Our understanding of the perilymph kinetics of drugs depends largely on data obtained by the analysis of perilymph samples. Although a number of studies have demonstrated qualitatively that perilymph samples may be contaminated by cerebrospinal fluid (CSF), and some investigations adopt specific methods to minimize CSF contamination of their samples, many other studies fail to consider the influence of this potential artifact on their measurements. In the present study we have attempted to quantify the degree of CSF contamination of perilymph samples taken from the basal turn of the guinea pig cochlea using the ionic marker trimethylphenylammonium (TMPA). TMPA solution was irrigated across the round window membrane while a TMPA-selective electrode sealed into the perilymphatic space continuously monitored perilymph TMPA concentration. After a period of TMPA loading, a perilymph sample was aspirated and its TMPA content determined. Differences between the sample concentration and the measured TMPA time course during perilymph loading and sampling were analyzed using a finite element computer model for simulation of solute movements in the inner ear fluids. The experimental results were consistent with the aspirated fluid sample from the cochlea being replaced by CSF drawn into the perilymphatic space through the cochlear aqueduct. The dependence of perilymph sample purity on the location of sampling and on the volume withdrawn was quantified. These relationships are of value in the design and interpretation of experiments that utilize perilymph sampling.