The control ofEigenmannia's pacemaker by distributed evaluation of electroreceptive afferences

Summary

This study analyzes the algorithm by which a distributed system of neurons evaluates specific features in sensory feedback and thereby controls a behavioral response.

The electric fishEigenmannia raises or lowers the frequency,f ₁, of its electric organ pacemaker in response to a neighbor's frequency,f ₂, which is slightly lower or higher respectively than its own frequency. This Jamming Avoidance Response (JAR) thus serves to increase the difference in frequencies,Δf=f ₂−f ₁, which is necessary to enhance the animal's electrolocation ability. The JAR is driven by electroreceptive afferences from the individual's electric organ discharges (EODs) which, in different parts of the body surface, are unevenly contaminated by EODs of the other fish. This can be demonstrated, in an open loop experiment, by replacing the silenced, nearly sinusoidal EODs of a curarized fish by a sine wave stimulus,S ₁, mixed with a similar stimulus,S ₂, which mimicks EODs of another fish. The JAR is controlled by joint modulations of instantaneous phase,H, and amplitude, ¦S¦, of the composite signal,S ₁+S₂. These modulations, if plotted in a two-dimensional state-plane withH and ¦S¦ as its axes, form a circular graph whose sense of rotation is counterclockwise for positiveΔfs and clockwise for negativeΔfs. The sense of rotation thus determines the direction in which the pacemaker frequency is to be shifted. Modulations ofH and ¦S¦ are encoded by specialized electroreceptors,T- andP-units respectively (Fig. 1).

Applying various forms of artificial computer generated modulations of phase and amplitude, many of which never occur in a natural situation, and by independently stimulating different parts of the curarized animal's body surface, we demonstrate the following:

• 1. The animal detects a modulation in phase,H, of the stimulus in an area,A, of its body surface by comparing the arrival time,t _A , of signals fromT-units in areaA with the arrival time,t _B , of signals fromT-units in a different area,B. Differential contamination of the animal's EOD by a foreign EOD in areasA andB will cause a periodic modulation of the interval,t _A −t _B . No JARs can be elicited without such modulations in at least some parts of the animal's body surface.

• 2. Increases as well as decreases of the amplitude, ¦S¦ _A , of the stimulus in areaA may either accelerate or decelerate the pacemaker, depending upon the momentary value oft _A −t _B . Amplitude modulations drive the JAR most likely via associated modulations inP-unit activity (Figs. 3–6).

• 3. The mechanism in (2) is such that, with areaA being more heavily contaminated than areaB, the pattern of joint modulations of ¦S¦ _A andt _A −t _B accelerates and decelerates the pacemaker for negative and positiveΔfs respectively. The respective modulations, ¦S¦ _B andt _B −t _A , in areaB, which yield a graph with a sense of rotation opposite to that in areaA, cause responses which are opposite but smaller in absolute terms than those in areaA. As a consequence, the net effect of this pairwise interaction between areasA andB is a shift in pacemaker frequency in the direction favored by the variables of areaA, ¦S¦ _A andt _A −t _B (Figs. 11, 14).

• 4. The JAR thus appears to be driven by a distributed system of pairwise interactions between neighborhoods ofT-units on the animal's body surface. These interactions control the effect of local modulations inP-unit activity upon the frequency of the pacemaker. The nature of this system does not require that the animal has any central representation of its own or any neighbor's EOD activity.

The findings in this paper are based on behavioral experiments on basically intact animals. The subsequent paper treats neuronal correlates of the JAR.