The behavioural literature on representativeness is able to provide general rules for selecting representative measurement systems. For example, using shorter intervals between observations will, in general, produce more representative samples. However, applied behaviour analysis would benefit from a data-based method of choosing a measurement system based on estimated duration of behaviour, desired representativeness (i.e., acceptable error), and the parameters of a measurement system likely to be practical in a clinical setting. Computer simulations were used to evaluate the effect of varying overall and bout durations on the representativeness of samples extracted using simulated momentary time sampling (MTS). A set of decision rules, in the form of 3D graphs, from which practitioners could select parameters of MTS, based on estimates of the dimensions of the behaviour (bout duration and overall duration) and acceptable error, were developed. The decision rules were applied to whole-week data sets obtained in a prior study to test their use with naturalistic data.