Changes in the Electric Organ Discharge After Pausing the Electromotor System Of Gymnotus Carapo

To navigate and communicate, weakly electric fish continuously produce electric fields by synchronously discharging the electrocytes within their electric organs (EOs) (Bennett, 1971; Heiligenberg, 1977; Kramer, 1990). The electrocytes are (except in the Apteronotidae) modified muscle cells in which spikes or excitatory postsynaptic potentials (EPSPs) are triggered in a 1:1 ratio upon input from chemical synapses, in some species at rates of up to 800 s⁻¹ (Bennett, 1971; Bass, 1986; Kramer, 1990). The electromotor system is unique among motor systems in several aspects. In most species, it is able to maintain an uninterrupted train of electric organ discharges (EODs) throughout the life of a fish without any significant pauses to allow the system to recover. Moreover, the time course of each discharge can be kept constant enough to allow recognition among individuals on the basis of their slightly different EOD waveforms (McGregor and Westby, 1992).

A number of weakly electric fish are known to interrupt their otherwise continuous train of discharges briefly when startled, probably to hide electrically from predators (Mormyriformes, e.g. Harder und Uhlemann, 1967; Harder et al., 1967; Gymnotiformes, Black-Cleworth, 1970; Westby, 1974, 1975). The present study uses this naturally occurring behaviour as a starting point to analyse how pausing the electromotor system affects waveform generation. I used the South American knifefish Gymnotus carapo, a pulse-type fish (discharge rate approximately 50 s⁻¹) that can readily be induced to pause and whose complex EOD has been studied in great detail (e.g. Bennett and Grundfest, 1959, 1966; Bennett, 1971; Trujillo-Cenóz et al., 1984; Trujillo-Cenóz and Echagüe, 1989; Lorenzo et al., 1988, 1990, 1993; Caputi et al., 1989, 1993; Macadar et al., 1989; Baffa and Lopes Correa, 1992). The complexity of waveform generation in this species together with the very detailed picture available of the underlying processes (recently reviewed by Caputi, 1999) make it an attractive species in which to explore the possibility of post-pause EOD changes.

The present study provides a detailed analysis of how the amplitudes of the three major deflections of the EOD and the temporal separation between these deflections change over the course of the first 1000 post-pause discharges. In addition, other more subtle waveform changes are described. The pause-induced changes described here are probably due to changes in the inner resistance of the electric organ and comprise both rapid initial changes (within 1 s) and slower changes back to equilibrium during continuous firing.

Twelve specimens of Gymnotus carapo (L.) (total length 10–22 cm) were used in this study. Fish were purchased from Aquarium Glaser, Rodgau, Germany, and were kept individually in tanks of either 60 cm×30 cm×30 cm or 70 cm×40 cm×30 cm (length × depth × height) equipped with standard heating and filtering equipment. Water conductivity ranged between 100 and 200 μS cm⁻¹ and pH between 6.5 and 7.5. Temperature was 26 °C unless stated otherwise. Each tank housed a porous shelter (with the bottom cut away to allow visual observations through the base of the tank) sited centrally in which the fish rested almost motionless during most of the day.

In the present experiments, pausing was elicited by the movement of a visual stimulus, by briefly touching the water surface with a plastic rod or by electrical stimuli generated by an isolated generator (DS 345, Stanford Research Systems). Electrical stimuli were most effective and were cycles (1–100) of sine waves (sometimes triangular waveforms) of various frequencies (100 Hz to 10 kHz), white-noise pulses of duration 1–200 ms or electrical short-circuiting and were applied using carbon dipole electrodes of the type used by Westby (1974, 1975). All stimuli were much shorter than the resulting pauses. The field strength of the electrical stimuli varied between 0.2 and 100 mV cm⁻¹ along the longitudinal axis of the fish and perpendicular to it. For electrical stimuli, either the waveform and amplitude or the location of the dipole electrode was varied from trial to trial to avoid habituation.

Standard recording electrodes were silver wires (length 10 cm), 50 or 60 cm apart in the smaller and larger tanks, respectively (distance to the fish approximately 20 cm), placed along the longitudinal axis of the shelter. All electrical connections were shielded, and heaters and filters were switched off during the experiments. Potential differences between the wires were amplified using two precision amplifiers (EG&G 5113, Princeton Applied Research) in either of two settings: a.c.-coupled with low-pass 6 dB point at 10 kHz, high-pass 6 dB point at 300 Hz (for simple recording of peak-to-peak amplitudes) or quasi-d.c.-coupled (6 dB point at 0.03 Hz) with low pass at 10 kHz.

