Areas of endemism are defined by the non-random distributional congruence of two or more taxa. They can arise by similar ecological and/or historical responses of the taxa or because the scale of mapping is unable to resolve multiple ecological and/or historical processes of relevance in a geographic area. Until now, two approaches existed for determining areas of endemism: (i) simple, non-algorithmic mapping of distributions in the search for significant overlap and (ii) algorithmic Parsimony Analysis of Endemism (PAE; parsimony analysis of a presence/absence matrix). Here we propose a third method, involving the pairwise comparison of taxon distributions in a presence/absence matrix and leading to a test of statistical significance for the departure of distributional congruence from that expected by chance. This approach has an advantage over the two other methods in excluding random distributional congruence (e.g., that which might arise when comparing the overlap of two taxa, each with a range of four grid cells, in a total habitable area of eight grid cells). There is also considerable flexibility introduced in starting with pairwise comparisons, rather than attempting to immediately summarize relationships among all grid cells in a bifurcating tree (the PAE approach). The method is implemented in a new program, AREND (AReas of Endemism defined with Null Distributions), written in the Python programming language. The results of an AREND analysis on a presence/absence matrix of 67 banksia taxa in Australia's Southwestern Botanical Province will also be presented, as will future refinements of the method.

Key words: areas of endemism, Australia, Banksia, null distributions, Proteaceae