Active Electroreception in Weakly Electric Fish
Summary and Keywords
American gymnotiformes and African mormyriformes have evolved an active sensory system using a self-generated electric field as a carrier of signals. Objects polarized by the discharge of a specialized electric organ project their images on the skin where electroreceptors tuned to the time course of the self-generated field transduce local signals carrying information about impedance, shape, size, and location of objects, as well as electrocommunication messages, and encode them as primary afferents trains of spikes. This system is articulated with other cutaneous systems (passive electroreception and mechanoception) as well as proprioception informing the shape of the fish’s body. Primary afferents project on the electrosensory lobe where electrosensory signals are compared with expectation signals resulting from the integration of recent past electrosensory, other sensory, and, in the case of mormyriformes, electro- and skeleton-motor corollary discharges. This ensemble of signals converges on the apical dendrites of the principal cells where a working memory of the recent past, and therefore predictable input, is continuously built up and updated as a pattern of synaptic weights. The efferent neurons of the electrosensory lobe also project to the torus and indirectly to other brainstem nuclei that implement automatic electro- and skeleton-motor behaviors. Finally, the torus projects via the preglomerular nucleus to the telencephalon where cognitive functions, including “electroperception” of shape-, size- and impedance-related features of objects, recognition of conspecifics, perception based decisions, learning, and abstraction, are organized.
Two orders of weakly electric bony fish (African mormyriformes and American gymnotiformes) show an ensemble of sensory modalities that have evolved in a convergent manner. This system has four main components: active electroreception, passive electroreception, mechanosensory lateral line, and proprioception. Two main efferences are involved in this reafferent system: the self-generation of a weak electric organ discharge (EOD) and the ability to move the body in order to optimize sensory reafference. Passive electroreception and lateral line mechanoception are driven by sources of energy external to the individual; active electroreception and to some extent proprioception (when responding to self-generated movement) are “active” systems that extract information carried by self-generated forms of energy. This multimodal ensemble of active and passive reafferent sensory modalities has motivated the comparison of this system with human haptic senses (Caputi et al., 2013).
This article focuses on active electroreception in which the presence of objects modulates the electric field created by the EOD.
Extensive bibliography on electroreception may be found in three books written by research leaders in the field (Fessard, 1974; Bullock & Heiligenberg, 1986; Bullock et al., 2005). In addition, the history and behavior of electric fish have been addressed by Moller (1995). Comprehensive essays on specific topics have been contributed by Heiligenberg (1977, 1991) and Kramer (1990). Progress in the field has been documented in special issues collecting review papers from the regular meetings of the electric fish community (Journal of Physiology, Paris 1979, 2002, 2008, 2016; Journal of Experimental Biology, 1999, 2013; and Journal of Comparative Physiology A, 1993, 2006). Only key contributions and reviews are cited in the text.
From Darwin’s Puzzle to the Formal Recognition of Electroreception
Charles Darwin (1872) identified electric organs morphologically in what are now known as weakly electric fish as a case of special difficulty for his theory. At that time, it was “impossible to conceive by what steps these wondrous organs have been produced. But this is not surprising, for we do not even know of what use they are” (Darwin, 1872). The clues for unveiling this incognita came from two previous observations (Moller, 1995). Electric eels were able to detect when a pair of wires was short-circuited inside their tank (Walsh & Seignette, 1772), and rays showed numerous orderly arranged openings of ampullary organs (Lorenzini, 1678, cited by Moller, 1995). While Walsh’s experiment showed the use of EODs for evaluating changes in the surrounding impedance (active electroreception), Lorenzini described the receptors providing rays as having the ability to receive the electric messages from conspecifics (electrocommunication).
An additional missing piece of the puzzle was that some African electric fish was provided by Erdl and confirmed later by Marcusen who showed a large brain, a giant cerebellum, and a nonscaled skin covered by “pores” innervated by the lateral line nerve (Moller, 1995). However, it was not until the beginning of the 20th century that Franz (1912) proposed that “[a] question for future research would be to consider whether perhaps the mormyrids themselves appreciate the electric shocks of members of their species and thereby adjust their own behavior …”
The answer to Franz’s question started to emerge when the head curator of the New York Aquarium, H. W. Coates, and the physicist, R. T. Cox, traveled through Brazil to study the bioelectricity of electric eels. In her essay about the trip, Cox’s wife (Shackelford, 1941) recounted how her husband threw a pair of wires into an electric eel pool and plugged the other ends into a city power outlet. To her amazement, eels came flocking to the wires’ endings. This was the first proof that electricity was the specific energy carrier of a nonhuman sense. This atypically reported discovery of electroreception was formalized when H. W. Lissmann (1951) reported that a wave electric fish, Gymnarchus niloticus, showed “(i) a mode of locomotion through waves passing over the extended dorsal fin and executed forwards and backwards with equal ease; (ii) a marked ability of the fish to avoid obstacles while swimming backwards … it seemed of interest to consider whether this (electric) organ may enable the animal to detect objects in the vicinity of its body.”
The presence of electroreceptors giving fish such an ability was first observed at the Makerere College Medical School of Uganda when P. G. Wright (1958) reported that the lateral line nerve was excited by the presence of a weak electric stimulus delivered from two electrodes near the skin. Soon after, Lissmann (1958) and Lissmann and Machin (1958) speculated that the skin pores innervated by the lateral line nerves were the specialized organs for sensing the self-generated electric field. The work of Lissmann and his colleagues provides a clear demonstration of the ability of these fish to detect the electric impedance of objects. Their work also provides evidence of electrocommunication, explains the mechanisms of electric imaging, and discusses the convergent evolution of the electric organs and their role in active electroreception in the two main taxa of electric teleosts.
Lissmann’s seminal studies were soon followed by the anatomical and functional description of electroreceptors (Fessard & Szabo, 1961; Bullock et al., 1961). The findings of an extreme sensitivity of Lorenzini’s ampulla to electric fields (Murray, 1962) and the discovery of their use for detecting fields emitted by conspecifics (communication) and prey (passive electroreception, Kalmijn, 1971) showed that electroreception is not limited to the actinopterigii.
During the next decade, it became evident that active and passive electroreception are the dominant senses in weakly electric fish and the existence of species-specific communication channel. The field of electroreception continues to grow and contribute to our general understanding of the nervous system.
The electric organ (EO) and the fish’s body are an inextricable imaging unit. Depending on the species, EODs may last between 0.1 and 10 milliseconds. This duration may be the same as or much smaller than the silent interval between EODs, yielding either a continuous wave (wave species) or discrete pulses (pulse species), respectively. Among pulse fish there is a major difference in electrogenic mechanisms. The EO of mormyrid fish is restricted to the tail, and the time course of the electric field is similar at every place, whereas the EO of gymnotid fish is distributed along the body and their EOD is characterized by a complex spatiotemporal field resulting from the sum of different components originating from different regions of the body (Fig. 1). These three strategies—wave fish, pulse fish with distributed Eos, and pulse fish with localized Eos—are associated with differences in signal encoding at electroreceptors and processing in central structures.