The use of stationary recording electrodes to observe changes in EOD amplitude that occur during the first 1000 post-pause EODs requires that the position of the fish with respect to the recording electrodes remains unchanged within that time. More precisely, any amplitude changes that occur due to motion of the fish during that time must be smaller than those induced by the pause.

In the initial phase of the present study (30 recordings for each of two fish), this was assessed using the following simultaneously obtained evidence. (i) Direct visual observations of the fish (either viewing it along its longitudinal axis or from below) were made to check for motion of the fish during the 1000 post-pause discharges. (ii) Video recordings of the fish were made during the pause and the onset of discharges with a resolution of approximately 1 mm and 1 °: The fish was videotaped from below at 50 frames s⁻¹ together with the output from a light-emitting diode that was switched on by the first post-pause discharge. (iii) Simultaneous recordings were made of the peak-to-peak amplitude of the EOD (V₃+V₄; see Fig. 2 for definitions) with two pairs of electrodes: one aligned with the longitudinal axis of the fish and the second (silver wire, separation 26 cm) oriented along a line almost orthogonal to the fish.

The visual observations were sufficient to detect linear motion (longitudinal, transverse or upwards). An amplitude change of 1 % would require a movement of at least 3 cm, which would be easily seen. Video recording proved to be an unnecessary improvement in the detection of translation-induced amplitude changes. The minimal detectable rotations of 1 ° in the video recordings would have led to amplitude changes of less than 0.1 %. This value is lower than the amplitude changes recorded (e.g. Fig. 1), and turns can also therefore be excluded as a cause of the post-pause amplitude changes. During simultaneous recording with longitudinal and orthogonal electrode pairs, turns can be detected using their different effects on the respective voltages: while one pair records a voltage increase, the other records a decrease. In contrast, if the amplitude of the source changes, both pairs record the same relative changes. Turns identified in the video recordings could readily be detected by these electrodes.

Direct visual observation was found to be sufficient to exclude translatory motion of the fish, and the use of a second pair of electrodes provided good resolution for turn-induced changes. Therefore, all recordings in the initial part of the study (240 000 EODs, in which only changes in peak-to-peak amplitudes were analysed) used these two methods.

Once it had been established that the regular patterns of post-pause changes in peak-to-peak amplitude (see Fig. 1) were not due to fish movement, only direct visual observation by the experimenter and recording with the longitudinally oriented electrodes were used in a later part of the study (130 000 EODs). This was possible because the earlier experiments had shown that whenever fish motion was visually detected there were also clear distortions in the post-pause changes in peak-to-peak amplitude (V₃+V₄, see Fig. 2). That is, any motion that escaped the experimenter’s attention would have been detected from changes to the regular pattern of post-pause changes in peak-to-peak amplitude.

A Microstar DAP 3200a/415 processor card with a 16-bit resolution A/D converter (programmed in DAPL language, Microstar Labs) was installed in a Pentium computer and controlled by a program written in Borland Turbo Pascal 7.0. A pause counter started when discharges ceased for more than 200 ms and stopped when the first post-pause EOD occurred. Generally, an online analysis of the first 1000 post-pause EODs was carried out. In all figures, original data are shown without smoothing or averaging so that the accuracy of the respective quantity can readily be judged. Unless stated otherwise, steady-state values were calculated as averages of the respective values in post-pause EOD numbers 900–1000.

Initial experiments used two orthogonally oriented pairs of electrodes as described above. Every 1.3 μs, the A/D converter received input from one of the two electrode pairs and switched to the other pair. In this way, electrode pairs were sampled simultaneously at a precisely maintained interval of 2.6 μs. The card’s processor then calculated online, for both channels, pulse onset (to determine the interval between successive pulses) and peak-to-peak amplitude. In addition, it was always checked that the relationship V₂<V₄<V₃ between the amplitudes of phases 2–4 of the waveform (see Fig. 2) held to ensure that peak-to-peak amplitude was always V₃+V₄.

