EXTINCTION AND THE NATURE OF THE METAPOPULATION - A CASE SYSTEM

Data on extinction for organisms occupying an archipelago of over 100 islands in the Bahamas is summarized. Lizards have much lower extinction rates than spiders, even for the same population sizes. Spiders naturally and experimentally become extinct more frequently in the presence than absence of lizards; islands with lizards have fewer spider species than no-lizard islands. The larger the island area, and (for no-lizard islands) the nearer to a potential source, the lower the extinction rate of spider species; both can be related to a tendency for smaller populations to become extinct more readily. Within particular spider species, population time series with higher variabilities in population size tend to have lower extinction rates. Mean population size of a series is positively correlated with temporal variability and negatively correlated with extinction rate; even when these correlations are taken into account in multiple regression, however, a negative relation of temporal variability and extinction rate often exists. A conceptual scheme including immigration and catastrophic mortality, in addition to normal stochastic birth and death rates, appears to accord at least qualitatively with these results. Insular spider populations in the Bahamas appear much closer to a BOORMAN-LEVITT metapopulation than to a LEVINS metapopulation, a similarity that seems to hold for organisms in general. A scheme is presented that favors such an outcome in part as an inevitable consequence of observational practicalities over time and space, as well as of a substantial degree of variation in isolate area.