Quantitative predictions of the course of acetylcholinesterase (AChE) activity, following interference of inhibitors and reactivators, are usually obscured by the time-dependent changes of all reaction partners. To mimic these dynamics we developed an in vitro model. Immobilized human erythrocyte ghosts in a bioreactor were continuously perfused while AChE activity was monitored by a modified Ellman method. The perfusion system consisted of two HPLC pumps with integrated quaternary low-pressure gradient formers that were programmed by a computer using commercial HPLC software. The combined eluates passed a particle filter (Millex-GS, 0.22 microm) containing a thin layer of erythrocytes that was immersed in a temperature-controlled water bath. The effluent passed a flow cell in a UV-vis detector, the signal of which was digitized, written to disc and calculated with curve fitting programs. AChE activity decreased by 3.4% within 2.5 h. The day-to-day variation of the freshly prepared bioreactor using the same enzyme source was +/-3.3%. Residual activity of 0.2% marked the limit of quantification. Following perfusion with paraoxon, pseudo first-order rate constants of inhibition were established that did not differ from results obtained in conventional assays. The same holds true for reactivation with obidoxime. The set-up presented allows freely programmable time-dependent changes of up to eight solvents to mimic pharmacokinetic profiles without accumulation of products. Due to some hysteresis in the system, reaction half-lives should be >3 min and concentration changes in critical compounds should exceed half-lives of 5 min. Otherwise, the system offers much flexibility and operates with high precision.