Living life with an electric touch

That some species of fish can produce electric fields (see Glossary) has long been known; for example, the strong electric organ discharges (EODs) of Nile catfish and Mediterranean torpedoes were known in ancient Egypt and classical Greece, respectively (Wu, 1984; Moller, 1995). Later, European colonization of America allowed old-world scientists to discover electric eels, whose modular electric organ (EO) structure inspired the first electric battery (Volta, 1800). At that time, it was already known that electric eels were able to detect a short circuit between a pair of wires inside their tank (Walsh, 1773). The weak EODs of the broadly distributed family Rajidae (Elasmobrachii), the superfamily Mormyroidea (Teleostei) from Africa and the order Gymnotiformes (Teleostei) from America (Fig. 1) puzzled Darwin, who wrote: ‘The electric organs of fishes offer another case of special difficulty; for it is 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). A possible communicative function was subsequently conjectured by Franz (1912, cited by Moller, 1995): ‘[a] question for future research would be to consider whether perhaps the mormyrids themselves appreciate the electric shocks of members of their species…’.

The subsequent development of electrophysiological techniques revealed that EODs result from the activity of specialized electrogenic units called ‘electrocytes’ (Albe Fessard and Buser, 1950; Keynes and Martins Ferreira, 1953; Bennett and Grundfest, 1959, 1961a,b; Bennett et al., 1961; Chagas and Paes de Carvalho, 1961). The EODs of several hundred species belonging to Gymnotiformes, Mormyroidea and Rajidae are species specific and consist of either continuous waves or series of discrete stereotyped pulses (Movie 1). Wave and pulse species are found in both Gymnotiformes (Coates et al., 1954) and Mormyroidea (Lissmann, 1951), indicating that these strategies evolved convergently in America and Africa.

Concomitantly, Hans Lissmann (1951) observed that Gymnarchus niloticus (the sole African wave species) was able to dodge obstacles while swimming in reverse (Movie 1), and proposed that weakly electric fish use their electric field as a carrier of sensory signals. This proposal was followed by a collection of inspiring articles in Journal of Experimental Biology (JEB). Together, these articles show that the signals resulting from the polarization of objects by self-generated electric fields inform a fish about object characteristics and location, and that the electric fields may also serve as communication signals (Lissmann, 1958; Lissmann and Machin, 1958; Machin and Lissmann, 1960). As a corollary, the significance of these signals for mate selection and for evolution was recognized – insight supported by the discovery of electroreceptor organs specifically tuned to the species-specific time course of the electric field (Wright, 1958; Fessard and Szabo, 1961; Bullock et al., 1961; Murray, 1962).

In the spirit of the JEB centenary, this Review re-visits old concepts in the search for new ideas for future research on EODs and their functions. This article is not intended to be a comprehensive review, but instead presents a fresh view of the mechanisms of electrogenesis and the electrosensory function of EODs in weakly electric fishes, while aiming to identify and prioritize unsolved questions and to stimulate the progress of this young but productive field.

Complex behaviors result from the combination of stereotyped actions organized by relatively simple neural circuits. These modular acts, so-called ‘fixed motor patterns’, are ‘ready-made “motor tapes” as it were, that when switched on produce well-defined and coordinated movements: the escape response, walking, swallowing, the pre-wired aspects of bird songs...’ (Llinás, 2002). Extending this concept, it was proposed that EODs are fixed action patterns in which the output is electrical rather than mechanical (Rodríguez-Cattáneo et al., 2008).

In motor systems, relatively large forces result from the sum of weak forces generated by muscle fiber contraction. The elasticity and viscosity of passive muscle elements allow them to integrate and filter the forces generated and apply them to the skeleton (Hoffer et al., 1992), resulting in a mechanical output that is a balance between force and displacement determined by the load moved (Hoffer et al., 1989). Similarly, in electromotor systems, thousands of identically oriented electrocytes stacked in a connective tissue sleeve form an EO. Each electrocyte contributes a small fraction of the total current that is conducted along the EO. In the absence of an external conductive medium (such as tank water), the potential of the electrogenic system to generate external currents is represented by the electromotive force, which corresponds to the drop in voltage caused by the circulation of the electrocyte-generated current along the fish's body in the opposite direction to the flow of current inside the EO (Cox and Coates, 1938; Bell et al., 1976; Caputi et al., 1989, 1993; Caputi and Budelli, 1995). This is analogous to the isometric contraction of a muscle group, which reflects the potential to produce movement.

When water offers an external path, a fraction of the EO-generated current leaks through the skin: the source becomes distributed and can be modeled as an array of electromotive forces and electric impedance pairs (see Glossary; Caputi and Budelli, 1995). Swimming in a heterogeneous conductive medium produces a self-stimulating close field, resulting from the trade-off between the changing distribution of impedance of the surrounding medium and the fixed distribution of impedance of the fish's body (Caputi et al., 1998a,b, 2002; Caputi and Budelli, 2006). Beyond a few body lengths, the shape of the field approximates that of a dipole (see Glossary) with a species-specific time course (Aguilera et al., 2001). Changes in the time course and spatial distribution of these electric fields are specifically detected by a mosaic of cutaneous electrosensory receptor organs distributed over the body. This produces a species-specific information channel – with high signal-to-noise ratio – that is tuned for receiving and processing the electric images (see Glossary) carried by self-generated fields and those of conspecifics.