For a finer analysis at a later stage of the study, only one pair of electrodes was used, sampling at 1.3 μs separation. For each of the first 1000 post-pause EODs, an online analysis was made of amplitudes V₂, V₃, V₄, of the time between V₂ and V₃ and between V₃ and V₄ and of the interdischarge interval (the time from V₂ to V₂ of the next discharge).

Complete waveform recordings of successive post-pause EODs (see Figs 2, 9) were also made at 1.3 μs resolution. Here, the processor card received trigger inputs from custom-designed counting modules (Max-Planck-Institut für Biologische Kybernetik, Tübingen, Germany) preset to the desired numbers (usually used to trigger recording of post-pause discharges 1, 20, 500 and 1000).

Relative changes in amplitude with respect to their steady-state values are usually reported throughout this paper. One reason for this is that the relative changes derived from the potential difference between the two recording wires should be the same had they been derived from differently positioned electrodes, e.g. during recording of field strength. Let the potentials of the two recording wires (as used in this study) be denoted by Φ₁ and Φ₂, and those of two other wires by Φ₃ and Φ₄. Because each of the potentials is proportional to the dipole strength (following from the superposition principle of electrostatics), the ratio (Φ₄—Φ₃)/(Φ₂—Φ₁) is unaffected by changes in dipole strength. In other words, changes in dipole strength should give rise to the same relative changes in Φ₄—Φ₃ and Φ₂—Φ₁. This linearity was confirmed (following the calibration described by Franchina and Stoddard, 1998) for both tank sizes used in the present study and for two dipole lengths (4 and 10 cm). Briefly, the dipole, positioned in the middle of the tank, delivered a 1 kHz sine wave (from an isolated generator, DS 345, see above) whose amplitude was varied to cover the full range of voltages seen in the experiments at the recording wires. At the same time, a pair of silver wires (1 cm apart; isolated except at the tips) recorded the field strength at 10 cm from the middle of the dipole. The resulting amplitudes were linearly related. Actual values corresponding to a 5 % relative change in Fig. 3A–C are 0.15 mV, 25.4 μV cm⁻¹ (Fig. 3A), 0.48 mV, 81.4 μV cm⁻¹ (Fig. 3B) and 0.32 mV, 54.3 μV cm⁻¹ (Fig. 3C).

Fig. 1A,B shows post-pause changes in peak-to-peak amplitude in two fish induced to pause periodically over the course of several days. Each data series shows the course of changes in peak-to-peak amplitude over the first 1000 EODs as a Gymnotus carapo resumed discharging after a pause of the indicated duration. Peak-to-peak amplitudes were V₃+V₄ in the notation of Trujillo-Cenóz et al. (1984) (see Fig. 2). For each set of 1000 post-pause EODs, amplitudes were normalized to the steady-state value within that series. The time of occurrence of each post-pause EOD is scaled logarithmically to emphasize the rapid initial decay in amplitude as the first post-pause EODs appear. Increasing pause duration affects the time course of amplitude changes. The first post-pause EOD changes from having a higher to having a lower amplitude than an EOD during steady firing, the rapid initial reduction in amplitude increases as does the slope of the final rebuild phase. It can also be seen that the rebuild phase seems to consist of an initial slow phase and a later faster phase. After pauses of short duration, these two phases seem almost to be separated by a small maximum that becomes a region of intermediate slope at large pause durations.

Fig. 1C,D shows how two prominent features of the plots, the size of the initial reduction between the amplitude of the first EOD and the minimum amplitude (initial drop, Fig. 1D) and the size of the first post-pause amplitude, A₁, normalized to steady-state amplitude (Fig. 1C), depend on the duration of the preceding pause. These experiments relate to the same two fish as in Fig. 1A,B and include data obtained over the course of a few days using stimuli of various origins, strength and duration. After pauses longer than approximately 20 and 30 s in both fish, the amplitude A₁ of the first discharge was smaller than the steady-state value (Fig. 1C). The initial drop increased with pause duration (P⪡0.001) (Fig. 1D).

