In human palaeontology, samples are generally collections of objects gathered according to morphological criteria and their statistical analysis usually calls for caution because of the relatively small number of objects. Testing normality for such samples is a critical problem. The procedure suggested by Filliben, however, is particularly suitable for small to moderate sample sizes. Moreover, it provides a graphical presentation of the data that permits to locate each object within the distribution range. We briefly review the principle of the method and give a computer program to carry out the necessary computations and plot. Finally, we illustrate the method by studying the distribution of the maximum length and the maximum breadth for a sample of skulls of the Upper Palaeolithic.