Before new drugs for the treatment of inner ear disorders can be studied in controlled clinical trials, it is important that their pharmacokinetics be established in inner ear fluids. Microdialysis allows drug levels to be measured in perilymph without the volume disturbances and potential cerebrospinal fluid contamination associated with fluid sampling. The aims of this study were to show: (i) that despite low recovery rates from miniature dialysis probes, significant amounts of drug are removed from small fluid compartments, (ii) that dialysis sampling artifacts can be accounted for using computer simulations and (iii) that microdialysis allows quantification of the entry rates through the round window membrane (RWM) into scala tympani (ST). Initial experiments used microdialysis probes in small compartments in vitro containing sodium fluorescein. Stable concentrations were observed in large compartments (1000 μl) but significant concentration declines were observed in smaller compartments (100, 10 and 5.6 μl) comparable to the size of the inner ear. Computer simulations of these experiments closely approximated the experimental data. In in vivo experiments, sodium fluorescein 10 mg/ml and dexamethasone-dihydrogen-phosphate disodium salt 8 mg/ml were simultaneously applied to the RWM of guinea pigs. Perilymph concentration in the basal turn of ST was monitored using microdialysis. The fluorescein concentration reached after 200 min application (585 ± 527 μg/ml) was approximately twice that of dexamethasone phosphate (291 ± 369 μg/ml). Substantial variation in concentrations was found between animals by approximately a factor of 34 for fluorescein and at least 41 for dexamethasone phosphate. This is, to a large extent, thought to be the result of the RWM permeability varying in different animals. It was not caused by substance analysis variations, because two different analytic methods were used and the concentration ratio between the two substances remained nearly constant across the experiments and because differences were apparent for the repeated samples obtained in each animal. Interpretation of the results using computer simulations allowed RWM permeability to be quantified. It also demonstrated, however, that cochlear clearance values could not be reliably obtained with microdialysis because of the significant contribution of dialysis to clearance. The observed interanimal variation, e.g., in RWM permeability, is likely to be clinically relevant to the local application of drugs in patients.