Two further aspects of the post-pause amplitude changes are worth mentioning. First, pauses of similar duration could be elicited regardless of the nature, amplitude or duration of the pause-inducing stimulus, and the same pattern of post-pause amplitude changes was observed in all cases. This clearly shows that the post-pause amplitude changes are not a trivial consequence of any direct influence an electrical stimulus might have on discharge generation. Second, while the position of the fish remained unchanged during recording of the first 1000 EODs after a pause (see Materials and methods), this was certainly not the case during the hours or even days between the experiments shown, for example, in Fig. 1. However, the fact that the steady-state amplitude is reached within each series of 1000 EODs provides a simple way of deriving the relative amplitude changes for each series. The clear scaling of these relative changes is evident from Fig. 1 and was found even though the longitudinal axis of the fish had changed by up to 40 ° between the experiments. Moreover, the time course of changes in relative amplitude could even (although with much greater scatter) be recorded with an electrode pair aligned at approximately 85 ° with respect to the longitudinal axis of the fish.

To a first approximation, the waveform did not change, for example the relationship V₂<V₄<V₃ held, for all 370 000 recordings of this study. However, given that the origin of the individual deflections of the EOD is well studied in Gymnotus carapo (see Caputi, 1999), a more detailed analysis of EOD waveform changes is rewarding. Fig. 2 shows the waveforms of the first (red), 20th (blue), 500th (green) and a steady-state (approximately the 1100th post-pause; black) EOD in a series recorded as a fish resumed discharging after having paused for 34 s. Changes in amplitudes V₃ and V₄ with time are clearly seen. Interestingly, V₄ is smaller for the first EOD than at steady state, indicating that the relationship between V₃ and V₄ may change over the course of the post-pause EODs. There are also more subtle changes. A saddle (arrow in Fig. 2) during the second phase is typical of the 20th post-pause EOD (it was not seen in the first five EODs and after the 50th). The fourth phase is longer in the 20th and shorter in the 500th post-pause EOD compared with the first EOD. Effects on phase 1 and phase 4 can only be analysed at a much larger amplification (see Fig. 6).

To provide a more complete picture of these changes, Fig. 3 (see also Fig. 5) shows the changes in V₂, V₃ and V₄ as well as other measures obtained simultaneously from the same set of 1000 post-pause EODs. Common to all amplitude changes (Fig. 3A–C) is a rapid initial decrease and a slow later rebuild phase. In addition, these changes in each individual amplitude showed scaling with pause duration and independence of stimulus type, as described above (Fig. 1). However, there are interesting differences among the individual amplitude patterns.

V₃ remains at its minimum value for approximately 100 EODs (Fig. 3B), while V₂ and V₄ increase immediately upon reaching a minimum (Fig. 3A,C). The recovery of V₃ also generally seemed to take longer than that of the other phases. In the first post-pause EOD, V₄ is much smaller and V₃ is larger than during steady-state firing. At its minimum, V₄ is only three-quarters of its steady-state value (Fig. 3C). Fig. 3D,E shows changes in the ratios V₂/V₃ and V₃/V₄. While V₃/V₄ declines rapidly to its steady-state value, V₂/V₃ shows a rapid initial increase and a second slow increase followed by a slow decline.

Fig. 3F presents V₄versus V₃ for each of the 1000 EODs. The relationship between V₃ and V₄ indicates the ability of currents generated during phase 3 to pass to and stimulate the electrocytes that produce phase 4 (see Discussion). During the first few and later post-pause EODs, the amplitudes seem to be almost linearly related and, between these, there is a phase, marked by the straight arrow in Fig. 3F, during which V₄ changes at constant V₃. The slopes (and their respective 95 % confidence intervals) of the initial and later phases are 0.464±0.029 (determined from the first 15 EODs; r=0.994) and 0.724±0.015 (for EODs 200–600; r=0.978). The larger slope in the later phase is highly significant (P⪡0.001; two-tailed t-test).

Fig. 4 shows the time course of changes in V₃/V₄ recorded in experiments in which a Gymnotus carapo had paused for different times from 2 to 86 s. After a longer pause, the initial value of V₃/V₄ increases. The subsequent slight increase in V₃/V₄ seems not to be present after short pauses.

