A Monte Carlo method based on Ripley's K function -a cumulative function related to the number of plants encountered at different distances from other plants -is used to test the null hypothesis of random distribution of shrub clumps in a desert dwarf shrub community in Namaqualand, South Africa, where Psilocaulon arenosum is the dominant shrub. The method takes into account the apparent regularity of pattern caused by the finite size (up to 2 m diameter) of the clumps. It is shown that the clump centres are significantly aggregated (compared to random expectation) at distances on the order of 1 m. Such aggregation is expected, as a simple result of regeneration near to seed sources, if the time between catastrophic droughts is short in relation to the time required for development of a non-aggregated or regular pattern determined by moisture competition. No significant regularity was detected at distances of 3 m or less. One subplot showed regularity above 3 m, but this pattern was not shown by the other subplot and may not be a competition effect. These results support a hypothesis of aggregation caused by regeneration pattern decaying slowly toward randomness as larger individuals compete.