The overall goal of this method is to record single-unit responses from an identified population of neurons. In vivo electrophysiological recordings from individual neurons are critical for understanding how neural circuits function under natural conditions. Traditionally, these recordings have been performed 'blind', meaning the identity of the recorded cell is unknown at the start of the recording. Cellular identity can be subsequently determined via intracellular(1), juxtacellular(2) or loose-patch(3) iontophoresis of dye, but these recordings cannot be pre-targeted to specific neurons in regions with functionally heterogeneous cell types. Fluorescent proteins can be expressed in a cell-type specific manner permitting visually-guided single-cell electrophysiology(4-6). However, there are many model systems for which these genetic tools are not available. Even in genetically accessible model systems, the desired promoter may be unknown or genetically homogenous neurons may have varying projection patterns. Similarly, viral vectors have been used to label specific subgroups of projection neurons(7), but use of this method is limited by toxicity and lack of trans-synaptic specificity. Thus, additional techniques that offer specific pre-visualization to record from identified single neurons in vivo are needed. Pre-visualization of the target neuron is particularly useful for challenging recording conditions, for which classical single-cell recordings are often prohibitively difficult(8-11). The novel technique described in this paper uses retrograde transport of a fluorescent dye applied using tungsten needles to rapidly and selectively label a specific subset of cells within a particular brain region based on their unique axonal projections, thereby providing a visual cue to obtain targeted electrophysiological recordings from identified neurons in an intact circuit within a vertebrate CNS. The most significant novel advancement of our method is the use of fluorescent labeling to target specific cell types in a non-genetically accessible model system. Weakly electric fish are an excellent model system for studying neural circuits in awake, behaving animals(12). We utilized this technique to study sensory processing by "small cells" in the anterior exterolateral nucleus (ELa) of weakly electric mormyrid fish. "Small cells" are hypothesized to be time comparator neurons important for detecting submillisecond differences in the arrival times of presynaptic spikes(13). However, anatomical features such as dense myelin, engulfing synapses, and small cell bodies have made it extremely difficult to record from these cells using traditional methods(11, 14). Here we demonstrate that our novel method selectively labels these cells in 28% of preparations, allowing for reliable, robust recordings and characterization of responses to electrosensory stimulation.