Passive and active electroreception during agonistic encounters in the weakly electric fish Gymnotus omarorum

Behavioural experiments of agonistic contests were carried out in dyads in which the weight of the smaller fish was 75%–95% of the weight of the larger fish (sixteen non reproductive adult fish, eight dyads), using the gate protocol (Silva et al 2007, Batista et al 2012, Zubizarreta et al 2012, Silva et al 2013). In this condition, it can be assumed that fish will fight over territory only.

To characterize the trajectories of approach in the evaluation phase of the agonistic contests and to evaluate the electrical cues used by the fish, fish were filmed from below in a glass-tank (110 cm × 80 cm × 25 cm). Three plastic gates initially separated the fish and ensured that animals were electrically isolated prior to the start of the experiment (figure 1(A)). The isolation was checked by placing a single fish in each compartment and recording its EOD in the other compartments. The partitions separating the fish were opened 10 min after lights were turned off and fish were removed from the test arena 10 min after resolution of the conflict.

Zoom In Zoom Out Reset image size

All experimental procedures were approved by the institutional ethical committee (Comisión de Ética en el Uso de Animales (CEUA), Instituto de Investigaciones Biológicas Clemente Estable, MEC, 007/02/2010).

Methods of the simultaneous recordings of EOD and behaviour have been previously described (Silva et al 2007). In brief, the tank was fitted with two orthogonal pairs of electrodes; each pair attached to opposites walls (figure 1(A), orange and yellow lines) connected to high input-impedance amplifiers (FLA-01, Cygnus Technologies, Inc.). Fish were held in separate partitions of the tank for 2 h before the experiment (water temperature: 20 °C–22 °C, conductivity: 100 μS). The experiments were performed in total darkness at night using IR-illumination (L-53F3BT, Fablet and Bertoni Electronics). The tank was filmed at 30 FPS (SONY CCD-Iris) and both the images and EODs were digitized online (Pinnacle Systems PCTV HD Pro Stick). The midpoint and head location of each fish was localized using a Matlab routine (figures 1(B) and (C)). These data were used to model passive and active EIs prior to the first attack (figure 1(D)). The first attack latency was defined as the time of the first aggressive physical contact (bite or nudge) towards the other fish. Conflict resolution was established at the moment of the third consecutive retreat of one of the fish without attacking back. This criterion unambiguously defined subordination status (dominant and subordinate); fish fulfilling this requirement were never observed to change their status in the following 10 min of interaction (Batista et al 2012).

To model the EO of G. omarorum, we used data published by Rodriguez-Cattaneo et al (2013). Internal tissues and the skin resistance values were obtained from Caputi and Budelli (1995).

The voltage difference between eight consecutive transverse planes of the fish placed at different sites of the body is mainly produced by the regions of EO encompassed by these planes and, according to Ohm's law, it is equal to the current (I) flowing through the internal tissues between each pair of planes times the resistance (R) of that section of the fish's body (figure 2(A)). Then, from the resistance of a given section of the fish body (R) and the measured voltage (V) across it, we were able to calculate the current causing the voltage drop: I = V/R.

Zoom In Zoom Out Reset image size

This is based on the simplifying assumption that for a small longitudinal region of the EO the electrocyte population is homogeneous and according to the simplest assumption, electrocytes within a short segment of the EO are oriented similarly and fire almost synchronously. Thus in the model, the current generated by the series of identical dipoles -mimicking the electrocytes inside a cylindrical body slice- is equivalent to another dipole. This is because the rostral pole of one dipole adds with the caudal pole of the next dipole. Consequently, all the intermediate poles are cancelled and the line of dipoles is equivalent to a single dipole with poles situated at the transverse planes limiting that piece of fish.

The longitudinal resistance (R) of a section of fish can be calculated from the geometry and resistivity (ρ).

$$\begin{eqnarray}&&{\bf{R}}={\boldsymbol{\rho }}\cdot {\bf{l}}/{\bf{S}},\end{eqnarray} \tag{ 1 }$$

where l and S are respectively the distance between recording electrode planes and the average cross-sectional area of the encompassed body portion.

