaker nucleus (CP/PPn), the response properties of sign-selective neurons reflect the behavioral properties of the jamming avoidance response to a still higher degree. A similar gradual rise in the specificity and acuity of the information extracted with increasing level of sensory processing within the neural hierarchy has been demonstrated in many other systems.
Fig. 5

Flow diagram of the neural pathway involved in central control of the jamming avoidance response in Eigenmannia sp. Separate types of electroreceptors, called P-type receptors (P) and T-type receptors (T), in the skin of the fish encode local amplitude and local phase, respectively. Different cell types in the electrosensory lateral line lobe (ELL) of the hindbrain process the amplitude and phase information independently. The phase-coding cells of the ELL project exclusively to lamina 6 of the dorsal part of the torus semicircularis (TSd) in the midbrain. The amplitude-coding cells of the ELL send axons to various laminae below and above this layer. A network of neurons within lamina 6 computes phase differences between any two points on the body surface, and thus the differential phase. Among various response characteristics of neurons in the layers below and above lamina 6 is the one of a class that responds to both amplitude and phase modulation. This convergence of amplitude and phase information appears to be achieved by vertical connections between the different layers. Various layers of the TSd project to the nucleus electrosensorius (nE). Stimulation of one area, the nE↑, causes accelerations of the electric organ discharge, while stimulation of a different area, the nE↓, results in decelerations. Single cells can be found in the respective areas that respond to the coincidence of phase lag while amplitude is increasing, and others that respond to phase advance while amplitude is increasing. Yet, in the natural situation presumably some number of such cells are necessary to activate the correct jamming avoidance response. The nE↑ innervates, via excitatory synapses (indicated by arrow), the central posterior/prepacemaker nucleus (CP/PPn) in the dorsal thalamus. The nE↓, on the other hand, provides, via inhibitory (GABAergic) synapses (indicated by circle), input to the sublemniscal prepacemaker nucleus (SPPn). Final motor control is achieved in the pacemaker nucleus (Pn) of the medulla oblongata. Neurons of the CP/PPn innervate pacemaker cells in the pacemaker nucleus. This input is mediated by AMPA-type of glutamate receptors. On the other hand, neurons of the SPPn innervate relay cells, which are also situated within the pacemaker nucleus. This input is mediated by NMDA receptors. As a final step, the relay cells project to spinal motoneurons that innervate the electric organ. Synchronous depolarization of the electrocytes comprising the electric organ generates the electric organ discharge. (After Heiligenberg 1991e ; Metzner 1999)

Despite this convergence, there are no decisive higher-order neurons that are individually essential, and thus indispensable, for the execution of the jamming avoidance response. In the natural situation presumably not a single cell, but a number of cells activate the correct jamming avoidance response. The advantage of such a distributed organization in sensory processing is that the system can not only tolerate local loss of sensors and computational elements, but also afford to use “cheap” (as Walter used to call them) sensors in the periphery that are rather broadly tuned. Sufficiently high acuity of perception can still be centrally obtained by robust mechanisms of averaging and interpolation.

As a corollary, novel neuronal systems might originate in the course of evolution from such “cheap” parts that existed already earlier. New functions are achieved by modifying networks originally adapted to other functions. Since the need for new functions in the future cannot be anticipated, some imperfection in the design of the neuronal hardware inevitably will happen. This latter aspect was a core element of Walter’s understanding of evolution and contradicts the opinion of many people, including biologists, who believe that biological systems always reflect optimal solutions to a given problem.

Tragic meets happiness

Walter Heiligenberg’s life and death are of the kind that writers use for novels. From childhood, he was very interested in classical music, and later in his life he even thought about giving up science and devote himself entirely to music. This interest in music was shared with his first wife, Zsuzsa, a pianist from Hungary, with whom he had three children. After Zsuzsa died of cancer in 1991, Walter Heiligenberg re-married—again a musician, Wendy Thompson, who played violin at the renowned Bayerisches Rundfunk Symphonieorchester in Munich. Wendy and Walter had met through Hansjochem Autrum, who lived in the same apartment house in Munich as Wendy. Since Autrum was an enthusiast of classical music and often attended concerts of the Bayerisches Rundfunk Symphonieorchester, they knew each other well. When Autrum on his first and only trip ever to the USA visited the lab of Walter Heiligenberg, Wendy Thompson came to La Jolla as well, because she played at that time in Los Angeles. In 1993 Walter and Wendy married.

