Gene flow and life history patterns.

Dispersal distances overestimate the gene-flow scale l (the square-root of the mean squared distance travelled from birth to reproduction) when egg laying is concentrated early in dispersal and when there is mortality during dispersal. If egg laying follows a square-root normal distribution in time, as it does in several Drosophila species, then l is reduced to about 0.6 of that estimated from dispersal alone, unless egg laying is concentrated in a very brief period. If mortality is such that the half-life is earlier than the mean fecundity day, then l will be reduced still more relative to the dispersal estimate, and will be very sensitive to small changes in mortality. Overestimating l yields overestimates of the amount of selection needed to maintain geographic differences in gene frequencies. If mortality increases in suboptimal habitats, then the neighborhood size will be smaller in those areas, because increased mortality decreases l. This means that l is smallest, allowing the greatest differentiation and genetic innovation, precisely where it is most needed. This lends support to Wright's shifting balance hypothesis. If l is adjusted to local conditions, then we do not necessarily expect a positive relationship between environmental and genetic heterogeneity. Data from Drosophila pseudoobscura are used to make the models realistic, and it is shown that l depends on the distances among traps or breeding sites. It is therefore essential to know the geometry of breeding sites and the life history parameters to estimate l.