The Phase-Lock Reflex

Walter Heiligenberg, Electric Knifefish, and the Jamming Avoidance Algorithm

Oscilloscope traces recorded in 1963 at Tokyo Medical and Dental University by neurophysiologists Akira Watanabe and Kimihisa Takeda captured one of the most reliable closed-loop reflexes in vertebrate neurobiology. Floating inside a narrow chamber in murky water, a South American glass knifefish (Eigenmannia virescens) continuously discharged the weak electric organ in its tail at a crystal-like sinusoidal frequency of 300.0 hertz. When Watanabe and Takeda switched on submerged carbon electrodes delivering an artificial sine wave at 302.0 hertz—two cycles per second above the fish's own signal—the fish immediately lowered its pacemaker frequency to 296.0 hertz. When the researchers switched the electrode frequency to 298.0 hertz—two cycles per second below the fish—the animal raised its frequency to 304.0 hertz. In every trial, the fish widened the frequency difference $|\Delta f|$ toward 6 to 8 hertz to prevent the neighbor's field from jamming its active electrolocation system.

First discovered by Hans Lissmann at Cambridge in 1951, wave-type weakly electric fish navigate nocturnal floodplains like the Amazon and Orinoco by surrounding their bodies with a continuous alternating dipole electric field of a few millivolts per centimeter. Whenever a conductive insect larva or resistive rock enters that field, it distorts the local current lines across the skin, casting an "electrical shadow" read by thousands of tuberous electroreceptor pores. If two neighboring Eigenmannia discharge within 1 to 5 hertz of each other, their sinusoidal waves interfere to create slow amplitude and phase beats that blind both animals to obstacles and prey.

At first glance, the Jamming Avoidance Response (JAR) appears to require a central command: surely the fish compares its internal medullary pacemaker signal (an "efference copy" of its own frequency $f_{1}$) against the incoming frequency $f_{2}$ to decide whether $\Delta f = f_{2} - f_{1}$ is positive or negative. Between 1973 and 1985 at the Scripps Institution of Oceanography in La Jolla, California, neuroethologist Walter Heiligenberg and his colleagues Theodore Bullock and Gary Rose demolished that assumption with surgical and biophysical experiments that mapped every neuron of the circuit from skin pore to motor pacemaker.

Heiligenberg injected Eigenmannia with curare-like drugs (flaxedil) that paralyzed the neuromuscular junctions of the tail electric organ, silencing the fish's real electric discharge while leaving its brain pacemaker ticking normally in the medulla. He then replaced the silenced field with an artificial sine wave $S_{1}(t)$ delivered through an electrode inserted into the fish's mouth and stomach, controlled from an external function generator independent of the brain pacemaker. When he added a second jamming signal $S_{2}(t)$ from external tank electrodes, the paralyzed fish's medullary pacemaker still shifted upward whenever $\Delta f < 0$ and downward whenever $\Delta f > 0$. The fish was not consulting its internal pacemaker at all; it computed the algebraic sign of $\Delta f$ purely from the composite waveform striking its skin:

How can a fish tell whether a neighbor's frequency is higher or lower without knowing its own frequency? Because the fish's electric organ sits inside its tail while the neighbor's field impinges from outside, the geometry of contamination varies across the body surface. Some skin patches—Zone A on the cheek—experience a strong local field $A_{1}$ mixed with a weak neighbor signal $A_{2}$, while another skin patch—Zone B on the flank—experiences a different mixing ratio. When two sine waves separated by $\Delta \omega = 2\pi \Delta f$ add together, the combined wave undergoes both cyclical amplitude modulation $|S(t)|$ and cyclical phase modulation $H(t)$ at the beat frequency $\Delta f$.

Plotted on a two-dimensional graph with local amplitude $|S_{\text{A}}|$ on the vertical axis and differential phase between Zone A and Zone B ($H_{\text{A}} - H_{\text{B}}$) on the horizontal axis, the beat cycle traces a closed Lissajous circle. If the neighbor's frequency is higher ($\Delta f > 0$), the state point rotates counterclockwise: phase lead coincides with rising amplitude, and phase lag coincides with falling amplitude. If the neighbor's frequency is lower ($\Delta f < 0$), the state point rotates clockwise.

The fish's skin contains two parallel classes of tuberous electroreceptors tuned to those two axes: P-units (probability coders), which increase their spike rate as local amplitude rises, and T-units (time coders), which fire a single spike locked strictly to every zero-crossing of the carrier cycle to report instantaneous phase. In the electrosensory lateral line lobe (ELL) of the hindbrain, spherical cells compare T-unit arrival times between body regions, while E-cells and I-cells detect rising and falling amplitude. These two streams converge in Layer 6 of the midbrain torus semicircularis. In 1985, Gary Rose and Walter Heiligenberg performed intracellular recordings from single sign-selective neurons in the torus semicircularis, demonstrating that pooling the distributed votes of thousands of local skin comparisons allowed Eigenmannia to shift its pacemaker reliably in response to differential phase shifts between head and tail of just 0.4 microseconds.