Stochastic Loss of Style Morphs From Populations of Tristylouslythrum Salicariaanddecodon Verticillatus(lythraceae)

Despite the theoretical importance of stochastic processes in evolution there have been few empirical studies of the interaction between genetic drift and selection on the maintenance of polymorphisms in plant populations. We used computer models to investigate the interaction between drift and frequency-dependent selection in affecting style morph frequencies in populations of tristylous species. Drift produces a distinct pattern of morph frequency variation involving: 1) the loss of the S morph and, to a lesser extent, the M morph; 2) no consistent bias in frequencies within populations; 3) a restricted pattern of variation involving a deficiency of one morph and equal excesses of the other two. Morph frequencies were surveyed in 137 populations of Lythrum salicaria from both its native range in Europe (N = 35) and recent adventive range in Ontario (N = 102), and 133 populations of Decodon verticillatus from four regions in eastern North America with different glacial histories to assess these theoretical predictions. There was a negative relationship between morph loss and population size in both species; the relationship was weaker in D. verticillatus than in L. salicaria. Morph loss was more frequent in the adventive than native range of L. salicaria, and in populations of D. verticillatus from glaciated northern regions compared with the unglaciated southern portion of its range. Simulations incorporating variation in life history, regeneration strategy and mating patterns revealed that the degree of morph loss was strongly influenced by year to year survival, clonal propagation, self-fertilization and departures from disassortative mating. Comparing the pattern of morph frequency variation between species supported these predictions. Morph loss was lower in self-incompatible L. salicaria (0% in Europe; 23% in Ontario), which reproduces through seed compared to self-compatible, clonal D. verticillatus (52%). A stochastic model provides the most parsimonious explanation for observed patterns of morph frequency variation in both species.