The detection of non-random patterns of distribution of species along a gradient.

A simple method for detecting non-random patterns of distribution of the boundaries of species is described. The method uses transects running across a community, where the number of upper and lower boundaries of species in each quadrat is recorded. The expected number of quadrats containing one or more boundaries can be calculated from the binomial distribution. The mean deviation of observed from expected number of such quadrats, for a set of transects, can be tested for departures from zero. Significant departures greater than zero indicate regular dispersion of boundaries. A mean deviation significantly less than zero indicates clustering of the boundaries. The method is unbiased and thus corrects previously published methods.