Cell-attached patch-clamp recording has been established as a major technique for investigating ion channel behaviour in a physiological setting, despite the problems which arise in analysing records containing more than one type of unitary current. To circumvent these problems, single-channel amplitude-based assignment of discrete single-channel events to different channel types becomes increasingly necessary. Surprisingly, a systematic evaluation of the validity of this method in determining single-channel parameters has not been performed to date. Using computer-simulated single-channel traces, and recordings from a biological preparation containing well-characterized ion channels (N1E-115 neuroblastoma cells), we have explored the accuracy by which amplitude-based separation recovers ion-channel parameters. Determination of gating kinetics after separation revealed that even a very small contamination in the selected population yields additional time constants in the probability density functions. Therefore, our results demonstrate that, whereas the use of amplitude-based separation is straightforward for determining slope conductance and reversal potential, it is prone to incorporate errors in establishing gating kinetics. Ways of identifying such errors are described.