1 The methods commonly used to estimate stem turnover rates (i.e. mortality and recruitment) in species rich tropical forests suffer from a previously unrecognized artefact. The estimated rate is not independent of the census period. 2 An average rate estimate will decrease with time if the sample population cannot be characterized as homogeneous. This artefact may have considerable significance for comparisons between permanent plot studies that have used different census periods. 3 We present a theoretical consideration of this census effect. The artefact will be severe when a fraction of the population has a very much higher mortality rate than the average. 4 Using a simple formulation we provide a mathematical proof that rate estimates will decline with increasing census periods for all but perfectly uniform populations. 5 The phenomenon of apparent rate decrease may be used to provide ecologically significant information about the diversity and dynamics of the population as it is related to the variance of life expectancies within the sample. 6 Such an artefact complicates evaluation of change over nonstandard time intervals and requires careful and detailed attention. Similar problems will effect any estimation procedure which cannot account for all the rate variation within a study population. Recognition of the problem is a necessary first step.