Hence:

$$\begin{eqnarray}&&{\bf{I}}=({\bf{V}}\cdot {\bf{S}})/({\boldsymbol{\rho }}\cdot {\bf{l}}).\end{eqnarray} \tag{ 2 }$$

The poles lying on the plane separating contiguous longitudinal pieces of the fish can be reduced to one by addition, and the EO can be represented by a set of poles equal in number to the planes limiting the experimentally studied regions of the fish. This method requires to identify whether there are abrupt transitions in the regional EOD waveform and to place gap limiting planes at the transition points (figure 2(B)). Since the shapes of the fish of a given species are usually the same (they are homotetic) and the head to tail voltage, outside the water, is almost the same, the parameters of a given fish can be obtained from another fish with different size. The new model can be made from the other by multiplying the parameters from the original with constants of proportionality. Figure 2(C) shows the models of two fish and the resulting head to tail voltages calculated underwater (Sanguinetti-Scheck et al 2011, Pedraja et al 2014).

Modelling of EIs was done using software developed by Diego Rother (2003). This model has two parts, a geometric reconstruction of the fish's body and a calculation of the transcutaneous field. The model was constructed under the following assumptions:

(1) All media are ohmic conductors. This means that the vector representing the current density at the point x (J(x)) is proportional to the vector electric field at the same point (E(x)). Then:

$$\begin{eqnarray}&&{\boldsymbol{J}}({\boldsymbol{x}})={\boldsymbol{\sigma }}({\boldsymbol{x}})\cdot {\boldsymbol{E}}({\boldsymbol{x}}),{\boldsymbol{\sigma }}({\boldsymbol{x}})\gt {\bf{0}}.\end{eqnarray} \tag{ 3 }$$

The proportionality constant σ(x), is 'the volumetric conductivity at the point x'.

(2) Given that the dielectric relaxation of the media is in general shorter than the minimum significant period of the EOD Fourier components, the model is an electrostatic approximation (Bacher 1983).

(3) The fish and other objects are immersed in an infinite water medium. The shapes of the fish body and objects are approximated by an external surface composed by triangles, allowing an approximation of the object shape that is limited only by the computation power available. Every object should be covered by a thin resistive layer (the skin in the case of the fish), which can be homogeneous or heterogeneous in resistance (magnitudes specified as desired).

The model is based directly on the charge density equation which, under our assumptions, implies that the charge generated by the sources (f(x)) is equal to the charge diffusion (∇J(x)) f(x) = ∇J(x), and therefore σ∇²ϕ(x) = σΔϕ(x) = f(x), where ϕ(x) is the local potential at point x.

This differential equation, the so-called Poisson equation, can be solved for the electric fish boundary using the boundary element method (BEM) as proposed by Assad (1997). Briefly, this method determines the boundary electrical distributions solving a linear system of S*N equations for S sources and N nodes, where the unknown variables are the trans-epithelial current densities and potentials that correspond to each node (for a detailed description of the method see Hunter and Pullan 2002. The known variables were the location of the nodes, the location and magnitude of the poles representing the EO inside the fish, the conductance of the internal tissues, the skin and the water. It is important to note that, differently from Assad's method, a set of important constraints of the model were those posed by the EO equivalent sources that we measured experimentally using the air gap method. In the first instance, only trans-epithelial current densities and potentials are calculated at the 'skin' nodes and these are then linearly interpolated in the triangles defined by the nodes. This allows the calculation of the potentials in the surrounding space. In this work, after the calculation of the transcutaneous current along the fish skin, a longitudinal section (on a horizontal plane) was taken to represent the EI (figure 2(D)).

EIs throughout this manuscript are represented by the root mean square transcutaneous current on each node. For an active EI, we calculated the difference between the EIs in presence of the contender and in its absence. To represent the temporal change of both active and passive EIs we reduced the complexity of the EIs by analysing changes in current density along one horizontal line (figure 2(D)). To visualize the spatio-temporal effect of behaviour on the EIs we stacked these simplified EIs for successive EODs to 2D temporal maps that represent the gradual change of the EIs over the body of the animals. Similar maps were created for simplified hypothetical trajectories. In these cases a marker to indicate the position of the maximum in each electrical image was superimposed to the 2D maps (figures 6 and 7 black lines).

Videos of both the evaluation phase (prior to first attack) and the contest phase (from first attack to contest resolution) were analysed. In 7 out of 8 dyads the larger fish was the winner (became dominant; binomial probability p = 0.03, marked with * in figure 3). This clear effect of size indicates that a pre-contest assessment of the size difference could be used by the animals to decide when to attack. If this were the case we would expect the larger fish to attack first. However, we found no significant differences in the fist-attack rates between small and large fish (binomial probability, no significant difference, ns p = 0.27, figure 3).