Just as bones and joints are essential for transmitting muscle actions, so the conductivity of the body links the EO output with the water, and the control of the fish's body movements is used to shape the self- and allo-generated fields (Fig. 2). Thus, the body of the moving fish can be considered as a self-controlled filtering device used to: (1) orient the field in order to polarize objects at the same time as orienting the electrosensory receptors to receive a self-generated ‘electric image’ of the nearby environment, and (2) elicit and receive allo-generated images. Consequently, one can decompose the electromotor act into two inextricable stages: (1) electrogenesis, defined as the action of transforming the fish's body into an electric source generating a stereotyped electromotive force pattern, and (2) imaging, which is the circumstantial result of the action depending on the impedance of the surroundings.

Electrogenesis by weakly electric fish involves precise coordination of a range of spatio-temporal processes at multiple levels of organization, from the sub-cellular level within the electrocytes to the whole-organism level, where the activities of multiple electrocytes are added. The electrocytes often show different excitable properties in the same species, and the electric fields produced are of different shapes and sizes. The currents generated by the electrocytes flow along the EO and close the circuit through the rest of fish's tissues and water. Thus, each deflection (see Glossary) of the EOD and the associated electric field results from the sum of currents generated at similarly oriented patches of membrane (Bennett and Grundfest, 1959). Some EOD deflections originate in the synaptic activation of a patch of membrane. Other deflections result from the propagation of a synaptically evoked action potential along the membrane until it reaches the opposite side of the electrocyte. Finally, in most species, the local sum of the ‘action currents’ (see Glossary) generated at the caudal faces of small and densely packed neighboring electrocytes of the tail region stimulates the rostral membrane faces of the same and adjacent electrocytes (referred to as ephaptic coupling, see Glossary; Bell et al., 1976; Fig. 3A).

The EOs of the different electrogenic fish families are analogous structures serving a common function. EOs appeared several times in evolution and consequently show differences in their lineage, organization, location, innervation pattern and control mechanisms, reflecting their different developmental origins. Nonetheless, they show commonalities that exemplify how evolution converges on similar solutions to solve common problems. For example, in pulse Mormyroidea and Rajidae, the EOs are short and localized at the tail region, whereas in the single species of wave Mormyroidea (Gymnarchus niloticus) and Gymnotiformes, the EOs extend across most of the length of the body. The EOs of Apteronotidae consist of a complex plexus composed of modified axons of spinal electromotor neurons (Waxman et al., 1972) and exhibit neurogenic electrocytes (i.e. electrocytes that are derived from neural tissue) at the terminal portion of such axons. The EOs of the rest of the weakly electric fishes consist of connective tissue sleeves inside which syncytial excitable units of mesodermic or muscular lineage (myogenic electrocytes) are connected in series by a gelatinous connective tissue (Szabo, 1961; Trujillo-Cenóz and Echagüe, 1989; Gallant, 2019).

All electrocytes, whether neurogenic or myogenic, possess extensive excitable membranes, species-specific non-linear voltage-dependent channel repertoires and specifically located activation sites. Thus, while the contribution of each electrocyte to the EOD depends on its shape, size, channel repertoire and activation mechanisms, the nature of the whole EOD depends on the diversity of electrocyte types, their distribution along the EO and the neural coordination mechanisms exhibited by each species (Box 1).

The repertoire and distribution of voltage channels in the electrocytes of different species are described more extensively elsewhere (Bennett, 1971; Sierra et al., 2005, 2007; Markham, 2013, 2019). Genome and transcriptome studies promise to rapidly enhance our knowledge of the genetic and evolutionary basis of electrocyte membrane properties (Zakon et al., 2006; Zakon, 2012; Gallant et al., 2014; Gallant, 2019). For example, one striking relevant discovery is the neofunctionalization of a voltage-gated Na⁺ channel gene (Zakon, 2012). In separate fish lineages, a duplicate originating from a muscle-type Na⁺ channel has been co-opted twice by novel myogenic EOs and once by a neurogenic EO (Zakon et al., 2006; Arnegard et al., 2010). In addition, advances in our understanding of the genetics of K⁺ channels and the regulation of the co-variation of the Na⁺ and K⁺ electrocyte channel dynamics are contributing to our understanding of the shape and duration of electrocyte action potentials (McAnelly and Zakon, 2000; Swapna et al., 2018).

In pulse Mormyridae and the wave Gymnotiformes of the family Apteronotidae, electrocyte shape is a major factor determining their contribution to the EOD. The short EO of Mormyridae consists of convoluted, topologically non-spherical (see Glossary) electrocytes that are tightly stacked and identically oriented. The propagation of the activation along the complex membrane structure determines the characteristics of the species-specific EOD (Alves-Gomes and Hopkins, 1997; Hopkins, 1999; Sullivan et al., 2000; Gallant, 2019) (Box 1, panels A and B). Apteronotidae electrocytes (Box 1, panel C) consist of specializations of the terminal axon in which two 100 μm diameter swellings are separated by a thin ‘hairpin’ turn. Nodes of Ranvier (see Glossary) at the end of each swelling are less excitable (Box 1, panel C, top, green bands), resulting in decremental conduction of the action potential. The currents leaving through these passive nodes (Box 1, panel C, red and blue arrows) generate the head-positive and -negative deflections of the EOD waveform shown below (Box 1, panel C, red and blue waves; Bennett, 1971; Waxman et al., 1972).

