Founder effects and silvereyes.

P of variation in nature have played a large role in the development of explanations for biological richness at the species level. One such inf luential pattern has been the morphological distinctiveness of small populations on islands at the periphery of a large continental land mass (1). If the island differs from the mainland in habitat or composition of the biological community, the morphological differences often can be explained in terms of adaptation by natural selection (2). But, if the environments are more or less the same, then an explanation has to be sought elsewhere. Fifty years ago, Ernst Mayr (3) suggested that the key to island evolution lies in the circumstances and immediate consequences of colonization. The model he proposed, called the ‘‘founder effects model,’’ postulated major changes in the genetic constitution of a newly established population that were set in motion by the reduced genetic variation carried by the few colonizing individuals in the founding event (4). The model was a bold explanation, because nothing was known of the genetic basis of the morphological traits. It has been challenged repeatedly on theoretical grounds (refs. 5–7; but see refs. 8 and 9) and has fared little better empirically (10). The preferred method of testing has involved creating in the laborator y new populations with a few individuals of small organisms with short generation times (11–13). These experimental studies have yielded some (11, 13) but generally little or no support (10, 12). Adopting a different tactic, Sonya Clegg et al. (14) have returned to the source of the problem, bird populations on islands, and in this issue of PNAS, they report the results of some novel tests. Silvereye is not the name of a Drosophila mutant but the name of a small, warbler-like bird that has colonized several islands from the mainland of Australia. Four colonizations occurred in historical times—from 1830 onwards— and can be dated precisely. Three others took place much earlier, at times that have been estimated at 3,000 – 4,000 years for the most recent and more than a million years for the most ancient. Thus, the sequence of colonizations has been reliably established; taking advantage of it, Clegg et al. (14) have asked whether the genetic variation both within and between populations ref lect substantial changes at the time of colonization, or whether changes have simply accumulated gradually over time. How could one distinguish between them? It can be done as follows. After the establishment of a new population, alleles at selectively neutral loci will be lost by random genetic drift and gained by mutation. Drift can be expected to predominate in small populations, at least for some time until a new drift-mutation equilibrium is attained, and populations will gradually diverge. If drift and mutation are all that happen, the population does not rapidly increase in size, and gene flow from neighboring populations is negligible or nonexistent, then the magnitude of withinand between-population changes will be a function of time. In contrast, if founder events have occurred, the effects may be so strong as to overwhelm the effects of longterm drift yet vary among populations in a manner unrelated to their time of origin. Therefore, against the background of long-term drift, a signal of founder effects may stand out as sharp reductions in genetic variation within populations and large differences between neighboring populations. This pattern is what Clegg et al. (14) looked for. They chose to work with allelic variation at six microsatellite loci because the alleles are likely to be selectively neutral (unless the loci happen to be closely linked to loci subject to selection). Their analyses failed to detect the signal of founder effects at any of the steps in the sequence of recent colonizations. Nevertheless, they found that allelic diversity gradually declined with repeated colonizations of new islands. The individual reductions are small, but the cumulative changes are large. From first to last in the sequence of recent colonizations, the mean number of alleles per locus dropped by almost half. Because the last population in the sequence is the youngest, one cannot explain this result by long-term genetic drift. Instead, the pattern seems to ref lect a small loss of alleles at each colonization, although hardly on the scale envisaged in the original formulation of the founder effects model; it is certainly not enough to kick-start a genetic revolution (1, 3). Expected heterozygosity does not follow this pattern, as would be expected because it is less sensitive than allelic diversity to sampling effects (15). It remains roughly the same among the new populations regardless of their age but declines sharply as one proceeds from young to old ones, which is inconsistent with the pattern the authors expected from the founder effects model (3). Agedependence hints at multiple bottlenecks over time. An additional factor (not analyzed in the paper) that might contribute to this pattern through its effect on demography and long-term effective population size is island size. Regardless of these causes, the conclusion drawn by the authors is that the patterns of variation at microsatellite loci principally reflect the effects of long-term genetic drift. Consistent with this conclusion, the oldest populations differ genetically the most from the presumed original source, and the newest populations differ the least. Other studies of bird populations have drawn similar conclusions (16–19) without being able to test for founder effects. An important feature of this paper is the use of a model to supplement the empirical analysis by simulating colonization with and without genetic bottlenecks. By using a Bayesian approach, the authors sought to determine the sizes of the effective founder flocks, i.e., the number of individuals who bred. If the colonizing flocks were small, then founder effects may