The changes in the duration of the fourth phase seen in Fig. 2 make it advisable also to include measures of changes in the time structure of the EOD in the present analysis. Two simple measures were chosen: the time between the extreme deflections during phases 2 and 3 (d₂₃) and that between phases 3 and 4 (d₃₄) phase. While these variables offer the advantage that their values will not be affected by post-pause amplitude changes and can be easily included in a simultaneous analysis of post-pause EODs, the precision obtainable is limited by the breadth of the peaks. The resulting large scatter is evident in Fig. 5, which shows d₂₃, d₃₄ and d₂₄ (the time between V₂ and V₄) obtained from the same 1000 post-pause EODs shown in Fig. 3. Despite the scatter inherent in using the timing of minima and maxima, general patterns can be seen. The time between the second and third peaks d₂₃ (Fig. 5A) shows an initial rapid decrease of approximately 40 μs and then increases back to its steady-state value. In contrast, the time between the third and fourth peaks d₃₄ (Fig. 5B) shows an initial increase of approximately 60 μs and then decreases back to its steady-state value. As a result of these opposing changes, the time between the second and fourth peaks (d₂₄) changes very little (Fig. 5C). Again, the patterns of post-pause changes in d₂₃ and d₃₄ are affected by the duration of the preceding pause: the initial reduction in d₂₃ and respective lengthening of d₃₄ increased in size with increasing pause duration, e.g. for the fish in Fig. 5, the changes were —80 μs and +100 μs, respectively, after a pause of 199 s.

The EOD of Gymnotus carapo starts with postsynaptic potentials (PSPs) of electrocyte membranes that cannot spike (Lorenzo et al., 1988; Macadar et al., 1989). Hence, an analysis of this phase might give hints as to whether pause-induced changes in PSP size could be involved in post-pause amplitude changes. However, under no condition (even at high temperatures and/or with long pause durations) examined in this study were post-pause effects seen in phase 1. An example is given in Fig. 6A in which the initial first phase is shown of the first (red), 20th (blue), 500th (green) and 1000th (black) EOD produced after a pause of 34 s duration. The increase in slope occurs as phase 2 begins. Only crude measures could be obtained at this point because the bending is not very sharp, but there was no indication of pause-induced changes. Therefore, the slope was analysed in the region between the broken lines in Fig. 6A. In this example, the slopes (and their 95 % confidence intervals) were (in V s⁻¹): 1.750±0.088 (EOD 1), 1.610±0.090 (EOD 20), 1.611±0.076 (EOD 500) and 1.720±0.100 (EOD 1000). None of the differences is significant (P>0.05; two-tailed t-tests).

In all recordings, a very small head-positive deflection was seen after the final phase (phase 4), approximately only 1/30 of the head-positive amplitude V₃. Most interesting are the changes that can be seen in this small deflection over the course of the first few post-pause EODs. In each part of Fig. 6B–G, two traces are shown, both starting at the end of phase 4: the dashed curve is for the 1000th post-pause EOD (i.e. steady-state), the continuous curve is for the 20th post-pause EOD.

Fig. 6B–D shows this comparison for experiments at 32 °C in which a fish had paused for different durations from 7 to 23 s. The deviations seen in the 20th post-pause EOD increased in strength as pause duration increased. A secondary peak (reverberation) becomes evident, and the original head-positive deflection is reduced. Fig. 6E,F shows a similar pattern for experiments at 26 °C. Deviations (e.g. the secondary peak) between the steady-state EOD and the 20th EOD become more prominent after a longer pause. At 21 °C (Fig. 6G), comparable changes in the small head-positive phase were only seen after very long pauses.

Similar effects are seen when comparing the first post-pause EOD with one produced during steady-state firing. In this case, the pause-induced reverberations differ from the saddle shown in Fig. 2 for EOD 20, which was never seen for the first post-pause EOD. No attempt was made to determine the final post-pause EOD number at which reverberations could be seen: however, they were never found after EOD 500.

It is interesting to note that there is a trade-off between pause length and temperature (Fig. 6B–G). Similar patterns of reverberations can be achieved at a lower temperature with a longer pause duration or at a higher temperature with a shorter pause duration.