The Roles of the EODs
EODs are critical for two major functions: active electroreception of objects differing in conductance from water and electrocommunication in which electric fish send and receive information with conspecifics, a subject not treated here.
The current generating units of the EOs are called electrocytes. They are synchronously and rhythmically excited by commands from the nervous system almost without interruption from larval stages to death. Electrocytes respond to these commands with brief pulsatile action currents that flow through the surrounding tissue. These currents polarize the fish’s body, transforming it into a distributed current source flashing electric fields polarizing, in turn, the surrounds. The electromotive force characterizing the fish’s body as an electrical source is a species-specific spatiotemporal pattern. The temporal aspect of the EOD generated at each region of the fish’s body is determined by the responses of individual electrocytes to neural commands, whereas the spatial aspect is determined by the location, density, and responsiveness of the electrocytes (Bennett, 1971; Caputi, 1999, 2011).
In the fish’s natural habitat, shores, stones, water surface, plants, and other animals with impedances different from that of water introduce local disturbances in the electric field (referred to as “object perturbing field”). In the same way that the moon reflecting the sun’s light can be regarded as a luminous object, objects in a fish’s electric field can be regarded as electric sources caused by the polarizing effect of the fish’s EOD field (“object stamps on the fish’s electric field”). The “object-perturbing field” due to the object is projected onto the skin where it is sensed by the fish’s electroreceptors. The contribution of each single object to the transcutaneous pattern is called the “electric image of the object” (Pereira & Caputi, 2010, Fig. 2).
Electric images of the scene are a complex sum of the effects of nonconductive objects that act as mirrors blocking the passage of current and conductive objects comparable to color glasses that let currents pass currents, both adding their own sensory flavor. Current flow caused by the fish’s EOD is accompanied by polarizations of both water and objects. An object that is more conductive than water projects an electrical image on the skin that is the opposite polarity of the image projected by an object that is less conductive than water, both objects being polarized by the same EOD. Therefore, what matters for electric imaging is the differential polarization of objects and water.
Temporally, the active electric sense can be considered an “instantaneous” sense since polarization of surrounding objects occurs at near the speed of light.
Spatially, the active electric sense is a short-range sensory modality. Electric current density decays rapidly with distance because of the quadratic increase in the distribution surface and the presence of two opposite poles. A decay of similar nature occurs during the image projection stage when the polarization of the object reflects back onto the skin of the fish. Image attenuation with distance increases as the product of both effects. The contrast of the image of a small moving object receding from 1 to 4 cm apart from the skin is reduced about 250 times, vanishing as visual images do in fog. Moreover, while the visual environment is focused on the retina and thus has a point-to-point correspondence with the focused surface (apposition images), electric images are spread over the fish skin, increasing in size as their distance increases (superposition images), and images of two objects became fused with increasing distance. This reduces even more the spatial resolution of active electroreception.
The Roles of Body Conductance and Movements
The fish’s body is a highly conductive path for self-generated currents inextricably present in the scene mediating important effects such as: (1) funneling the signals of the object and conspecifics to an electrosensory fovea, (2) selective object polarization, and (3) a juxtacutaneous hypersensitive fringe.
First, because fish conductivity is larger than that of water, because fish have a predominantly longitudinal dimension, and because the cross-sectional area becomes smaller toward the tail, electric currents (generated either by the fish itself or by external sources such as other fish) find a main “highway” reorienting them toward the snout. The electrosensory mosaic in this region of the skin shows (1) the largest electroreceptor density; (2) the largest variety of electroreceptor types; and (3) the largest central representation compared to the rest of the body. The ability to control the position of this region allows the fish to enhance the resolution of the electric image in the perioral region as occurs when eye movements are used to bring the image of the objects of interest onto the fovea. Because of all these arguments, this region has been likened to as an “electrosensory fovea” (Castelló et al., 2000, Fig. 3).
Second, the control of body curvature allows the funneling of such currents toward objects of interest. Current through the perioral region acts like a sprinkler on the object, increasing its relative polarization and the contrast of the image at the electrosensory fovea. In addition, the snout specializations in pulse mormyriformes add extra degrees of freedom to the foveal regions. For example, the species Mormyrus rume proboscirostris has a long proboscis that allows it to manipulate its eggs by means of water jets (Schugardt & Kirschbaum, 1998), and a fingerlike mobile chin appendage in Gnathonemus petersii is used for exploration on the bottom of a stream in the search of food (Amey-Özel et al., 2015).
Third, objects acting as secondary sources near the skin, besides projecting a primary image, can also reciprocally polarize the fish’s body. This recursive loop of mutual polarization between object and fish’s body can only be significant for shaping object image when the surfaces of the object and the skin are at very short distance. This juxtacutaneous fringe of “hypersensitivity” allows fish to evaluate the “electric texture” of object surfaces.
Peripheral Encoding of Electric Images
Electroreceptive mosaics are characterized by different electroreceptor types and their distribution over the skin. It is a constant finding across species that the variety and density of receptors decreases from head to tail and that the peak density is largest in the perioral region (electrosensory fovea, section 3.2, Fig. 3).
Electroreceptor organs show a peculiar form of “electroelectrical” transduction. While the stimulus is a transcutaneous electric field, the response is a change in ionic conductance of the plasma membranes in a population of sensory cells forming part of an electroreceptor organ. These sensory cells synapse on primary afferent fibers that respond with trains of action potentials encoding the local transcutaneous electric field as a series of interspike intervals.
Electroreceptor organs can be grouped into two broad categories: ampullary and tuberous. Ampullary receptors are responsible for passive electroreception. They are the most abundant and are present in cartilaginous and bony fishes. Although they respond to the EOD, and thus they might be involved in active electroreception, the behavioral roles of these responses have not been studied extensively. They project to a cerebellum-like structure that allows for the generation and subtraction of negative images of predicted or expected sensory patterns. For example, corollary discharge signals originating in the nucleus commanding the EOD in mormyriformes or in the respiratory centers generating respiratory potentials as gills are opened and closed in elasmobranchs, eliciting negative images or representations of the electrosensory input pattern that has followed recent motor commands. This representation is stored in the synaptic weights at synapses between parallel fibers conveying the corollary discharge and the dendritic trees of the neurons the cerebellum-like structures. Tuberous receptors appear to be specifically designed to respond to the EOD, and their structure and function are treated here in detail.