Zoom In Zoom Out Reset image size

While these data indicate that contenders either cannot remotely assess size difference or, do not use this information prior to engaging in agonistic behaviour, the trajectory of the fish suggests that the fish used electric information of their contender in general. In figure 4 we show the absolute angular body deviation of two example dyads (upper panels) and the time sliding-window cross correlation of these data (bottom panels). Our results show that the turning behaviour was correlated, indicating that the fish have access to information on the conespecific behaviour and position even without visual input (experiments were conducted in darkness using IR-illumination). Similar data were found in five of the six dyads (R > 0.5, correlation coefficients; p < 0.05). This motivated us to analyse the EIs associated with the specific behaviour as this arguably is the most likely source of sensory information given the experimental design.

Zoom In Zoom Out Reset image size

The above data show that fish use electric information prior to the first attack. Depending on the source of electric information being used, two different approach strategies can be expected: (1) using the passive EI information, following the field lines generated by the contender's EOD, or (2) using the active EI information, following the field lines generated by the perturbation produced by the presence of the contender on its own EOD.

Figure 5 shows the pre-contest phase of six dyads. While the left column shows the trajectories of the individual fish, images to the right show the temporal series of EIs for these trajectories. EIs were calculated along a horizontal line (as showed in figure 2(D)), both for the active (middle) and passive (right) EIs. The smaller fish is represented in red and the larger fish in green throughout the figure. The insets in the middle and right columns show the maximum amplitude of each EI as a function of and proximity between contenders.

Zoom In Zoom Out Reset image size

The first dyad (figure 5(A), left column) shows an indirect approach made by both fish that resulted in the bigger fish making the first attack. When the fish are close, they are placed with the heads almost touching each other and the bodies drawing an angle close to 100°. In this situation the maxima of the active and passive EIs are located at the head of both fish. In the bigger fish a second peak in the passive and active EIs appear on its right trunk. In the smaller fish a second peak appears on the left trunk, the side facing the contender, only for active images. Note that, while the amplitude of the maxima of the EIs is comparable between both fish, it peaks in the small fish for the final frames of the approach sequence (insets). The second dyad (figure 5(B)) shows a direct frontal approach that resulted in the bigger fish attacking first. Both passive and active EIs are maximal at the rostral regions and the peak values of the sensory images are comparable throughout the approach in both fish (albeit slightly larger in the smaller fish, see insets). In the third dyad (figure 5(C)) fish approached each other indirectly. Initially they were oriented almost perpendicular to each other and the approach ended with fish facing each at an angle close to 120°, at which point the smaller fish attacked. At the beginning of the approach the EIs are located both in the frontal and caudal regions of both fish. As fish increased the distance between each other and turned, the EIs decreased in amplitude but remained maximal at the head until the distance between fish decreased again for the final section of the approach. During this phase the EIs increased, peaking at the rostral region of both fish but with a second maximum at the tail. The active EI maximum is higher in the larger fish at the beginning while it is smaller in this fish for the later phase. The passive EI maxima are higher in the smaller fish throughout the sequence.

The sequence shown in figure 5(D) is the only initial approach sequence of the dyads where fish initially approach each other, but no physical contact was made and the larger fish retreated after the smaller fish had approached the larger fish in a roughly orthogonal manner. Prior to the larger fish retreating, passive and active EIs are maximal in the caudal regions with a second weaker peak in the frontal region in both fish. The maxima of both EI were larger in the larger fish (inset). However, the magnitude of the second EI-peak at the head region was constantly larger in the smaller fish. In the fifth dyad (figure 5(E)) fish approach along perpendicular trajectories until the larger fish reaches the collision point and orients towards the approaching opponent which in this case attacked first. During the perpendicular approach the active and passive EI maxima are located in the caudal region in the larger fish while they are found in the frontal region for the smaller fish. While the magnitude of the active EI peak is higher in the bigger fish, it is the opposite for the passive EI. The approach trajectory in the sixth dyad (figure 5(F)) was also almost perpendicular between the fish and resulted in the smaller fish attacking the larger one from behind. At this point the EIs are maximal in the rostral part of the smaller fish while they are found in the rostral and caudal region at the right side of the larger fish. The magnitude of both active and passive EIs was higher in the smaller fish (inset, note that the difference is higher when comparing EIs in the head region only as opposed to the absolute peaks). A common finding of the otherwise variable behaviour shown in figure 5 is that passive and active EIs are consistently centred on the head and tail regions, almost irrespective of the relative orientation between contenders. This can be explained in part by the elongated body shape tapering off towards the tail. This reduces the cross-sectional area towards the tail, increasing the resistivity and hence funnelling electric currents (generated either by the fish itself- or by external sources) towards the head region (the region that also contains the highest density of electroreceptors, (Castello et al 2000, Bacelo et al 2008). As a result the maximal current densities (and transcutaneous voltages) are at the head. A second trend that emerges from this analysis is the finding that the animal that perceives the larger passive EI amplitude is the one more likely to initiate an attack. While this was not significant given the low sample size (binomial probability, p = 0.09, n = 6), it was found in 5 out of 6 cases. This indicates that information related to the attack is evaluated using passive EIs.

