A method of multichannel analysis is proposed in which the mean rate in each channel is presented directly as counts per unit time interval, independent of the total experimental time. The method may also be applied to the direct representation of relative probability. Essentially the total storage capacity of a mean rate analyser is divided into two independent halves. One half contains the latest assessment of the mean rate per channel and os coupled to the display system, while the other half contains the counts currently being collected and has random access from the data input. Because the two halves have different functions and rates of working it is possible to combine different storage media to provide optimum operation at minimum cost. In computer applications the method reduces both program complexity and execution time on machines having short word lengths. It also permits half the total bit capacity required to be provided by a suitable drum store. The display provided is essentially static in amplitude but has increasing precision as the experiment continues.