Encoding of Self-Generated Electric Images by Tuberous Receptors
In general, tuberous electroreceptors consist of an epithelial capsule having a round flask shape. They are composed of a sheath of flattened cells interconnected by numerous desmosomes and tono-filaments that result in large resistance to electric currents. The bulb expansion at the bottom generally consists of a single sensory chamber connected to the exterior through a channel filled with loosely packed epithelial cell. Mormyromast electroreceptors of pulse mormyriformes have two sensory chambers, however, innervated by separate afferent fibers. A detailed account of the morphology in different taxa can be found in the books edited by Fessard (1974) and Bullock and Heiligenberg (1986).
Electroreceptors are broadly tuned to the spectral power density of the EOD. In some cases, they also show specific responses to the different time course of the stimulus even when the power spectral density remains constant (Fig. 4).
Tuning is important for three main reasons: First, tuning to the self-generated EOD increases the differences between self- and non–self-generated signals and the differences between conspecifics and other sources. Second, changes in the amplitude or phase spectra may inform the fish about the presence of complex impedances caused by capacitive elements of the object. In the presence of more than one type of electroreceptor with different tuning, this opens the possibility of perceiving more than one dimension in the electric image as, for example, occurs in color vision or tone audition. Third, individuals are able to discriminate conspecifics among other electric fish and also to distinguish male from female by detecting temporal and spectral cues in the EODs. Electroreceptor tuning to species-specific EODs is seasonally modulated in such a way that the sharp frequency tuning of electroreceptors parallels the changes induced by sexual hormones to the power spectral density of the EODs (Meyer & Zakon, 1982).
The mechanisms of frequency tuning in tuberous electroreceptors are not well understood. It is thought that the channel, capsule, intrinsic properties of the sensory cells, and the responsiveness of afferent fibers determine how these receptors and their afferents respond to changes in the amplitude and time course of the transcutaneous electric fields. Owing to the asymmetry between the plug and wall impedance, it is probable that the electric currents stimulating sensory cells flow through a complex impedance path. Indirect evidence of the role of the intrinsic properties in tuning is the common observation of attenuated oscillatory responses to the self- and conspecific-generated electric fields at the receptor channel opening.
Differences Across Taxa in Electric Image Encoding
Gymnotiformes and mormyriformes have evolved in a convergent manner two main types of tuberous receptors innervated by, respectively, different subtypes of primary afferents that in turn encode differently the actively generated electrosensory images and project to different structures.
These two types of tuberous receptors give origin to the so-called fast and slow electrosensory pathways, a common organization of sensory processing observed in many vertebrate sensory modalities.
Signal Encoding at Fast Paths
In the first central relay of these paths, the latency of the single spike evoked by the EOD is on the order of 1–2 ms and shows only small changes (below 0.1 ms) with changes in stimulus intensity, indicating powerful temporal resolution.
In wave gymnotiformes and mormyriformes, fast-conducting afferents are called T or S units, respectively. They encode the local phase of each EOD cycle, although in high-frequency fish they might not follow the EOD one to one. Because frequency beat occurs when two conspecifics are close to each other, the phase difference of the field stimulating two adjacent regions of the body and the corresponding timing of this afferent subtype is modified. Therefore, the pattern of firing time across populations of this afferent subtype at different regions of the body informs the fish about the frequency beat caused by the sum of self- and conspecific-generated EODs.
In pulse gymnotiformes and mormyriformes, these afferents are called pulse markers and K-afferents. They encode the timing of the self- and allo-generated EODs. K-afferents and pulse markers are thick fibers showing a high-conduction velocity that allows the signals to reach the central nervous system in 1 or 2 milliseconds after the EOD. They fire single-spike tightly phase locked to the transition between outward and inward current deflections of the self-generated EOD and to similar voltage transitions generated by other electric fish. Type I electroreceptors are characteristic of the foveal region in pulse gymnotiformes, but Knollenorgans are extremely sensitive and absent from the fovea, though present on more caudal parts of the head and the flank.
Signal Encoding at Slow Paths
In wave fish, afferents arising from Type II receptors (P or S units) fire at a near fixed phase of the EOD cycle, but they skip an uncertain number of cycles between firings and their firing before a constant-amplitude stimulus is apparently random. In the presence of objects, the probability of firing is a smooth increasing function of the change in peak-to-peak amplitude of the local signal. An important property of gymnotiformes’ afferents is the presence of spike firing adaptation, which explains the apparent randomness of the spiking. This means that when an afferent fires at short intervals, either it needs some time to recover excitability or it requires an increase in stimulus intensity to fire again. This negative feedback on excitability obeys processes with two different time courses: a fast adaptation in which spike feedback effects decay following an exponential law, and a slow adaptation in which they decrease time following a power law.
The effective stimulus for the electrosensory system of wave fish is the temporal and spatial pattern of amplitude and phase of a sinewave-like carrier. Three main types of amplitude result from the addition of the self-generated EOD and another sinewave: (1) Objects behave as a virtual source of the same frequency as the EOD; (2) static conspecifics have similar frequencies and thus cause a wave beat; and (3) moving conspecifics cause changes in the envelope of the beat. In addition, communication signals consisting of transient changes in frequency cause changes in the phase and frequency of the self-EOD amplitude modulation.
Although the modulations caused by objects are generally local and vary slowly with the fish’s or objects movement, the frequency difference naturally observed between conspecifics causes amplitude modulation, with a repetition rate equal to the frequency difference. Receptor adaptation dynamics facilitates the synchronous firing among receptors belonging to relatively large patches of skin during the increasing phase of the beat and their relative silence during the decreasing phase. Sudden changes in frequency (so-called “chirps”) cause either transient increases or pauses of synchronism and convey communication signals. Therefore, object features are encoded in the probability of firing of individual receptors, and conspecific signals are encoded by the synchronism of afferent populations in a local neighborhood (Clarke et al., 2015).
In pulse fish, slow-conducting primary afferent fibers originating the slow electrosensory pathway fire trains of spikes. The latency and number of spikes in the train encode the waveform distortion of the local EOD.
At least two subtypes of afferents arising from Type II receptors have been described in the species of three families of gymnotiformes: (1) narrow-band fibers showing V-shaped tuning curves with best frequencies near the peak of the EOD power spectra and (2) wide-band fibers showing little frequency selectivity. Consistent with the hypothesis of an electrosensory fovea previously described, all subtypes of Type II receptors occur at the head and perioral region of pulse gymnotiformes where they show their highest density (10 times larger than at the back of the head and 100 times larger than at the body region), while wide-band fibers are homogeneously distributed.
Pulse mormyrids show complex receptor organs. Mormyromasts have two chambers—an outer chamber that communicates to the exterior at its apex and an inner chamber that communicates with the outer chamber. Typically, two fibers (type A) receive chemical synaptic contacts from the cells of the outer chamber, and a single fiber (type B) innervates the cell of the inner chamber. While A fibers respond only to change in stimulus amplitude, B fibers respond to both amplitude and EOD time course. In pulse mormyrids, mormyromast density also increases from caudal to rostral regions, having their peak at special structures as the Schnauzenorgan. As gymnotiformes, these fish’s regions show the largest central representation at the electrosensory lobe.
