OIKOS 88: 641 – 651. Copenhagen 2000 A microsatellite analysis of natterjack toad, Bufo calamita, metapopulations G. Rowe, T. J. C. Beebee and T. Burke Rowe, G., Beebee, T. J. C. and Burke, T. 2000. A microsatellite analysis of natterjack toad, Bufo calamita, metapopulations. – Oikos 88: 641 – 651. Although it is widely recognised that spatial subdivision of populations is common in nature, there is no consensus as to how metapopulation dynamics affect genetic diversity. We investigated the genetic differentiation of natterjack toads, Bufo calamita, in three regions of Britain where habitat continuity indicated the likely occurrence of extensive metapopulations. Our intention was to determine whether genetic analysis supported the existence of metapopulation structures, if so of what type, and to identify barriers to migration between subpopulations. Allele frequencies were determined across eight polymorphic microsatellite loci for a total of 24 toad subpopulations at three separate sites. Genetic differentiation was assessed using five measures of genetic distance, notably F ST , R ST , Nei’s standard distance D s ,  2 and the Cavalli-Sforza chord distance D c . B. calamita exhibited small but significant levels of genetic differentiation between subpopulations in all three study areas, and genetic and geographic distance correlations indicated isolation-by-distance effects in all three cases. The effects on correlation strengths of compensation for positive (sea, rivers, urban development) and negative (pond clusters) barriers to toad migration between the subpopulations in each area were also determined. D c , a measure which assumes that differentiation is caused by drift with negligible mutation effect, yielded the most plausible interpretation of metapopulation structures. Overall the patterns of genetic variation suggested the existence of a mixed metapopulation model for this species, with high levels of gene flow compatible with one version of the classical model but often supported by particularly stable subpopulations as in the mainland- island model. G. Rowe and T. J. C. Beebee (correspondence), School of Biology, Uni. of Sussex, Falmer, Brighton, UK BN19QG (t.j.c.beebee@sussex.ac.uk).– T. Burke, Dept of Animal and Plant Sciences, Uni. of Sheffield, Sheffield, UK S10 2TN. The study of metapopulation structures and dynamics has become a central theme in population ecology since the realisation that fragmented distributions are more common than spatially continuous ones (e.g. Levins 1970, Hanski and Gilpin 1991, 1996). However, meta- populations can take many forms which may have different genetic and evolutionary consequences de- pending on how interconnected the demes are, and how frequently and evenly extinction events occur (e.g. Mc- Cauley 1991, Harrison and Hastings 1996). The classi- cal metapopulation model invokes relatively high subpopulation turnover and recolonisation rates with equal probabilities across demes, but more realistic alternatives have been developed in light of increasing evidence from empirical studies. Thus the existence of source and sink subpopulations, with the former having a lower probability of extinction than the latter, forms the basis of a mainland-island model; and migration may be sufficiently frequent that local extinction is rare and what amounts to a single ‘‘patchy’’ population persists. These models make various predictions with respect to genetic differentiation and gene flow between subpopulations, and thus may have different evolution- ary implications. In the classical model, differentiation depends critically on the mode of recolonisation but can be high if colonists are rare and come only from a Accepted 5 July 1999 Copyright © OIKOS 2000 ISSN 0030-1299 Printed in Ireland – all rights reserved OIKOS 88:3 (2000) 641