In the remaining species, electrocytes are topologically spherical (see Glossary) and the location and timing of their activation by their synaptic contacts determine the time course of their contribution to the EOD. In wave species (including Sternopygidae, which are Gymnotiformes), Gymnarchus niloticus, some species of Gymnotidae (Gymnotus cylindricus, Gymnotus obscurus) and most Hypopomidae, electrocytes are innervated on their caudal face only (Bennett, 1971). The ways in which the rostral faces of singly innervated electrocytes are recruited is species dependent and diverse. The probability and timing of rostral face recruitment depends on membrane excitability (Bennett, 1971; Markham, 2013, 2019) and also on the excitatory effects of ephaptic currents generated by neighboring electrocytes (i.e. EO autoexcitability; Bell et al., 1976; Caputi et al., 1998b; Rodríguez-Cattáneo and Caputi, 2009; Rodríguez-Cattáneo et al., 2013).

During the process of EO auto-excitability, action currents synchronously elicited at the caudal electrocyte face are ‘leakily’ funneled along the EO and close the circuit through the surrounding tissue. A fraction of the action currents generated by each electrocyte are conducted along the EO. The conduction distance increases with the ratio between the sleeve and tissue resistance (i.e. the space constant of the sleeve; Fig. 3A). The sum of currents from neighboring electrocytes elicits a phase-locked activation of the opposite, non-innervated faces (Albe-Fessard and Buser, 1950). Thus, EO autoexcitability increases with electrocyte density. In some species, auto-excitability at the tail region is large enough to cause repeated activation of the rostral and caudal faces, thus generating multiple waves of activity (Rodríguez-Cattáneo et al., 2013; Waddell et al., 2016). Seasonal increases in water conductivity offer a parallel path for closing the circuit of ephaptic currents. Moreover, seasonal and sex-dependent modulation of electrocyte excitability facilitates the recruitment of rostral electrocyte faces in many pulse species of Gymnotiformes (Fig. 3B,C).

In most Gymnotidae and in Rhamphychtidae, electrocytes can be innervated on the rostral face, at the caudal face or on both faces. The coordinated activation of the synaptic activities of electrocytes with different innervation patterns allows them to contribute to the whole EOD with a variety of waveforms. Additionally, in these species, different patterns of electrocyte innervation occur at different regions of the EO; their coordinated activity gives rise to very complex EODs (Caputi, 1999; Rodríguez-Cattáneo et al., 2013; Waddell et al., 2016).

Finally it should be mentioned that some Hypopomidae species – for example, Steatogenis elecgas (Hypopomidae; Bennett, 1971) – show a second EO at the head region. Furthermore, a whole clade of small-sized Gymnotus, including G. javary and G. coropinae, show a head-specific expansion of the EO (Castelló et al., 2009; Rodríguez-Cattáneo et al., 2013).

In almost all species of weakly electric fish, inputs of the spinal electromotor neurons controlling the EO originate from a command nucleus, the output of which is followed one-to-one by the EOD. The command nucleus stimulates a medullary relay nucleus whose axons project to spinal electromotor neurons, which in turn activate the electrocytes (Bennett, 1971; Dye and Meyer, 1986; Bass, 1986; Caputi et al., 2005; Caputi, 2011).

Mormyridae and Rajidae emit electric pulses at irregular intervals that depend on the integration of descending signals from the command nuclei (Bennett et al., 1967; Grant et al., 1999). These species have a compact EO composed of very similar electrocytes located at a similar conduction distance from the command nucleus; this facilitates the synchronous activation of all the electrocytes.

In Apteronotidae, small rhythm-setting and large relay neurons are intermingled within the premotor command nucleus (Elekes and Szabo, 1985; Moortgat et al., 2000). In apteronotids, all neurons of the electromotor system are also able to spontaneously fire action potentials at a regular frequency (i.e. they display pacemaker properties). Bulbar command neurons fire at the highest frequency – driving the EOD above 2000 Hz in some species – which is the highest frequency of neural firing observed in nature. Relay neurons send thick, fast-conducting fibers that phase lock the intrinsic rhythmic firing of electromotor neurons, whose axonal terminal specializations function as electrocytes (Bennett, 1971; Smith, 2013). The successive activation of the arms of the hairpin-shaped electrocytes (Box 1, panel C) determines the biphasic discharge whose duration matches the firing interval of the command pacemaker nucleus (Waxman et al., 1972).