Because of receptor encoding variety at the fovea, both pulse gymnotiformes and mormyriformes evaluate differences between objects of different impedance, including capacitance. Importantly, although receptors evaluate changes in time course, it is not likely that fish strictly distinguish capacitance from resistance because for such a purpose more parameters in the EOD waveform are required to be evaluated than receptor types occur. Therefore, rather than the human concepts of capacitance and resistance, these fish evaluate an impedance-related feature that could be defined as “electric color.”
Early Processing of Electric Images at the Brainstem
Although the organization of electrosensory pathways in fast and slow branches is similar among the wave and pulse of both African and American electric fish, there are also many differences.
First, signal processing is heavily dependent on the presence of a corollary discharge associated with the EOD motor command in pulse mormyriformes, but such a corollary discharge is not present in other taxa.
Second, the EOD emission regime makes a major division. Wave and pulse gymnotiformes fish, both lacking corollary discharges, have stimulus pattern and receptor encoding properties that determine different central processing of sensory signals on a very similar central circuitry “hardware.”
Third, afferent projection into the electrosensory lobe is different. In gymnotiformes, primary tuberous afferents trifurcate after entering the brainstem. Their branches project in somatotopic register to each of three maps at the electrosensory lobe (afferent fibers from ampullary receptors project to a fourth map). While afferents giving origin to the fast pathway project onto a specific type of second-order neuron, the spherical neuron, which in turn projects directly onto the mesencephalon, afferent fibers giving origin to the slow pathway project to the granule layer of cerebellum-like cortex where they make synaptic contacts with several types of second-order neurons.
In mormyriformes, Knollenorgan’s afferent projects onto ovoid adendritic cells of a deep nucleus in the electrosensory lobe, which in turn projects onto the mesencephalon. Type A and B fibers from mormyromasts project to two different maps in the cerebellum-like cortex of the lobe (afferent fibers from ampullary receptors project to a third map).
The wave fish Gymnarchus niloticus and Sternopygus macrurus are special cases and are not treated here in detail.
The Fast Path Consists of a Filter Followed by a Time Comparison Circuit
The primary afferent population of the fast electrosensory pathway in all electric fish encodes transcutaneous electric fields as a pattern of relative latencies. Mixed chemical and electrical synaptic contacts transfer information between primary afferent terminals on their postsynaptic target’s adendritic (mormyriformes) ovoid or pausi-dendritic (gymnotiformes) spherical cells. These neurons in turn project onto the mesencephalon. Although the target location and circuit details differ among taxa, it has the common feature of being a time comparison circuit (Fig. 5) and in all cases appears to be involved in the trade-off between self- and allo-generated species specific signals.
Special Features of Wave Fish
Spherical neurons of the electrosensory lobe project to large and small neurons either in the magnocellular nucleus (pulse fish and Apteronotidae) or layer VI of the torus semicircularis (Eigenmannia). Each spherical cell axon makes immediate excitatory contact on large neurons and then after some wandering through the nucleus terminates with excitation on small neurons. The long traveling distance between the contacts with large and small neurons adds a significant delay. Large neurons, in turn, project on several of the small ones that also receive input from other regions of the skin. This network implements a delayed timing detection circuit as that described by Jeffress (1948) for auditory localization. It is thought that small cells act as coincident detectors with a submicrosecond precision. Therefore, depending on the amplitude and time course of the stimulus at the skin, different sets of small neurons would be activated. One particular role of this circuit is to encode the difference in phase between two skin regions when a wave fish interacts with a conspecific.
To avoid jamming introduced by the slow modulation of their EODs, the faster fish increases, and the slower fish decreases, their respective EODs frequencies. This results in a higher frequency difference as well as a higher frequency modulation of the signal at the skins of both fish. This behavior, referred to as the “jamming avoidance response,” requires encoding and comparison of the periodic changes in peak-to-peak amplitude and the difference in the zero-crossing times between two regions of the fish’s body. This behavior is organized automatically at the brainstem and reveals the great acuity of the wave fish’s neural networks.
Despite their independent evolution, both Eigenmannia and Gymnachus have evolved an identical social behavior that shares the same computational algorithm. In Eigenmannia, differential phase modulation is decoded by a Jeffress’s circuit at layer VI of the torus semicircularis. Large neurons are phase-locked to the zero crossing of the stimulus cycles, but small ones are recruited in a facultative way upon the phase difference between different regions of the body. In Gymnarchus, phase comparison occurs at the electrosensory lobe where a specific neuron type responds to differential-phase modulation in the range of microseconds.
In both taxa, the torus semicircularis shows neurons responding to differential phase or to amplitude modulation alone as well as cells responding to combinations of the modulation in amplitude and differential phase. These neurons project directly and indirectly through the tectum opticum on the nucleus electrosensorius. At this nucleus, a transformation from distributed, topographically organized processing to discrete cell clusters that control antagonistic motor responses is performed. Electrosensorius neurons show robust sign selectivity for the difference in frequency between self- and non–self-generated EODs. Two pathways control the pacemaker rate. One controls frequency increases via excitatory projections on a diencephalic prepacemaker nucleus, and the other that controls frequency decreases through inhibitory projections on the sublemniscal prepacemaker nucleus, which is tonically active in the absence of jamming signals (Heiligenberg, 1977, 1991; Kawasaki, 1993).
Special Features of Pulse Mormyrids
Three lines of evidence suggest that the Knollenorgan fast pathway is used for sensing conspecific signals only. First, Knollenorgans show a very low threshold and small latency variability with the amplitude of the stimuli. Second, at the electrosensory nuclei, this path receives a corollary discharge that generates a strong inhibition on the somata of the second-order neurons at the exact arrival timing of the spikes elicited by the self-generated EOD. In consequence, the axons of second-order neurons only carry information on the timing of conspecific EOD to the nucleus exterolateralis. Finally, exterolateralis lesions block behavioral responses to electric communication signals but leave unscathed the active electrolocation abilities.
At the electrosensory lobe nuclei, endings from fast Knollenorgan primary afferents converge with the terminals of central origin that convey a corollary discharge signal coupled to the EOD motor command onto the second-order cells of the nucleus of the electrosensory lobe. This corollary discharge input causes a strong inhibition to arrive just before the primary afferent input evoked by the fish’s own EOD. Because of this corollary discharge-driven inhibition, downstream, neurons in the mesencephalic exterolateralis nuclei where fibers from the nucleus of the electrosensory lobe terminate never receive the self-generated signal. However, they receive signals evoked by conspecific EODs, occurring out of the brief inhibitory window.
