The conductance, number, and mean open time of ion channels can be estimated from fluctuations in membrane current. To examine potential errors associated with fluctuation analysis, we simulated ensemble currents and estimated single channel properties. The number (N) and amplitude (i) of the underlying single channels were estimated using nonstationary fluctuation analysis, while mean open time was estimated using covariance and spectral analysis. Both excessive filtering and the analysis of segments of current that were too brief led to underestimates of i and overestimates of N. Setting the low-pass cut-off frequency of the filter to greater than five times the inverse of the effective mean channel open time (burst duration) and analyzing segments of current that were at least 80 times the effective mean channel open time reduced the errors to < 2%. With excessive filtering, Butterworth filtering gave up to 10% less error in estimating i and N than Bessel filtering. Estimates of mean open time obtained from the time constant of decay of the covariance, tau obs, at low open probabilities (Po) were much less sensitive to filtering than estimates of i and N. Extrapolating plots of tau obs versus mean current to the ordinate provided a method to estimate mean open time from data obtained at higher Po, where tau obs no longer represents mean open time. Bessel filtering gave the least error when estimating tau obs from the decay of the covariance function, and Butterworth filtering gave the least error when estimating tau obs from spectral density functions.