ORIGINAL ARTICLE An improved method for the identification of areas of endemism using species co-occurrences Sinos Giokas* and Spyros Sfenthourakis INTRODUCTION Traditionally, areas of endemism (AoE) are considered to be the potentially fundamental units in historical biogeography and conservation biology (Crisci et al., 2003). However, there is no general consensus regarding their definition and, consequently, on the analytical methods for their delineation. According to Linder (2001), an AoE should have certain fundamental features: (1) a single history usually resulting from vicariance events (with the notable exception of oceanic islands where biotas are shaped by dispersal), (2) a size smaller than the entire study area, (3) no geographical overlap with other AoE, (4) host at least two taxa with ranges restricted to it, and (5), if possible, the ranges of the species endemic to the AoE should be maximally congruent. The latter prerequisite lies in the core of the debate. Linder (2001, p. 893), for example, stated that AoE ‘… are delimited by the congruent distribution of at least two species of restricted range’, whereas at the other extreme, Harold & Mooi (1994, p. 262) defined them as ‘…geographical region(s) comprising the distribution of two or more taxa that exhibit a phylogenetic and distributional congruence and having their relatives occurring in other such-defined regions’. The major difference between these two opposing definitions is that the latter does not Section of Animal Biology, Department of Biology, University of Patras, Patras, Greece *Correspondence: Sinos Giokas, Section of Animal Biology, Department of Biology, School of Natural Sciences, University of Patras, GR-26500 Patras, Greece. E-mail: sinosg@upatras.gr ABSTRACT Aim To develop a simple method that (1) combines the notions of biotic elements (groups of taxa with ranges significantly more similar to each other than to the ranges of other taxa) and of areas of endemism (AoE, areas of non-random distributional congruence among taxa), and (2) overcomes the constraints of a previously suggested null model-based method that cannot deal with disjunctions and is strictly grid-dependent. Location We used test data sets from southern Africa and Crete. Methods First, we used a null-model approach to detect pairs of species that have a significant degree of co-occurrence, in order to determine biotic elements. Subsequently, we used a parsimony analysis of endemicity to delineate candidate AoE, and multivariate analysis to define groups of biotic elements on the basis of species interactions (co-occurrence, mutual exclusion, neutral) using only the species detected in the previous step. We applied this method to the well known data set for Sciobius in southern Africa, as well as to endemic invertebrates of Crete (Greece), in order to evaluate its performance. Results Our results are very similar to those of previous analyses, and produce meaningful delineation of AoE and biotic elements in both data sets. The method is flexible regarding null models and significance levels, and eliminates noise in the data. Main conclusions We offer a simple method that provides reasonable identification of both biotic elements and AoE, produces good-fit statistics, reduces uninformative or junk output, and reduces computational time. Keywords Areas of endemism, biotic elements, co-occurrence, Crete, historical biogeog- raphy, null models, PAE, Sciobius. Journal of Biogeography (J. Biogeogr.) (2008) 35, 893–902 ª 2007 The Authors www.blackwellpublishing.com/jbi 893 Journal compilation ª 2007 Blackwell Publishing Ltd doi:10.1111/j.1365-2699.2007.01840.x