At the exterolateralis nuclei, pulse mormyridae express an “anticoincidence detection” mechanism. Small neurons do receive inhibitory synapses from large neurons excited by axons having contralateral receptive fields. These neurons mediate an inhibitory effect that, depending on the stimulus, duration may anticipate and block the excitatory effects of the direct fibers. This initial temporal analysis done at the exterolateralis is further refined in the nucleus exterolateralis pars posterior, where units tuned to ranges of pulse durations have been identified physiologically (Xu-Friedman & Hopkins, 1999).
Special Features of Pulse Gymnotiformes
Although the circuit is similar in pulse gymnotiformes, its function might be different from that in wave fish for the following two reasons: First, in Brachyhypopomus, the responses of the nucleus magnocellularis to male and female conspecifics differ, suggesting that this path is encoding some time course features of the EOD and might be involved in the identification of potential mates. Second, the spherical cells of Gymnotus are characterized by a long refractory period that has important consequences during conspecific interactions. There is a strong recruitment of spherical neurons by the self-generated EOD which prevents their firing during the first half of the inter-EOD cycle. This poststimulus low-responsiveness window has important consequences when considering that the presence of a conspecific EOD just before or together with the self-generated causes a transient increase of the pacemaker rate, followed by a slow relaxation to the baseline. As a consequence of these transient accelerations, a large part of conspecific EOD effects are blocked by the refractory period (Nogueira & Caputi, 2013).
The Slow Path: Comparison Between Input and Central Expectation Signals
Most of our knowledge of image processing by the slow pathway comes from the long-term studies of the wave gymnotiform fish Apteronotus albifroms led by L. Maler and J. Bastian and their school (see reviews by Krahe & Maler, 2014; Marsat et al., 2012; Chacron et al., 2011) and the pulse mormyriform fish Gnathonemus petersii led by C. Bell and K. Grant and their school (Sawtell & Bell, 2008). Therefore, most of the following description is based on results from these two species (Figs 6 and 7).
Commonalities in Early Electrosensory Processing
In the electrosensory lobe, the skin surface is represented in maps (two in mormyriformes and three in gymnotiformes, plus an additional map for ampullary signals). There are similar strategies for image processing within each map, but each map has a different role and there are differences in mormyriformes and gymnotiformes. There are multiple subtypes of neurons, but both taxa have in common the presence of a dual encoding system in each electrosensory map. First, efferent neurons can be primarily grouped in those that are excited (E) and those that are inhibited (I) by the presence of a conductive stimuli at the center of their receptive field. E and I cells are found in pairs, arranged in topographic columns responding to the same primary afferent input. Thus, the output of the electrosensory lobe is carried by somewhat “mirror-image” encoding systems. This push-pull design (repeated by evolution in many sensory systems; e.g., there are “on” and “off” cells in the retina) appears to be necessary to compensate for the nonlinearity of the neurons and neural networks as processing devices.
Second, inhibitory granule cells project on neighbor granule cells, generating a center-on, surround-off opposition pattern that serves as a “Laplacian filter” to increase the local spatial contrast of the incoming signals. Excitatory and inhibitory granule cells project on E and I efferent cells that respond to either the excitation or inhibition at the center of their receptive fields and with an opposite pattern to the surround.
Third, in each electrosensory map, efferent neurons work as comparators. E and I can be morphologically distinguished by the differential presence of basal dendrites. The E cells’ so-called basilar (in gymnotiformes) or fusiform (in mormyriformes) neurons have prominent basilar dendrites that are excited by afferent input. The I cells, so called nonbasilar (in gymnotiformes) or large ganglion (in mormyriformes) neurons, have small, short basilar dendrites and receive afferent information through inhibitory interneurons, implementing a signal inversion that allows the fish to encode decreases in EOD amplitude below threshold for E cells. In their apical dendritic trees, both cell types receive electrosensory feedback from the nucleus praeminentialis, proprioceptive afferences, and central descending commands. In the case of gymnotiformes, differences in the size of apical dendritic trees also imply a different weight of apical (central commands) and basilar (afferent signals) among neurons subtypes (Figs 6 and 7).
Fourth, because motor commands and the consequent changes in body position strongly affect electrosensory input, unambiguous identification of objects requires integration of information about motor commands and body position with electrosensory information in the sensory centers. Information about body position is provided in both gymnotiformes and mormyriformes by a strong projection of proprioceptive signals (and in the case of mormyrids a corollary discharge from spinal cord) carried by the parallel fibers from the granule mass of the eminentia granularis posterior terminating in the molecular layer of the electrosensory lobes. These fibers inform the electrosensory lobe about the relative position of tail, body, and head. In pulse mormyrids, parallel fibers also convey time-structured corollary discharges associated with the electromotor command. There is time-dependent associative anti-Hebbian synaptic plasticity in the contacts between parallel fibers and apical dendritic trees. Systematic association of an afferent input pattern with distributed parallel fiber input allows the fish to store “negative” images of the past activation that, when summated with actual input, are able to remove the predicted signals and detect novelty.
Fifth, an additional main component of the central input compared with the electrosensory input is the feedback from upper centers, in particular from an intermediate relay nucleus of the brainstem (praeminentialis). In fact, in both mormyriformes and gymnotiformes as well as other electroreceptive but nonelectric fish, the praeminentialis nuclei receive the output of some efferent neurons of the electrosensory lobe and project back onto the apical dendrites of principal cells though direct and indirect (via the eminentia granularis and the parallel fibers) pathways (Fig. 6). This electrosensory feedback has been involved in a global automatic gain control of the response tending to adapt the dynamic range of the system to the mean value of the current signals, and in a localized gain control that may act as a “searchlight,” enhancing the response of some regions of the network to signals of interest (Berman & Maler, 1999).
Finally, much is known about the role of the torus semicircularis and tectum opticum in the organization of electromotor and skeletomotor automatic behaviors triggered by the electrosensory afference. These behaviors are specific to different taxa and with the exception of the jamming avoidance response are not treated here. Interestingly, there is evidence of sparse coding of electrosensory signals in the torus (Chacron et al., 2011). In particular, some neurons specifically respond to electrocommunication signals and presumably to sensory cues of objects. Torus and tectum both project to the preglomerular nucleus. This diencephalic nuclear group is functionally analogous to the thalamus since it is the most important link between the brainstem and the dorsal telencephalon.
Special Features of Pulse Mormyriformes
The EOD of pulse mormyriformes is generated at the tail region, and as a result, its time course is similar all over the skin except for the deformations caused by the presence of objects. Therefore, the effective stimulus for the electroreceptive system is the temporal and spatial pattern of the field created by each EOD. In addition, electromotor and skeletomotor corollary discharges effects are shown on the electrosensory lobe (Fig. 7).