Pausing affects not only the individual EOD but also the intervals between successive post-pause EODs. It is therefore interesting to examine the degree to which the post-pause EOD changes might be attributable to post-pause changes in the temporal discharge pattern. Fig. 7 gives an example of successive interpulse intervals after a fish had paused for 10 s. The initial intervals are much larger than the average interval at steady state. After approximately 50 intervals, a minimum is reached, and the interval duration then increases slowly to its steady-state value. The pattern differs from the EOD changes described above in that the pattern of post-pause interval changes seems to be much more variable and is not determined by the length of the preceding pause. However, its general pattern is similar to those reported above in that it also shows a rapid initial decline and a slow recovery. However, no relationship between the changes in the inter-discharge interval and any of the post-pause EOD changes analysed in the present paper could be found. For example, Fig. 8A–D shows that there is no simple relationship between peak-to-peak amplitude and interpulse interval duration of the first 1000 post-pause EODs after various pauses in four different fish. In all four plots, there are regions in which the steady-state amplitude is independent of interpulse interval: amplitude remains constant despite variations in interpulse interval. Independent variations in amplitude and interval are also evident in the early stages of post-pause onset of discharging, e.g. see the horizontal segments in Fig. 8B–D (changes in interpulse interval do not affect amplitude), zigzagging in Fig. 8C,D and vertical segments in Fig. 8C (changes in amplitude occur without respective changes in interpulse interval).

Similar plots of amplitudes V₁–V₃ and durations d₂₃ and d₃₄ also showed no relationship between the changes in these quantities and in the EOD interpulse interval.

Fig. 9 illustrates the effects of temperature on the course of changes in peak-to-peak amplitude (Fig. 9A,B) and on post-pause waveform changes (Fig. 9C,D). In the example shown, after a similar pause duration, the peak-to-peak amplitude of the first post-pause EOD is larger than that of a steady-state EOD at 21 °C but smaller at 33 °C (Fig. 9A,B). The post-pause waveform changes described above (see Fig. 2) are almost absent at 21 °C, whereas marked changes can be seen (after a similar pause duration) at 33 °C. Therefore, there seems to be a trade-off between temperature and pause duration: to induce the same effects at a lower temperature, a pause should be longer.

A simple interpretation for this effect is illustrated quantitatively in Fig. 10 for the initial drop in peak-to-peak amplitude. Fig. 10A shows how the initial drop depends on pause duration for three temperatures (21, 26 and 31 °C). The scatter is much greater than in the corresponding Fig. 1D, obtained at a fixed temperature of 26 °C. However, it is clear that a similar trade-off between pause duration and temperature exists for the initial drop as for the reverberation phenomena illustrated in Fig. 6B–G: a longer pause duration is necessary to achieve the same initial drop at a lower temperature. This could be the result of processes occurring during the pause that determine the size of the initial drop in amplitude after the pause. Pause duration can therefore be rescaled by assuming a fixed temperature coefficient (Q₁₀) for these imaginary processes and then rescaling pause duration accordingly. The remaining scatter in a linear regression between the initial drop and the logarithm of rescaled pause duration is then determined, and a new Q₁₀ is assumed. This iterative process is repeated until the least scatter is obtained. Fig. 10B gives the results of this process on the data presented in Fig. 10A. The data points now follow a reasonably smooth curve. The Q₁₀ thereby assigned to the hypothetical process is approximately 8.

While a pause duration versus temperature trade-off determines the initial drop in amplitude and the reverberation effects (Fig. 6B–G), it should be stressed that it cannot explain all the characteristics of the post-pause EOD changes. For example, post-pause changes in peak-to-peak amplitude recorded at 21 °C were never exactly the same as at 33 °C, even with appropriate pause durations. While two pause durations can be found where post-pause amplitude changes recorded at two different temperatures show the same initial drop, their patterns always differ in other ways, e.g. in the normalized amplitudes of the first EODs.

