Island distribution patterns of species in thirty-one families of West Indian land birds are analysed. Species-area regressions of individual families are ambiguous because the slopes of these regressions are a function of family size. Rarefaction gives the result of all islands supporting close to the expected number of families based on a random draw, although small islands tend to be family-poor and large islands family-rich. The expected number of species in each family was generated from a hypergeometric model, sampling without replacement from the total list of species. With this model no island has fewer species in any family than expected by chance, although some islands have more species than expected. Confamilial sympatry in the Columbidae and Mimidae is consistently greater than expected with this technique. Some assumptions of a random draw of species from an equivalent species pool for all islands are biologically unrealistic. However, the hypergeometric model avoids many biases inherent in species-area regressions of families and is preferable to regression analysis for preliminary identification of unusual family distributions.