First, there is a nonplastic pathway converging with afferent fibers on granule neurons at the input layer of the electrosensory lobe network that facilitates afferent spikes arriving within a given window after the EOD. Precise arrival times following the self-generated EOD of spikes relative to the EOD command corollary discharge are likely to be a natural code. Therefore, in the electrosensory lobe of G. petersii, afferent fibers contact granule cells and basilar dendrites of the E cells that are directly and also indirectly excited by the afferent train. Granule cells are also contacted by central descending axons conveying an excitatory corollary discharge that is timed at the most likely arrival time of the primary afferent volley. In this way, granule cells work as an “AND” gate for facilitating the entrance of self-generated signals into the electrosensory lobe network.
Second, the apical dendrites of efferent cells receive a massive input. One of the inputs comes from the nucleus praeminentialis that, as in gymnotiformes, exerts a sort of gain control. The other input is from parallel fibers originating from granule cells of the eminentia granularis. This input conveys skeletal-motor corollary discharge and proprioceptive information about body posture and movements (Requarth & Sawtell, 2014). However, the best studied parallel fiber input is a second type of corollary discharge in which different parallel fibers fire at different delays following the EOD motor command. This EOD command corollary discharge excites both the dendrites of the efferent cells and the dendrites of Purkinje-like interneurons known as medium ganglion cells that project onto the efferent neurons and are also affected by afferent input from the periphery. Responses of both efferent neurons and medium ganglion cells to the temporally distributed corollary discharge inputs of parallel fibers are plastic. Parallel fiber synaptic contacts that are activated in coincidence with a strong sensory input causing centrifugal dendritic spikes are depressed, but those that are activated with other timings increase. As a result of these changes in the apical inputs, the responses of efferent cells to the electrosensory stimuli are also changed. In this way, the efferent cells become driven by a central expectation that reproduces as a negative image the average past responses of the efferent cells. This negative image cancels out the expected input and enhances the response to novel features in the incoming afferent input patterns (Fig. 7). The functional role of dendritic spikes in the control of synaptic weights through anti-Hebbian time-dependent plasticity was first described in these synaptic contacts at the electrosensory lobe of pulse mormyriformes (Bell et al., 1993, 1997).
Special Features of Gymnotiformes
In the genus Apteronotus, afferent adaptation dynamics (Section 4.1.1) cause signals evoked by the presence of objects to be encoded by modulations in the spiking rate of individual afferents, but amplitude modulations caused by fish interactions are encoded by the synchronous discharge in subpopulations of afferents distributed over broad regions of the skin. Therefore, stimulus and afferent dynamics cause a partition between those in which the population of afferent spikes is synchronized by conspecific generated beats and those in which the probability of firing still follows stimulus amplitude but fire asynchronously across the population.
Afferent projection patterns and intrinsic properties of the neurons enable a fish to extract different behaviorally relevant features from the same spike firing temporal patterns reaching the different maps of the electrosensory lobe.
These maps are distinguished by differences in the size of the receptive field and in the temporal integration dynamics. Receptive fields of the second-order neurons are different between maps. They increase 25 times in size and 5 times in the degree of overlap, in going from the centromedial to the lateral maps. In addition, the cells of the lateral map show a higher firing threshold and stronger after-hyperpolarizing potentials than those of the centromedial map. Therefore, lateral map cells are best suited to respond to the massive periodic excitation provided by the synchronized afferent input driven during conspecific interaction, and the output cells of the centromedial segment have lower spike thresholds and weaker after-hyperpolarization, keeping them best suited to respond to nonsynchronized afferent activity evoked by low-frequency communication-associated and object-associated signals.
The efferent neurons of the lobe have different roles depending on their location at different layers. Deep cells receive minimal feedback input, have no antagonistic surround, and are relatively linear encoders of the modulation in amplitude of the self-generated EOD. These cells project onto the praeminentialis nucleus, which in turn projects back to the apical dendrites of the apical and intermediate efferent neurons and also on some interneurons projecting locally. Superficial and intermediate cells have center surround fields and nonlinear responses to the afferent input owing to the convergence on their apical dendrites of the praeminentialis electrosensory feedback and the proprioceptive and central afference (Bastian et al., 2004). This circuit shows a working memory similar to that observed in mormyrids.
Slow-frequency adaptation allows the primary afferent of the slow path to encode the distance of a given object from the skin surface regardless of its approaching or receding speed and speed as the rate temporal derivative. However, the electrosensory lobe neurons do no show a symmetric pattern of response during looming and receding. These neurons better encode the temporal derivative of contrast than the center surround contrast. Therefore, E and I cells do not have strict mirror responses, but considered together, they cooperatively produce a firing rate representation of a moving object’s position that is symmetric under reversal of motion direction. It is thought that downstream neurons at the torus are able to merge both signals.
Although much less is known about central sensory processing in pulse gymnotiformes, the patterned activation of the network of the electrosensory lobe causes a frame-to-frame processing of electric images, as in pulse mormyrids (sharing the EOD regime) in a similar structure as shown by wave gymnotiformes (closely related from a genetic point of view). In fact, pulse gymnotiformes show a remarkable ability to detect novelty, requiring a working memory that is probably implemented by receptor adaptation, recurrent feedback, and plastic mechanisms, as those described earlier in this article for pulse mormyriformes (Fig. 8).
Cognitive Functions and the Potential Role of the Pallium
Recent studies of the telencephalon of gymnotiform fish have opened a new field of research to which electric fish can contribute. The forebrains of weakly electric fish are relatively large and well differentiated, suggesting that it might be the neural substrate required for learning and executing complex tasks. Lesion experiments suggest that memory storage of spatial location of landmarks occurs in the pallium. Interestingly, the dorso pallium of gymnotiformes contains cells that specifically respond to novel electrosensory stimuli, with sustained discharge for many seconds. This finding and the immediate early gene expression associated with learning suggests that these transient “up states” initiate consolidation of short-term memories (Harvey-Girard et al., 2011).
Although this is a new field of research, these findings match the strong behavioral evidence that electric fish are able to perform cognitive tasks that are considered to be mediated by telencephalon in other vertebrates.
For example, the mormyrid Gnathonemus petersii is able to coordinate its movements with increases in EOD sampling during the exploration of objects. The calculation of the electric flow during natural probing behavior in Gnathonemus petersii (Mormyridae) provides evidence for intentional fish movements to optimize sensory patterns received on a sensory surface when evaluating an object or the environment (Nelson & MacIver, 1999; Hofmann et al., 2013).
Strong evidence has also been presented that electric fish are able to learn object attributes based on electrosensory cues as, for example, of a sort “electric color,” “shape,” and “size” of objects and can use them for decision making (von der Emde, 2006). In addition, G. petersii and S. macrurus can recognize three-dimensional orientations and configurations and extract and generalize spatial features solely with their electrical sense. Strikingly, G. petersii is able to learn object shapes using the electric sense, to transfer such internal representation into visual expectations, and to visually recognize without previous visual exposition the shape learned using the electric sense (Schumacher & von der Emde, 2016).