In the remaining species, including the families Gymnarchidae, Sternopygydae and all pulse Gymnotiformes, the EOs stretch over 90% of the length of the body; consequently, the distance from the electrocytes to the pacemaker nucleus increases along the body. To produce a stereotyped EOD output waveform, the coordinated activity of about a thousand similarly oriented electrocytes in these species is required. However, the electrocytes are located at varying distances (up to 30 cm) from the pacemaker. The mechanism by which coordination is achieved has been studied in Gymnotus omarorum and Brachyhypopomus gauderio. Synchrony results from the combination of three ‘delay-line’ mechanisms (see Glossary): (1) differential conduction velocities increase the conduction time of the shorter fibers of the bulbo-spinal tract (Lorenzo et al., 1990, 1993); (2) the farthest electromotor neurons are activated with shorter latency by the convergence of more synchronized inputs (Caputi and Trujillo-Cenóz, 1994; Caputi and Aguilera, 1996); and (3) differential conduction lengths increase the conduction time of electromotor neuron axons projecting to more rostral electrocytes (Bennett, 1971; Trujillo-Cenóz et al., 1984; Caputi and Trujillo-Cenóz, 1994). This results in a traveling wave sweeping the body at 400 m s⁻¹ (Caputi, 1999). Despite this speed, the delay between the activities of equally innervated faces of the electrocytes at the abdominal and caudal portions of the EO is a large fraction of the duration of an action potential. This delay increases the complexity of the electric field in pulse Gymnotiformes (Box 1, panels H–J) and in some wave emitters belonging to Sternopygydae and Gymnarchidae, where a phase opposition between the EOD components generated at the more rostral and more caudal regions of the fish's body may be observed.

In addition, in most species of Rhamphychthydae and Gymnotidae, the EOs are composed of various types of electrocytes (Box 1, panels D–F), differing in their innervation pattern, size and membrane properties. In these species, electrocytes of different innervation types (which consequently contribute to the EOD with currents of different time courses) are contained in different tubes in the same body region. Each electrocyte contributes to the EOD with either a monophasic discharge, corresponding to the activation of a single face, or a multiphasic discharge, corresponding to the sequential activation of differently oriented electrocyte regions. Moreover, the activity of different electrocytes is driven by cholinergic synapses that may evoke either smooth and long end plate potentials (see Glossary; Lorenzo et al., 1988; Sierra, 2007) or sharp and brief action potentials, according to their size and channel repertoires (Bennett and Grundfest, 1959; Macadar et al., 1989). In these electrocytes, current direction alternates between caudal to rostral and vice versa owing to a precise sequence of activation of caudally and rostrally oriented faces. This is achieved by two mechanisms, depending on whether the electrocyte is doubly (Box 1, panels D and F) or singly innervated (Box 1, panels E and G; Trujillo-Cenóz and Echagüe, 1989; Macadar et al., 1989; Caputi et al., 1989).

Singly innervated electrocytes contribute to the EOD with a mono- or bi-phasic discharge, the latter of which is achieved by EO auto-excitability (Albe-Fessard and Buser, 1950; Bell et al., 1976; Caputi et al., 1998b; Rodríguez-Cattáneo and Caputi, 2009) (Fig. 3A). The probability of recruitment and synchronization of neighboring electrocytes by these ephaptic mechanisms increases with the space constant of the connective sleeve and also with the progressive decrease in electrocyte size and inter-electrocyte distance from abdominal to tail regions. This explains the increase along the EO of the final EOD wave observed in all pulse Gymnotiformes (Caputi et al., 1989, 1994, 1998a,b) (Box 1, panels H–J).

Doubly innervated electrocytes in Gymnotidae (Box 1, panel D) and Rhamphichthyidae (Box 1, panel F) contribute tri- and tetra-phasic discharges, resulting from the combination of a precise sequence of synaptic activation. In these electrocytes, the synchronized synaptic input to identically oriented faces causes the initial EOD deflection. This is followed by a phase-locked synaptic activation of opposite faces and finally by the successive activation of opposite faces in a ‘reverberating’ way, implemented by EO auto-excitability.

In pulse Gymnotiformes, different regions of the EO contribute with waveforms of different time course to the whole EOD. In Brachyhypopomus gauderio, the different regional contributions mainly depend on electrocyte size and density and their coordination mechanisms. The red band in Box 1, panel H, represents the location of the caudally singly innervated electrocytes shown in Box 1, panel G, which are present all along the EO. In Gymnotus omarorum and Rhamphychtis rostratum, there are both doubly innervated electrocytes (Box 1, panels D and F; blue bands in panels I and J) and singly innervated electrocytes (caudally innervated electrocytes are located along the red band in Box 1, panels I and J; rostrally innervated electrocytes are located along the green band in panel J), which are specifically located at different regions of the EO and contribute with different waveforms to the whole EOD (Caputi et al., 1989, 1994, 1998a,b). The present hypothesis regarding control of the EOD time course in Gymnotus omarorum is depicted in Box 2.

EODs of weakly electric fish are key for three functions. In Mormyroidea and Gymnotiformes, (1) they carry signals informing the fish about nearby objects, (2) they are signs recognized by other fish and (3) they are symbols of a communication system. Each of these functions is discussed in more detail below. In Rajidae, EODs are emitted rarely in the absence of stimuli generated by other fish (Kramer, 1996). In these species, it is thought that the main role of EODs is to convey sexual calls (Sisneros and Tricas, 2002).