Following their marriage, Walter traveled back and forth between La Jolla and Munich, where Wendy continued to live and work. On 8 September 1994, he headed for San Diego. He was invited to give, on his way back, a lecture at Pittsburgh, and since his plane from Europe had arrived in Chicago earlier than expected, he decided to take a flight—US Air 427—different from the one for which he was scheduled. Just a few minutes before landing at Pittsburgh, the Boeing 737 went into a nosedive and crashed. All 132 people aboard the aircraft, included Walter, perished. Eighteen days after his death, Wendy gave birth to their daughter, Maria Clara. In its final report released 5 years after the crash, the National Transportation Safety Board concluded that the aircraft’s rudder, a moveable control surface hinged to the tail fin, became jammed for still unknown reasons, forcing the plane into an almost vertical roll at about 6,000 ft. (2,000 m).

Walter Heiligenberg’s way of doing science

Walter Heiligenberg’s fame as a researcher, which has not diminished even 12 years after his death, is not only based on his scientific achievements, but also on his rather unconventional way of doing science. Everybody who knew Walter remembers him as the most alive person they had ever met. His optimism was unlimited. He appeared to be driven by an inexhaustible source of energy that let him succeed where others did not even try. He disliked bureaucracy and formalities. Even at international conferences, he would give his talks clad in white turtle-neck sweater, tennis shorts, and beach sandals. He came into his laboratory every day, including Sundays and holidays, and it was certainly not more than a handful of days that he took off per year. He seemed to draw his satisfaction from running his own experiments, every day and all day long—patiently sitting at his “rig” (as he called it) that was packed with physiological equipment, exposing the brain of the tiny fish, performing delicate surgeries, searching over days and weeks for specific cells, until he finally found the right ones. This event was shared with the rest of the lab by a loud scream “look, look—there it is!”.

Every associate was also amazed how Walter managed to keep up with the literature, reviewed and edited manuscripts for this journal (as he did for many years), write grant applications, give stimulating lectures, receive visitors from all over the world, and maintain a large correspondence, without the assistance of a secretary, while at the same time regularly attended his wife’s and other concerts. He avoided overpopulating his lab, but welcomed pre- and postdoctoral associates with diverse backgrounds. At the time of his death, he had Ph.D. students taking their degrees in physics, marine biology, neuroscience, and philosophy. These students with diverse backgrounds formed the basis for the multidisciplinary approach employed in Walter’s studies. This approach included behavioral studies, neurophysiological experiments, neuroanatomical investigations, and computational analyses, and, shortly before his death, Walter and his group even entered the field of molecular biology.

We doubt that Walter ever formally sat down with students to discuss their progress. Nevertheless, he was an extremely inspiring advisor. He led by example, not by forcing a graduate student or a postdoc into a certain direction. This freedom had an enormous impact upon the creation of new directions of studies. He encouraged people of his group to go to other laboratories to learn new techniques and enter new fields. A good number of postdocs, and even graduate students, started completely new projects while still in his lab, independently of his own work, but backed by his unspoken support. They took these projects with them when they left the lab and established their own research groups. These projects include exploration of the neural and endocrine basis of communicatory behavior in weakly electric fish (Meyer 1984; Meyer et al. 1984; Keller et al. 1986, 1991; Kawasaki et al. 1988; Metzner and Heiligenberg 1991; for reviews see Metzner 1999; Zupanc 2002); the study of plasticity of sense organs and neuronal systems in adult fish (Zupanc and Heiligenberg 1989; Zupanc and Zupanc 1992; for review see Zupanc and Maler 1997; Bastian and Zakon 2005; Zupanc 2006); and the phylogenetic analysis of electric fishes using a combination of morphology, electrophysiology, and mitochondrial sequence data (Alves-Gomes et al. 1995). A significant number of papers were published from his lab without his name listed as a co-author—a degree of independence unthinkable in most other labs. It is especially this way of doing science and of opening new vistas that will continue to inspire, for many years to come, his students and others who had the privilege to know him. The review articles in this issue, written by former students, collaborators, and friends, give impressive testimony of this.


We thank Masashi Kawasaki, Mark Konishi, and Marianne M. Zupanc for helpful comments on the manuscript. This article was written while G. K. H. Z. was a visiting scholar at the Department of Neurosciences of UCSD in La Jolla.

Copyright information

© Springer-Verlag 2006