The electrosensory world of weakly electric fish is constructed inside a fuzzy-bounded electrosensory bubble that moves with the fish. This bubble can be divided in three major concentric regions distinguished by the nature of signals and the rules determining how spatial resolution and contrast of images decay with object distance. Electroreceptor organs are distributed over a complex sensory organ, the skin of the fish. This mosaic is able to encode the timing, amplitude, time course, and spatial pattern of the EOD. The active electric sense is integrated with two other cutaneous senses (passive electroreception and lateral line mechanoception) and also with proprioceptive signals informing about body posture. These signals, as well as electromotor and skeleton motor corollary discharges present only in pulse mormyrids, serve to build expectation signals that are subtracted from the incoming inputs at the earliest sensory relays. Sparse encoded signals at the mesencephalon project back to lower brainstem regions, initiating automatic behaviors that improve electroreception as, for example, the jamming avoidance and novelty responses. They are also carried up through the preglomerular nucleus to the telencephalon, where they serve important cognitive tasks such as perception, decision making, and learning.
Journal of Physiology-Paris (1979). SPECIAL ISSUE, 75(4).Find this resource:
Journal of Comparative Physiology A (1993). SPECIAL ISSUE. Guest editors: C. Bell, K. Grant, and C. Hopkins, 172(4).Find this resource:
Journal of Experimental Biology (1999). SPECIAL ISSUE: Electroreception and Electrocommunication. Guest editors: Ray Turner and Len Maler., 202(10).Find this resource:
Journal of Physiology-Paris (2002). SPECIAL ISSUE. Guest editors: G. von der Emde, K. Grant, and J. Meek, 96(5).Find this resource:
Journal of Comparative Physiology A (2006). SPECIAL ISSUE. Ed: G. Zupanc. 192(6).Find this resource:
Journal of Physiology-Paris (2008). SPECIAL ISSUE: Electrosensory Systems. Guest editors: G. von der Emde, K. Grant, J. Engelmann, and B. Carlson, 102(4–6).Find this resource:
Journal of Experimental Biology (2013). SPECIAL ISSUE: Electric fishes: neural systems, behaviour and evolution. Guest editors: Rüdiger Krahe and Eric Fortune, 216(13).Find this resource:
Journal of Physiology-Paris (2016). SPECIAL ISSUE. Guest Editors: L. Maler, M. E. Castelló, and L. Quintana.Find this resource:
Bioinspiration and Biomimetics (2016). SPECIAL ISSUE. Guest editor: F Boyer.Find this resource:
Amey-Özel, M., Von Der Emde, G., Engelmann, J., & Grant, K. (2015). More a finger than a nose: The trigeminal motor and sensory innervation of the Schnauzenorgan in the elephant‐nose Fish Gnathonemus petersii. Journal of Comparative Neurology, 523(5), 769–789.Find this resource:
Bastian, J., Chacron, M. J., & Maler, L. (2004). Plastic and nonplastic pyramidal cells perform unique roles in a network capable of adaptive redundancy reduction. Neuron, 41(5), 767–779.Find this resource:
Bell, C. C. (1990). Mormyromast electroreceptor organs and their afferent fibers in mormyrid fish. III. Physiological differences between two morphological types of fibers. Journal of Neurophysiology, 63(2), 319–332.Find this resource:
Bell, C. C., Caputi, A., & Grant, K. (1997). Physiology and plasticity of morphologically identified cells in the mormyrid electrosensory lobe. Journal of Neuroscience, 17(16), 6409–6423Find this resource:
Bell, C. C., Caputi, A., Grant, K., & Serrier, J. (1993). Storage of a sensory pattern by anti-Hebbian synaptic plasticity in an electric fish. Proceedings of the National Academy of Sciences, 90(10), 4650–4654.Find this resource:
Bennett, M. V. L. (1971). Electric organs. Fish Physiology, 5, 347–491.Find this resource:
Berman, N. J., & Maler, L. (1999). Neural architecture of the electrosensory lateral line lobe: adaptations for coincidence detection, a sensory searchlight and frequency-dependent adaptive filtering. Journal of Experimental Biology, 202(10), 1243–1253.Find this resource:
Bullock, T. H., Hagiwara, S., Kusano, K., & Negishi K. (1961). Evidence for a category of electroreceptors in the lateral line of gymnotid fishes. Science, 134(3488), 1425–1437Find this resource:
Bullock, T. H., & Heiligenberg, W. (1986). Electroreception (p. 711). Hoboken, NJ: John Wiley & Sons.Find this resource:
Bullock, T. H., Hopkins, C. D., Popper, A. N., & Fay, R. R. (Eds.). (2005). Electroreception. New York: Springer Science+Business Media.Find this resource:
Caputi, A. A. (1999). The electric organ discharge of pulse gymnotiforms: the transformation of a simple impulse into a complex spatio-temporal electromotor pattern. Journal of Experimental Biology, 202, 1229–1241.Find this resource:
Caputi, A. A. (2011). Detection and generation of electric signals—Electric organs. In Anthony P. Farrell (Ed.), Encyclopedia of fish physiology—From genome to environment (pp. 387–397). Amsterdam: Elsevier and Associated Press.Find this resource:
Caputi, A. A., Aguilera, P. A., & Castelló, M. E. (2003). Probability and amplitude of novelty responses as a function of the change in contrast of the reafferent image in G. carapo. Journal of Experimental Biology, 206(6), 999–1010.Find this resource:
Caputi, A. A., Aguilera, P. A., Pereira, A. C., & Rodríguez-Cattáneo, A. (2013). On the haptic nature of the active electric sense of fish. Brain Research, 1536, 27–43.Find this resource:
Castello, M. E., Aguilera, P. A., Trujillo-Cenoz, O., & Caputi, A. A. (2000). Electroreception in Gymnotus carapo: Pre-receptor processing and the distribution of electroreceptor types. Journal of Experimental Biology, 203(21), 3279–3287.Find this resource:
Carr, C. E. (1986). Time coding in electric fish and barn owls. Brain, behavior and evolution, 28(1–3), 122–133.Find this resource:
Chacron, M. J., Longtin, A., & Maler, L. (2011). Efficient computation via sparse coding in electrosensory neural networks. Current Opinion in Neurobiology, 21(5), 752–760.Find this resource:
Clarke, S. E., Longtin, A., & Maler, L. (2015). Contrast coding in the electrosensory system: parallels with visual computation. Nature Reviews Neuroscience, 6, 733–744.Find this resource:
Darwin C. (1872). El origen de lase species pormedio de la selección natural. Editorial Grijalbo S.S., 1957 Mexico, DF.Find this resource:
Fessard A. (1974). Electroreceptors and other specialized receptors in lower vertrebrates. Handbook of Sensory Physiology. Berlin: Springer.Find this resource:
