Suction electrodes are widely used for recording compound action potentials (CAPs) from peripheral nerves or central tracts. Unfortunately, the recordings obtained with suction electrodes often vary over time, making quantitative measurement of CAP amplitude difficult. We developed an equivalent electrical model which predicts that the magnitude of a recorded potential will be linearly related to the resistance of the electrode with a nerve inserted. Mathematical procedures were developed that allow correction of virtually all variability inherent in this type of recording; this variability may arise from resistance drift, variable stimulus artifact, or potentials generated as a result of the current of injury. The validity of the theoretical analysis was confirmed experimentally using rat optic nerves. The magnitude of the CAP and electrode resistance varied spontaneously by as much as 100% over time, due to changes in electrode resistance and size of the stimulus artifact. Because the CAP was linearly related to resistance, it was therefore best quantified by the slope computed from this relationship. The stimulus artifact, unlike the CAP itself, was shown to be independent of recording electrode resistance and therefore only resulted in a variable offset to the area vs resistance linear relationship; the slope of this relationship was unaffected. In the absence of stimulation, a steady negative DC potential was recorded from the optic nerve, which was greatest immediately after dissection, and was also a linear function of electrode resistance. In contrast to CAP amplitude, the latencies of the component peaks within the CAP were not significantly altered by changes in electrode resistance. The experimental results confirmed the validity of the electrical model and demonstrated that suction electrodes can be a very reliable and quantitative recording method if the signals are properly corrected.