SUMMARYUsing two dimorphic species with diallelic incompatibility. Primula farinosa \^. and Armeria maritima (Miller) Willd., we tested the hypothesis that an association of seedling performance with seed size, and an inverse association of seed size with seed numher, tnight, contrary to most fitness models, select for plants which set relatively few seeds.In P. farinosa we discovered positive relationships between seed number per eapsule and capsule nutiiber per inflorescence. As capsules are formed in most flowers, scapes with tew flowers set the fewest seeds per capsule. This suggests that conspieuousness to pollinators and/or plant resource play an important role in reproductive success in this species. Seed size and seedling performance were also positively associated.For P. farinosa, which has multi-seeded fruits, we discovered an inverse association between seed size and seed tiumher for capsules containing more than 65 seeds. We suggest that stabilizing selection may occur for flower number per inflorescence in this species, as inflorescences with 8-11 capsules tend to set many seeds which tnay be inefficiently small. This tends to confSrm our hypothesis.However, for A. maritima there is a positive association between seed size and seed number per inflorescence. For such plants with single-seeded fruits there should be directional selection for large flower number per inflorescence, a trend which may have led to the evolution of the capitulum.

Using the outcrossing Primula farinosa and its autogamous selfing relatives P. scotica, P. scandinavica and P. stricta, we compared the fitness of light and heavy seeds. Heavy seeds germinated in greater numbers and more quickly. In competition with seedlings grown from lighter seeds, heavy seeds produced larger rosettes. In P. farinosa such seedlings went on to produce more seeds, and in two populations heavier seeds, than plants from lighter seeds. After transplantation to natural populations, seedlings of P. farinosa derived from heavy seeds produced larger rosettes, more flowers and seeds than those from lighter seeds in certain populations so that seedlings born of heavy seeds were much fitter than seedlings from lighter seeds. Average seed weight varied in inverse proportion to seed number per capsule. The autogamous species produced on average about twice as many seeds per capsule as P. farinosa. In P. scotica and P. stricta this difference appears to be due in part to assured fertilization, but this high fecundity did not cause disadvantageously light seeds. As these species produced fewer capsules per scape, their overall seed production was on average no greater than for P. farinosa. P. farinosa tradedoff fitness between capsules with large seed numbers, which donated more offspring to the next generation, and those with small seed numbers, whose heavy seeds would be more likely to reproduce themselves in the next generation. We conclude that low fecundity in outcrossing species might at times be advantageous.

For a number of generations, we followed the frequency, performance and fecundity of an initial mixture (50 : 50) of triazine resistant (R) and susceptible (S) but otherwise near-isogenic lines of Brassica rapa (for 3 yr) and of Chenopodium album (for 2 yr) in neighbouring experimental garden plots, which differed in aspect (north-facing versus south-facing), shading, and transparent cover from precipitation. Each of eight treatments was replicated four times per species. Seed from each plot was kept separate and sown back in its plot of origin to provide the next generation. For both Brassica and Chenopodium, R frequency changed between generations. In Brassica R frequency declined consistently from an initial 0n5 and, after 2 yr, R plants had disappeared in all replicates in two treatments and averaged 0n126 in the remaining treatments. However, R frequencies in the smaller second generation samples in each year were less consistent. In Chenopodium, R frequency increased consistently from an initial 0n5 in all treatments and replicates during 1995 to average 0n772 on south-facing sites and 0n675 on northfacing sites. However, in 1996 Chenopodium R frequency did not change significantly from that in 1995 for any treatment. Brassica S plants were more fecund that R plants, except in shaded conditions, and south-facing plots produced significantly more S seeds than north-facing plots. Both Brassica plants and plots were more productive in 1996 than in 1995. By contrast, Chenopodium was much less productive in 1996, a season with more climatic extremes between plots than 1995. In 1995, R plants were more fecund than S plants in south-facing plots and in covered conditions, but this difference was not detected in 1996. South-facing Chenopodium plots were more productive than north-facing plots in 1995, but the reverse was the case in 1996. Cover resulted in an increase in seed production in Chenopodium on south-facing plots in 1995, but a decline in 1996. We conclude that selection might act against triazine-resistant plants in B. rapa under the climatic conditions in the UK, but in C. album, which is more susceptible to extreme conditions of light and water stress, triazine-resistant phenotypes might predominate in low stress cool moist conditions. We suggest that a relationship might exist between the overall climatic tolerance of a species, and climate thresholds beyond which herbicide-resistant phenotypes might spread. Thus, in the absence of herbicide, it is not inevitable that the frequency of R phenotypes will decline in field conditions.