Electric fish monitor modulations in sensory feedback from their own electric organ discharges (EODs) to locate moving objects and to detect interfering EODs of their neighbors. The gymnotoid genus Eigenmannia minimizes detrimental effects of jamming by EODs of its neighbors by shifting its own EOD frequency away from a neighbor's EOD frequency that is too close to its own. Since the animal lowers its own frequency if its neighbor's frequency is higher and raises its frequency if its neighbor's frequency is lower, this jamming avoidance response (JAR) requires that the animal determine the sign of the difference frequency, Df, between the interfering EODs. Eigenmannia obtains this information by evaluating modulations in the amplitude and phase which its nearly sinusoidal EOD signal experiences due to the interference with the neighbor's EODs. The necessary logical operations are executed in the dorsal torus semicircularis, an analogue of the inferior colliculus of higher vertebrates, and are similar to operations underlying directional hearing. By intracellular labeling of physiologically identified cells we have identified the anatomy and functional characteristics of neurons involved in the processing of amplitude and phase information. The JAR is controlled by hierarchical and parallel processing of information in several laminae of somatotopically ordered neurons. Phase differences between signals received by electroreceptors in different parts of the body surface are computed in lamina 6. Information about differential phase is then relayed to multipolar cells in the deeper laminae 8, b and c, which also receive information about modulations in local signal amplitude. These cells are excited by a rise or fall of amplitude as well as by a lead or lag of phase. According to their responses to either of these two variables, these neurons can be divided into four classes. These classes encode all information necessary for the control of the JAR and project to the optic tectum. Dynamic properties and sensory specificities of the JAR are not found in individual, properly tuned neurons but rather emerge statistically from the joint effects of a large population of imprecisely tuned neurons. This system is characterized by a distributed pattern of organization and by the absence of a small number of key neurons whose malfunction would jeopardize the behavioral response. The complexity of this neural machinery appears unnecessary for the logically simple task of the JAR, and it suggests that this behavior was acquired later in evolution by being derived from more general motor responses to moving objects.(ABSTRACT TRUNCATED AT 400 WORDS)