A post-pause initial reduction in amplitude was seen for amplitudes V₂–V₄ (Fig. 3A–C). This suggests that only after a certain number of post-pause discharges have occurred are the slow processes started that restore these amplitudes to their steady-state levels. One group of experiments attempted to determine directly whether this was so and the number of post-pause EODs needed to turn on such processes. Fig. 11 gives an example of a double-pause experiment used to estimate this hypothetical number of discharges. After the test fish had paused its discharges for duration p₁, it again started to fire. However its nth EOD (n was set by the experimenter to between 1 and 500 using the counting modules described in Materials and methods) triggered a stimulus that silenced the fish again for a duration p₂. It was therefore possible to study the onset of discharging after the two pauses p₁ and p₂ as a function of the number of EODs that had occurred between the two pauses. If the number n of EODs preceding the second pause p₂ were sufficient to start regenerative processes, then the amplitude of the first EOD after pause p₂ should be larger than that of the last EOD before p₂. If the n EODs had no effect, the pattern of amplitude changes should continue at the level reached before the second pause. The available experimental evidence is, unfortunately, limited because the second silencing stimulus often did not cause discharging to stop immediately. The smallest number of intermediate discharges whose effect could be studied was 15. It is clear, however, that the first post-pause discharges after the first and second pauses had similar amplitudes if there were more than 20 intermediate discharges. Therefore, regenerative processes appear to be triggered even after only a few post-pause discharges.

The present study has demonstrated that the EOD of Gymnotus carapo undergoes a variety of changes when the electromotor system is reactivated after a pause. One of the most striking features of these post-pause changes is the time scale over which they occur: a rapid initial reduction in amplitude of 15 % may occur within less than 1 s. The subsequent slower rebuild to steady-state levels occurs within 1000 post-pause EODs or approximately 20 s. The changes are easy to study in unrestrained fish because G. carapo usually remains motionless during these periods and simple methods can be used for checking this. In the following discussion, an attempt is first made to interpret the post-pause changes in terms of a detailed model available for discharge generation in G. carapo (Caputi, 1999). The dynamics of the amplitude changes will then be described quantitatively by a simple heuristic model with activity-dependent kinetics.

According to a detailed model of waveform generation within the complex EO of Gymnotus carapo (for a recent review, see Caputi, 1999), the following processes occur during the EOD. It starts with a slow head-negative phase generated by PSPs at the rostral faces of abdominal electrocytes, which are unable to fire a spike (Lorenzo et al., 1988; Macadar et al., 1989). Later, spikes in the rostral faces of central electrocytes generate the head-negative second phase of the EOD. After this phase, the caudal faces of rostral, central and caudal electrocyte groups are activated, giving rise to the strong head-positive third deflection. The final head-negative phase starts when currents produced during the third phase trigger action potentials in the rostral faces of purely caudally innervated electrocytes. All electrocytes active during the individual steps act in complete synchrony.

The pattern of post-pause EOD changes described here could have resulted from a variety of factors. (i) The relative timing of when the different electrocytes become active might change, thereby leading to partial cancellation between successive deflections. (ii) The synchrony of electrocyte activation within each phase might be affected, and reduced synchrony would diminish the amplitude. (iii) The efficacy of the synapses between electromotor neurons and electrocytes might be changed. (iv) The current produced by each electrocyte might change. (v) The externally recorded EOD might also be affected by changes in the inner resistance of the EO (e.g. see Bell et al., 1976; Heiligenberg, 1977; Caputi et al., 1993; Hopkins, 1999). For example, a decrease in inner resistance would yield an increase in the externally recorded amplitude because of the increase in current flowing across the external resistance between the electrodes.

Two findings make it likely that the last factor listed above is the main cause of the changes described here. First, no effect of the pause on the slope of the purely PSP-derived first phase could be seen. This suggests that, at least for the rostral side of the abdominal electrocytes, the pause has not led to changes in synaptic efficacy. This finding does not rule out post-pause changes in PSP size as being the cause for the changes in deflections V₂ to V₄, but it seems to make them rather unlikely. Second, the increased ability of a given constant V₃ to yield a larger-amplitude V₄ (as seen, for example, in Fig. 6F) indicates directly changes in the inner resistance over which the action current (of the third phase) must flow to trigger the fourth deflection.

Such changes in inner resistance could explain the overall pattern of the amplitude changes. They could also explain the initial increase in the time d₃₄ during the first few post-pause EODs (Fig. 5B): at an increased inner resistance, only at a later time is the action current set up during the third phase large enough to trigger the fourth phase. The common dependence of the EOD changes on pause duration could also simply be explained by having changes in inner resistance as their common cause.

