Patch clamp records with small numbers of channels can be modelled effectively in a Bayesian framework. A statistical model was formulated in which the underlying signal is piecewise constant, any conductance level is permitted and there are no long-range constraints. (The levels used to fit two sections of data are independent of each other if the sections are sufficiently separated in time.) The posterior probability of a chosen set of conductance levels and transition times can be calculated from the values of these parameters and the experimental data. A Markov chain Monte Carlo procedure [StarkJA97], in which the number of transitions is allowed to vary, is used to draw a sequence of fits from their posterior distribution.

The method is applied to filtered data contaminated with coloured noise and is illustrated with results from experimental data.