Analogous to light in vision, the EOD-associated field is the signal carrier that stimulates the electroreceptive mosaic of electric fish. However, unlike the visual sense, a self-generated pattern of transcutaneous current density is generated independently of the presence of any object (see discussion on different active systems in Box 3). When objects are present, the self-generated field causes a charge redistribution in the objects according to the difference in permittivity (see Glossary) between the object and the water (Lissmann and Machin, 1958; Caputi and Budelli, 2006; Nelson and MacIver, 2006; Benda, 2020). Polarized objects behave as virtual electric sources, so-called ‘stamps of the objects in the electric field’, just as the sun-illuminated moon behaves as a luminous object (Caputi and Budelli, 2006). For simple objects relatively far from the body, such a virtual source can be represented by two components: the local polarizing field and the object's ‘imprimence’ (Lissmann and Machin, 1958). The EOD field provides an ego-centered perceptual coordinate system that moves with the fish, relaying upon the moving body the information about the object's distance, altitude and azimuth. The imprimence is a summary of the object's characteristics, including its material (i.e. impedance relative to the water, which causes changes in the stimulus time course) and geometry (i.e. size and shape; Rasnow, 1994, 1996; Pereira and Caputi, 2010; Boulier, 2013; Ammari et al., 2013). All this information is integrated in the object's ‘stamp’. According to this concept, the field that stimulates the skin (the ‘re-afferent signal’) can be considered as the sum of two fields: the first directly generated by the fish EOD that exists in the absence of the object, and the second indirectly generated by the fish EOD, virtually generated by the object stamp (the ‘ex-afferent signal’; von Holst and Mittlesteadt, 1969; Fig. 4). The electrosensory system of Mormyroidea and Gymnotiformes is organized to extract the ex-afferent information that represents the object and is contained in the re-afferent image projected on the skin. The geometry and location of the object appear to be encoded by the spatial pattern of the image (Rasnow, 1996; Caputi et al., 1998a,b; Budelli and Caputi, 2000) (Fig. 4), whereas the object's material appears to be encoded in the time course of local stimuli (Fig. 4 and Box 4).

To understand how electric fish evaluate an object's geometry and location, one must take into account that unlike the apposition images (see Glossary) formed in the vertebrate retina, the electrosensory mosaic receives superposition images (see Glossary). Signals originating from every point in space project with different weights onto different points on the skin and, consequently, each electroreceptor is stimulated by the weighted sum of the signals generated at different points in the surrounding space (Caputi and Budelli, 2006; Pereira and Caputi, 2010) (Fig. 4). These relative weights show large variations when the objects are close by, causing the largest increments in electric images at the region of the skin closest to the object. In addition, objects of different shapes and size located at about the same site relative to the fish's body generate images of different profiles. In this case, the stamp-generated field resembles the field of a multipole in which components project with different weight onto different regions of the skin. By contrast, when the fields of objects that are far away reach the skin, they resemble that of a single dipole: distant objects have similar image profiles varying by their location. Consequently, the stimulus image of distant objects is of low intensity, broad and of low contrast (Rasnow, 1996; Sicardi et al., 2000). Object discrimination from the background vanishes at a distance of about one-half of the fish length, but the limits of the ‘sensory bubble’ moving with the fish are not clear: this can be compared to a human seeing through the fog (Pereira et al., 2012). Furthermore, images of more than one object soon become fused, reaching the limit of ‘electric acuity’ at about one-half of the distance corresponding to the object's detection threshold (Pereira et al., 2012).

Because the net current through the skin is null (Lagrange, 1773), electrosensory images consist of regions of skin where the current density increases at the expense of other regions, where it decreases (Caputi et al., 1998a; Pereira and Caputi, 2010). Signal projection weights depend on the ratio between the distance to the object and the curvature radius of the skin. When this ratio is small, the images of nearby objects represented here have a center-surround ‘Mexican hat’ profile (i.e. with a broad, low brim and a high central peak; Fig. 4). This profile indicates the presence of a filter commonly used for border detection in image processing (Marr and Hildreth, 1980). Most animals implement this filter at the neural level using a mechanism called lateral inhibition, but in electric fish, this filtering occurs twice – firstly at the physical image-generation stage (i.e. a pre-receptor mechanism) and secondly in the early sensory centers.

On the side of the fish, the Mexican hat filtering profile is observed. However, in front of the fish, large objects globally alter the flow of currents, causing a bell-shaped increment (those more conductive than water increase current flow and those less conductive decrease it), the spatial profile of which does not depend on the shape of the object but on the shape of the snout (Fig. 4). In contrast, small objects close to the skin generate a Mexican hat profile. In this case, there is also a reciprocal polarization between the fish's body and the object. This causes a juxta-cutaneous fringe of ‘hypersensitivity’ that allows the fish to evaluate the ‘electric texture’ of object surfaces (Caputi et al., 2011, 2013). This, and the increase in current density at the perioral region due to body funneling, increases the contrast of objects at the region where electroreceptor organs have the highest density and functional variety (this is the so-called ‘electrosensory fovea’; see Glossary; Castelló et al., 2000; Bacelo et al., 2008). Placing this fovea at the large curvature regions appears to be an optimal design for evaluating whether an object will fit into the mouth (Movie 2).