If this picture were correct, the following interpretation would be imposed on the other phenomena reported. The differences between the individual phases (Figs 2–5) could be attributed to the different spatial effects inner resistance changes have on the different groups of electrocytes, as discussed by Caputi et al. (1993). In particular, the failure to see any post-pause effects on V₁ in the present study would mean that the changes in inner resistance do not affect the PSP-produced current of the abdominal electrocytes. It should be noted that, if the more rostral electrocytes are indeed less affected by the resistance change, two other observations of this study could also be explained: the initial increase in amplitude of the second phase relative to the third (V₂/V₃; Fig. 3D) and the initial decrease in the time d₂₃ (Fig. 5A). The first follows because resistance changes would affect V₂ less than V₃, and the second could be imagined from the effect that a relative increase in amplitude of the second phase with respect to the third could have on the resulting (superimposed) waveform. Also, the finding of reverberations (Fig. 6B–G) only after a pause but not during steady firing of the EO could be interpreted accordingly. Bennett and Grundfest (1966) have described a nonlinear current–voltage relationship in gymnotid electrocytes different from Hodgkin–Huxley-type conductivity changes. When current was passed through these electrocytes, their membrane resistance jumped to a tenfold larger value. This nonlinear phenomenon has been noted to lead to reverberations at the end of a current pulse and might, therefore, account for the slight head-positive deflection at the end of the fourth phase. According to this view, the finding of additional reverberations after a pause, but not during steady firing, could be due to the current stimulating the caudal faces of the doubly innervated electrocytes being smaller after the pause than during steady firing.

The saddle seen in the course of the second deflection around the 20th post-pause EOD does not fit easily into this scheme. Also, it could be argued that the changes in times d₂₃ and d₃₄ would indicate that pause-induced changes in synchronisation also play a role in determining the amplitude changes. However, it must be noted that d₂₃ and d₃₄ are constant after approximately the 200th post-pause EOD (Fig. 5), when the amplitudes V₂–V₄ are still rising towards their steady-state values (Fig. 3).

The regular pattern of changes described here for the main deflections of the EOD of Gymnotus carapo with its fast initial phase and slow recovery phase (Fig. 3) and their dependence on pause duration suggest that a simple description of the underlying dynamics should be possible. Below, the simplest type of model with which to describe the present findings is assessed. The heuristic value of such a model would be to characterize further the dynamics of the underlying processes, although their detailed nature remains unknown. This latter point must be stressed: the following equations are, at this point, only a quantitative description of the dynamics of the changes.

A further equation is needed to account for the changes that occur immediately after an EOD. A simple assumption is that A is changed to a smaller value ηA as if each discharge had consumed a constant fraction (1—η) of the store.

Not much can be said about the nature of the processes involved. As discussed above, they are probably involved in lowering the inner resistance of the electric organ. While individual ion channels that could be involved in such changes are expected to have Q₁₀ values of between 2 and 4 (Hille, 1984) a chain reaction in which, for example, the net rate constant is the product of the individual ones could give a total Q₁₀ representing the sum of the individual ones and could be the basis for the high Q₁₀ that was used in Fig. 10B.

The lack of any pause-induced changes in the purely PSP-derived first phase of the EOD and the pattern of changes in the amplitude ratio V₃/V₄ make it likely that the post-pause amplitude changes are mainly due to activity-dependent changes in the inner resistance of the electric organ of Gymnotus carapo. The fast initial phase of these changes occurs within less than 1 s. It would therefore be most interesting to determine whether G. carapo or other species are able actively to exploit fast changes in inner resistance to modulate their EODs rapidly. To date, the fastest spontaneous amplitude changes (of approximately 40 %) of pulse-type EODs were reported in Brachyhypopomus pinnicaudatus (Franchina and Stoddard, 1998) to occur within 10 min and could be caused by a protein-kinase-A-induced increase in the Na⁺ current of each electrocyte (Zakon et al., 1999).

While the present study has stressed the interpretation of the pause-induced changes in terms of waveform generation and the modelling of the general pattern of amplitude changes, it should also be noted that these findings provide a new experimental approach for studying the implications of the rapid EOD changes during electrolocation.

It is a pleasure to thank the referees of this paper for their highly stimulating comments and suggestions that considerably helped to improve the manuscript. Cordial thanks also to Sam Rossel and Klaus Vogt for critically reading an earlier version of the manuscript. The experiments comply with the ‘Principles of Animal Care’, publication No. 86-23, revised 1985, of the National Institute of Health and also with the laws of Germany.