Clustering of sites was investigated using Ghat. Ghat measures the distance from each point to its nearest neighbour (Diggle 1983). This can be compared with a Monte Carlo simulation of random distributions which are based on the Poisson distribution. The simulated distributions must have the same boundary and number of points as the distribution in question. The number of iterations can be increased to enhance the confidence of the test; 100 iterations were used here. The highly non-random distribution of the tertres tumulaires is shown clearly in Figure 11. It is outside the expected envelope for a random distribution. Tertres are strongly, and statistically significantly, clustered, with most sites very close to another site.