Object impedance is encoded as changes in the time course of the local stimuli over the skin. In vision or audition, the frequencies of the stimulating spectra are combined in a precise way, giving origin to the idea of object ‘qualia’. For example, the same color or chord can be generated by the combination of different frequencies in a family of well-defined proportions, and the set of stimuli generating the same qualitative sensation can be considered equivalent. This provides the animal with the ability to qualitatively classify objects according to their color or the sounds according to their timbre. In Gnathonemus petersii and Gymnotus omarorum, families of time courses of a local stimulus serve to encode object impedance, providing these fish with a sort of ‘electric color’ (Budelli and Caputi, 2000). As explained in Box 4, the stimulus space of these pulse fishes can be partitioned into families of stimuli defined as those having the same relationship between the peak-to-peak amplitude and the ratio of the deflections (von der Emde and Ronacher, 1994; Aguilera and Caputi, 2003; Gottwald et al., 2017, 2018). As in other senses, encoding of qualitative aspects is based on the presence of two types of electroreceptors with differential sensitivity to changes in amplitude and waveform at the foveal and perifoveal regions (von der Emde and Bleckmann, 1992, 1997; Rodríguez-Cattáneo et al., 2017; Caputi and Aguilera, 2019). Interestingly, qualia discrimination is not only used for evaluation of object impedance. The electric color of the images generated by the EOD of other fish is also used to discriminate conspecifics from allo-specifics and male from female conspecifics (Hopkins and Bass, 1981; Waddell and Caputi, 2020) (Movie 3).

Finally, there is an important dynamic aspect to address (see also Box 3). In most species of weakly electric fish, the EO is not present in the head region; consequently, current leaves the snout and spreads out in a cone within which the time course of the polarizing field is relatively constant, and its amplitude and orientation are site specific (Caputi and Aguilera, 2020). The fish can orient this ‘foveal field’ towards objects of interest, using the snout to direct a ‘polarizing current beam’, simultaneously orienting the fovea to optimize reception of the electric image (Caputi and Aguilera, 2020) (Fig. 2). The fish can then orient its head (and, in Gnathonemus petersii, a chin appendage; Bacelo et al., 2008; Amey-Özel et al., 2015) towards objects of interest. This enhances image resolution by both increasing the object's stamp and allowing the fish to receive the stamp-generated field at the most dense and qualitatively rich area of the sensory mosaic (Caputi et al., 2013; Caputi and Aguilera, 2020). Foveal movement, exploring an object's surface (Toerring and Moller, 1984; Nelson and MacIver, 1999) (Movie 2), combined with sensory adaptation mechanisms enhancing transient changes in the re-afferent signals (Bell et al., 1993, 1997; Caputi et al., 2023), improve the evaluation of electric texture, just as fingertip stroking movements do when humans evaluate tactile texture (Caputi et al., 2013; Rodríguez-Cattáneo et al., 2020) (Movie 2). Consequently, within the ‘electrosensory bubble’ there are two electrosensory receptive fields: one a peripheral, omnidirectional field (Snyder et al., 2007) adapted for object detection and the other a foveal field adapted for close object examination (Caputi and Aguilera, 2020).

A sign is an event whose occurrence indicates the probable occurrence of something else. Just as a bark is a sign of a dog's presence and location, an allo-generated EOD waveform is a sign of the presence, identity and location of a nearby sympatric individual. However, just as a barking dog may interfere with signal perception when one is listening to music, a fish close by may interfere with active electrolocation. This trade-off between signature and interference is solved in different ways by different taxa.

Pulse Mormyridae distinguish allo-generated stimuli from self-generated input through the use of an EOD command-driven inhibitory corollary discharge that separates these stimuli to distinct processing pathways. There is a dedicated sensory path for receiving allo-generated signals (Bell and Grant, 1989); in this pathway, the corollary discharge blocks self-generated sensory inputs in the first processing relay in the brain, eliminating self-interference (Bell and Grant, 1989). By contrast, in the dedicated sensory path for active electroreception, the corollary discharge gates the signals, letting in those generated by the self-generated EOD (Bell, 1989). Pulse Gymnotiformes, which produce electric signals in a regular rhythm, show fast and transient increases in EOD frequency when the EOD of another fish precedes or coincides with the self-generated EOD, thus reducing the probability of future coincidence (Westby, 1975).

Both Mormyroidea and Gymnotiformes have evolved a fast electrosensory path that responds with a single spike to allo-generated EODs (Hopkins, 1986; Castelló et al., 2008; Carlson, 2019). The difference in latency between primary afferent spikes generated on different sides of the fish informs the fish of the orientation of the field emitted by a conspecific. These signals are analyzed in both cases by a Jeffress' circuit (see Glossary), similar to that observed in the auditory systems of birds and mammals (Carr, 1993). Consistent with this ability to detect field orientation, these fish have an electrotaxic reaction – they align their body with the field orientation, swimming along the field lines (Schluger and Hopkins, 1987; Shieh et al., 1996; Hopkins, 2005; Waddell and Caputi, 2020) (Movie 3). This side-to-side orientation of the body to the stimulus is perhaps analogous to the way in which humans orient their head toward an interlocutor during verbal communication. Experiments using constant field stimulation indicate that the side-to-side gradients caused by the disparity of the fish's body and field angle, and the gradient of the time derivative of the stimulus at the head region (which is a cue for identity and approaching side) are sufficient to evoke the electrotaxic response in pulse Gymnotiformes (Waddell and Caputi, 2021), although several other models have been proposed (Schluger and Hopkins, 1987; Hopkins, 2005) and have been tested in artificial devices (Boyer et al., 2013).

In pulse teleostei, species- and sex-specific EOD time courses let other fish know the identity of the emitter. Playback of synthetic EOD series using either random sequences or phase-locked stimuli showed the importance of EOD time course rather than the power spectral density or the timing pattern to cue species and sex recognition of a distant emitter (Hopkins and Bass, 1981; Bass and Hopkins, 1983; Waddell and Caputi, 2020). In addition, when pulse gymnotiforms are exposed simultaneously to two playback time courses, they are attracted to the source better representing the same species and the opposite sex (Waddell and Caputi, 2020, 2021).

In the case of wave fish, amplitude modulations of the EOD have a beat frequency equal to the difference between the two EOD frequencies. When fish move relative to one another, or there is more than one fish, one can observe slow and less predictable ‘envelopes’ consisting of variation in the degree of amplitude modulation (Stamper et al., 2019; Benda, 2020). Central processing of EOD beats allow Eigenmannia (a gymnotiform; Movie 1) and Gymnarchus to avoid interference from conspecifics. These species have evolved a similar mechanism for jamming avoidance: they evaluate both the amplitude and phase of the temporospatial pattern of interference to identify whether their own EOD has greater or lesser frequency than the interfering signal (Watanabe and Takeda, 1963; Bullock et al., 1972; Heiligenberg, 1991; Kawasaki, 1993). This elicits jamming avoidance and socially evoked responses that consist of shifting their own EOD frequency either up or down, which increases the beat frequency sufficiently to reduce the interference pattern. Although much is known regarding the central processing of envelopes in Apteronotidae (Metzen and Chacron, 2014, 2019), understanding the neuroethological meaning of these signals is a future challenge.

Communication involves congruent changes in the behavior of two or more individuals coordinated by messages ciphered in a common code. EOD timing patterns and self-deformation of the species-specific EOD time courses appear to be symbols of a code that is currently only partially understood. Evidence for this are the ‘rasps’ and ‘chirps’ expressed by both Mormyridae and Gymnotiformes. These signals involve explosive and brief increases in EOD repetition rate, and often occur during courtship or aggressive encounters – they are discussed in more detail below.

More than 10 stereotyped timing patterns have been identified during courtship, foraging and aggressive interactions in mormyrids (Hopkins, 1988; Carlson, 2002; Arnegard and Carlson, 2005; Gebhardt et al., 2012; Carlson and Gallant, 2013; Worm et al., 2021). Amongst these, rasps are bursts of a few pulses separated by longer intervals. During courtship, male rasps are followed by a female response and vice versa (Hopkins, 1981). This ‘rasp matching’ suggests that a dialog is occurring between males and females (Wong and Hopkins, 2007). In pulse Gymnotiformes, chirps consist of rapid increases in the pacemaker activity that often overwhelm the EOD coordination mechanisms, causing a reduction in amplitude and major changes in the time course of the field. Chirps appear to be the symbols that make up sex communication messages in Brachyhypopomus and social ranking signals in Gymnotus (Silva et al., 2007; Silva, 2002, 2019).

In wave fish, chirps consist of fast accelerations of the EOD. Chirps emitted under different circumstances vary in the onset, degree and duration of the frequency modulation. They are frequently ‘answered’ with other chirps during aggressive and sexual encounters, indicating that they are symbols of a communication code (Smith, 2013; Zupanc et al., 2006; Gama-Salgado and Zupanc, 2011). Chirps provoke fast changes in the amplitude and phase modulations of EODs that are well discriminated at the level of electroreceptors and in the central electrosensory lobe (Benda et al., 2005; Hupé et al., 2008), and there is indirect evidence that self- and allo-generated chirps are segregated by corollary discharge mechanisms at the preglomerular complex, which is the sole brainstem gate to the forebrain (Wallach et al., 2022). Wave fish also produce elaborate ‘electric serenades’ during courtship. In the presence of females, Sternopygus males display frequency increases and decreases that range from 50 to 150 Hz, and female Eigenmannia produce elaborate ‘warbling’ during courtship and spawning (Hopkins, 1988).

Although the rapid rate of knowledge acquisition in this field makes it hard to predict the most fruitful research directions, here I suggest some areas that are likely to be particularly productive. Firstly, further research into the membrane channels expressed in electrocytes is likely to be beneficial. The astonishing neural firing properties of Apteronotidae are conferred by a particular type of voltage-gated Na⁺ channel (Thompson et al., 2018); the acetylcholine receptor has also been widely studied in electrocytes (Keesey, 2005). Elucidating further details of the function of ionotropic and metabotropic membrane channels expressed in electromotor neurons and electrocytes would greatly contribute both to our understanding of the functional evolutionary adaptations of electric fish and to our knowledge of neuronal intrinsic and synaptic properties in general.

Weakly electric fish also offer a good opportunity to investigate the genetics and development of EODs. The species-specific EODs of mormyriforms offer an opportunity to explore how functionally significant electrocyte membrane shapes are genetically encoded and developmentally expressed. In addition, although the EOD of the larvae of all pulse Gymnotiformes are monophasic (Pereira et al., 2007), some electrocytes of adult Gymnotus and Rhamphychthys are doubly innervated. What are the developmental processes underlying electromotor neuron differentiation and their target affinity?

Investigations focused on the evolution of EOs are also likely to be productive. For example, what evolutionary forces led to the appearance of supplementary EOs as seen in the genus Steatogenys and head extension of the EOs as seen in Gymnotus coropinae, Gymnotus javary and other members of the same clade? Where are the electromotor neurons that innervate these EOs, and what are their functions? Furthermore, the distinction between myogenic and neurogenic EOs is an interesting one. There are actually two known species of weakly electric fish that exhibit mixed neurogenic and myogenic components in their EODs: the activity of the posterior electromotor nerve innervating the tail region of Gymnotus carapo and Gymnotus sylvius has enough strength to contribute to the electric field (Rodríguez-Cattáneo et al., 2013). Are there other similar examples?

It is known that EO auto-excitability is plastic: habitat variables, including temperature and changes in conductivity, cause acclimatory changes in EO auto-excitability (Kramer and Kuhn, 1993; Moller, 1995; Caputi et al., 1998b). The role of electrocyte membrane excitability in plasticity is well documented (Markham et al.; 2009; Markham and Stoddard, 2013; Markham, 2019). It is also known that both major groups of weakly electric fishes express a number of genes – including two collagen genes, a glycosyltransferase and dystrophin (Gallant et al., 2014; Gallant, 2019) – that are likely to determine the space constant of the connective sheath and the magnitude of the longitudinal ephaptic current supporting EO autoexcitation. Is the expression of these genes modulated seasonally?

There are also many questions surrounding the sensory roles of EODs, and comparisons made across the different groups of weakly electric fish are likely to be informative. For example, the patterns of the time course of the received signals are of relatively low spatial complexity in Mormyridae, intermediate complexity in Gymnarchus and wave Gymnotiformes, and high complexity in pulse Gymnotiformes. What are the sensory consequences of these different degrees of complexity? Is it possible to electrolocate precisely by evaluating the patterns of the stimulus time course? What is the communication role of the complex fields generated by pulse fish close to the skin? The hypothesis of electric color is an intriguing one – in order to confirm this idea, it would be necessary to identify electric color-sensitive neurons in higher centers. Are there specific neurons that are involved in identifying conspecifics' color? Another question regarding the sensory aspects of the EOD relates to circadian changes: it is known that pulse Gymnotiformes show different emission patterns across the circadian cycle, including changes in the electric field amplitude (Brachyhypopomus gauderio: Franchina and Stoddard, 1998; Stoddard et al., 2006; Silva et al., 2007; Migliaro, 2018; Migliaro et al., 2018; Vazquez et al., 2023) and pattern of repetition (Gymnotus spp.: Forlim and Pinto, 2014; Camargo et al., 2023), even in isolated fish. What are the sensory roles of these changes?

Finally, one of the most intriguing aspects of EO function is its possible role in communication. A language is a structured system of communication that consists of grammar and vocabulary. I have previously argued that the alternation between chirps, rasps and other typical EOD patterns during interactions between weakly electric fish suggests that these features could be part of a vocabulary. Exploring whether there are systematic alternations in these patterns during fish communication may unveil whether there are specific rules suggesting a sort of grammar.

Seven decades after Lissmann's discovery of the functional role of the EOD and its importance in evolution, research on an effector system that humans do not have – and on a sensory system that does not evoke human-intuitive sensations – has greatly contributed to the understanding of the function and evolution of sensory–motor loops. First, electrogenesis is now well understood from the cellular to the whole-organism level. This understanding has led to the discovery of new variants of voltage-gated channels supporting astonishing electrophysiological properties and has revealed generalizable concepts, such as the combination of delay lines for achieving synchronous and coordinated activities. Second, the study of electric fish has helped us to understand the imaging process in active systems, thanks to the clear separation between the energy types of the signal carrier (the electric field generated by the electro-motor system) and those involved in positioning the carrier source, the electrosensory mosaic and prereceptor conditioning the signals (animal movements controlled by the skeleto-motor system). Knowledge gained from studying these fish has also contributed to general imaging theory, by improving our understanding of generalizable concepts such as an object's stamp and imprimence. Furthermore, the more than 500 known species of weakly electric fish show both different solutions to the same neurobiological problems (for example, whereas mormyrids use electrocyte geometry to produce complex waveforms, gymnotiforms use multiple innervation and diversity of electrocytes) and the use of similar mechanisms to solve different problems (for example, to evaluate the time course of the stimulus to actively evaluate the object impedance and to recognize a conspecific). Unveiling how variation in the organization and time course of the EODs contributes to different strategies for species survival, as well as understanding the genetic encoding and phenotypic expression of the EOD, is a major challenge that might shed light on general evolutionary mechanisms.

The author thanks Dr Kirsty Grant for her invaluable scientific comments and great help with English language editing, the anonymous reviewers for their contributions to improve the manuscript and Mr Marcelo Casacuberta for the supplementary movies.