To learn how images depend on the size and relative position of the fish, we studied the images produced in simplified collinear, orthogonal and 45° approaches to the head of a stationary fish (fish being 8 and 16 cm). The goal of this analysis is to understand the potential contribution of the two sub-modalities used in electrolocation (active and passive) in agonistic behaviour.

When assuming collinear behaviour (figures 6(A)–(C)) the maxima remain localized at the tail when a fish is being followed, while they are maximal at the head in the follower fish or if animals face each other. Note that the EI-magnitude is almost always bigger in the smaller fish. The exceptions are the cases where the large fish approaches the caudal part of the smaller fish and vice-versa. In the first case the passive EI is larger in the larger fish, whereas in the latter case the active EI is bigger in the large fish (compare figures 6(B) and (C)). In figures 6(D) and (E) we model orthogonal approaches. As expected, both active and passive EIs are maximal at the head for the approaching fish. Interestingly, we find that the EIs are bigger in the smaller fish when it is approached by a larger distant fish (figure 6(E)). In this case the maxima of the EIs in the smaller fish for active and passive images are found at the tail and head, respectively. As expected, at closer range, both passive and active EIs are located in the middle of the trunk of the stationary fish, irrespective of its size. If both animals move along orthogonal trajectories (figure 6(F)) the maxima are located at the head region facing towards the contender with the magnitude being bigger in the smaller fish. For an approach at an angle of 45° (figures 6(G) and (H)) the maximum is always at the head of the approaching fish while the maxima in the stationary fish transverse from caudal to rostral. Again, EI magnitude is bigger in the smaller fish, except when the larger fish is approaching.

Zoom In Zoom Out Reset image size

A frequently observed behaviour in electric fish is repetitive back-and-forth swimming along an object ('va et vient', Toerring and Belbenoit 1979). Such behaviour is known to aid in electrolocation (Hofmann et al 2014), but it also occurs in social interactions. We thus decided to study the impact of this behaviour on passive and active EIs, both for parallel and antiparallel orientations, assuming a fixed lateral distance of 2.5 cm between fish. In figure 7 the upper panels show results for active EIs, while lower panels show the corresponding passive EIs. Data in the left column summarizes results for differently sized fish (the smaller being half the size of the other fish), while results on same-sized interactions are shown in the right column. As expected, antiparallel orientation results in maximal EIs close to the tip of the head (figures 7(A) and (B)) in both EIs and reach peak amplitude when contenders reach zero head-to-head distance. After this, the maxima move to the side of the trunk that faces the contender. Notably (see below), EIs now become double-peaked and the maxima finally move towards the tail. For same-size interaction (figure 7(B)), the EIs are similar between contenders, due to the symmetry of the scene. When both fish face in the same direction, the maxima of the EIs are on the head of the approaching and on the tail of the fish being approached fish (figures 7(C)–(E)). When the approaching fish gets closer, EIs maxima move to the tail of the approaching fish and to the head of the approached fish. Remarkably, there are two inversions of the movement direction of the maximum, which finally are situated at the tail of the moving and the head of the stationary animal. This is particularly evident for the passive EIs. Both for parallel and anti-parallel orientations of the fish, two-peaked EIs occur. The distance between the peaks changes with difference in sizes between fish, an effect also seen in figures 7(A) and (B). From this, we hypothesized that G. omarorum may determine the size of a contender, using the distances between the two maxima during antagonistic displays.

Zoom In Zoom Out Reset image size