Fessard, A., & Szabo, T. (1961). Mise en évidence dun récepteur sensible à lélectricitédans la peau des mormyres. Comptes redus hebdomadaires des séances de l’academie des sciences, 253(17), 1859.Find this resource:
Franz, V. (1912). Das Mormyridenhirn. Zool. Jb. (Abt. 2), 32, 465–492.Find this resource:
Günther, ACLG (1864). Description of a new species of Mormyrus. Proceedings of the General Meetings for Scientific Business of the Zoological Society of London, January 12, 1864, Page 22, Plate II.Find this resource:
Harvey-Girard, E., Tweedle, J., Ironstone, J., Cuddy, M., Ellis, W., & Maler, L. (2011). Long-term recognition memory of individual conspecifics is associated with telencephalic expression of Egr-1 in the electric fish Apteronotusleptorhynchus. Journal of Comparative Neurology, 518, 2666–2692.Find this resource:
Heiligenberg, W. (1977). Principles of electrolocation and jamming avoidance in electric fish: A neuroethological approach (Vol. 1). New York: Springer Science & Business Media.Find this resource:
Heiligenberg, W. (1991). Neural nets in electric fish. Cambridge, MA: MIT Press, p. 179.Find this resource:
Hofmann, V., Sanguinetti-Scheck, J. I., Künzel, S., Geurten, B., Gómez-Sena, L., & Engelmann, J. (2013). Sensory flow shaped by active sensing: Sensorimotor strategies in electric fish. Journal of Experimental Biology, 216(13), 2487–2500.Find this resource:
Hopkins, C. D. (1976). Stimulus filtering and electroreception: tuberous electroreceptors in three species of gymnotoid fish. Journal of Comparative Physiology, 111(2), 171–207.Find this resource:
Hopkins, C. D., & Bass, A. H. (1981). Temporal coding of species recognition signals in an electric fish. Science, 212, 85–87.Find this resource:
Jeffress, L. A. (1948). A place theory of sound localization. Journal of Comparative Physiology and Psychology, 41, 35–39.Find this resource:
Kalmijn, A. J. (1971). The electric sense of sharks and rays. Journal of Experimental Biology, 55(2), 371–383.Find this resource:
Kawasaki, M. (1993). Independently evolved jamming avoidance responses employ identical computational algorithms: a behavioral study of the African electric fish, Gymnarchus niloticus. Journal of Comparative Physiology A, 173(1), 9–22.Find this resource:
Krahe, R., & Maler, L. (2014). Neural maps in the electrosensory system of weakly electric fish. Current Opinion in Neurobiology, 24, 13–21.Find this resource:
Kramer, B. (1990). Electrocommunication in teleost fishes: Behavior and experiments (Vol. 29). New York: Springer Science & Business Media.Find this resource:
Lissmann, H. W. (1951). Continuous electric signals from the tail of a fish, Gymnarchus niloticus. Nature, 167, 201–202.Find this resource:
Lissmann, H. W. (1958). On the function and evolution of electric organs in fish. Journal of Experimental Biology, 35(1), 156–191.Find this resource:
Lissmann, H. W., & Machin, K. E. (1958). The mechanism of object location in Gymnarchus niloticus and similar fish. Journal of Experimental Biology, 35(2), 451–486.Find this resource:
Marsat, G., Longtin, A., & Maler, L. (2012). Cellular and circuit properties supporting different sensory coding strategies in electric fish and other systems. Current Opinion in Neurobiology, 22(4), 686–692.Find this resource:
Meyer, J. H., & Zakon, H. H. (1982). Androgens alter the tuning of electroreceptors. Science, 217(4560), 635–637.Find this resource:
Migliaro, A., Caputi, A. A., & Budelli, R. (2005). Theoretical analysis of pre-receptor image conditioning in weakly electric fish. PLoS Computational Biology, 1(2), e16.Find this resource:
Moller, P. (1995). Electric fishes: History and behavior. New York: Chapman & Hall.Find this resource:
Murray, R. W. (1962). The response of the ampullae of Lorenzini of elasmobranchs to electrical stimulation. Journal of Experimental Biology, 39(1), 119–128.Find this resource:
Nelson, M. E., & Maciver, M. A. (1999). Prey capture in the weakly electric fish Apteronotus albifrons: Sensory acquisition strategies and electrosensory consequences. Journal of Experimental Biology, 202(10), 1195–1203.Find this resource:
Nogueira, J., & Caputi, A. A. (2013). From the intrinsic properties to the functional role of a neuron phenotype: an example from electric fish during signal trade-off. The Journal of experimental biology, 216, 2380–2392.Find this resource:
Pereira, A. C., & Caputi, A. A. (2010). Imaging in electrosensory systems. Interdisciplinary Sciences: Computational Life Sciences, 2(4), 291–307.Find this resource:
Requarth, T., & Sawtell, N. B.(2014). Plastic corollary discharge predicts sensory consequences of movements in a cerebellum-like circuit. Neuron, 82, 896–907.Find this resource:
Sawtell, N. B., & Bell, C. C. (2008). Adaptive processing in electrosensory systems: Links to cerebellar plasticity and learning. Journal of Physiology-Paris, 102(4), 223–232.Find this resource:
Schugardt, C., & Kirschbaum, F. (1998). Sozial- und Fortpflanzungsverhalten von Mormyriden (Nilhechten). In H. Greven & R. Riehl (Eds.), Verhalten von Aquarienfischen (pp. 87–98). Bornheim: Birgit SchmettkampVerlag.Find this resource:
Schumacher, S., de Perera, T. B., Thenert, J., & von der Emde, G. (2016). Cross-modal object recognition and dynamic weighting of sensory inputs in a fish. Proceedings of the National Academy of Sciences, 113(27), 7638–7643.Find this resource:
Shackelford, S. (1941). Electric eel calling: A record of an artist’s association with a scientific expedition. New York: Charles Schribner’s Sons.Find this resource:
von der Emde, G. (2006). Non-visual environmental imaging and object detection through active electrolocation in weakly electric fish. Journal of Comparative Physiology A, 192(6), 601–612.Find this resource:
Watson, D., & Bastian, J. (1979). Frequency response characteristics of electroreceptors in the weakly electric fish, Gymnotus carapo. Journal of Comparative Physiology, 134(3), 191–202.Find this resource:
Wright, P. G. (1958). An electrical receptor in fishes. Nature, 181, 64–65.Find this resource:
Walsh, J., & Seignette, S. (1772). Of the Electric Property of the Torpedo. Philosophical Transactions of the Royal Society of London, 63, 461–480.Find this resource:
Xu-Friedman, M. A., & Hopkins, C. D. (1999). Central mechanisms of temporal analysis in the “Knollenorgan” pathway of mormyrid electric fish. Journal of Experimental Biology, 202(10), 1311–1318.Find this resource: