In species with sporophytic SI systems, such as species from the plant families Brassicaceae and Asteraceae, dominance interactions among certain alleles in the expression of the pollen and pistil phenotypes are known (Bateman 1952; Kowyama et al. 1994; Hatakeyama et al. 1998; Kusaba et al. 2002). In models that incorporate dominance, recessive alleles are expected to occur at higher equilibrium frequency than dominant alleles, the so-called ‘recessive’ effect (Cope 1962; Imrie et al. 1972). This arises from a general property of frequency-dependent selection models: at equilibrium all SI phenotypes are expected to have equal frequencies (‘isoplethy’, Finney 1952). Under serial dominance of allele specificities, a given SI phenotype comprises fewer different SI genotypes as recessivity increases, until finally the most recessive phenotypic class is represented by a single allele in a homozygous state. Therefore, isoplethy implies that allelic frequencies should increase with recessivity (see Uyenoyama 2000a for a complete coverage of this issue). Although this phenomenon was described long ago (e.g. Bateman 1952), evolutionary models of sporophytic SI systems in finite populations have only recently been explored in detail (Schierup et al. 1997; Uyenoyama 2000c). In a sporophytic system with a linear dominance hierarchy in both the pistil and pollen phenotypes, dominance/recessivity is expected to be a major determinant of an allele's evolutionary dynamics. On the one hand, the probability that new alleles that have arisen by mutation will become incorporated in a population increases with their dominance. On the other hand, the higher equilibrium frequency of recessive alleles reduces the probability of loss by genetic drift (Schierup et al. 1997). Altogether, dominant alleles are simultaneously more likely to become incorporated and also more likely to become lost from the population. Dominant alleles are thus expected to have a substantially higher turnover rate (shorter lifespan) than recessive alleles. This conclusion, however, does not hold under all possible dominance relationships. Under a model with full co-dominance in the pistil and a linear dominance hierarchy in the pollen, for instance, very different evolutionary dynamics arise (Schierup et al. 1997), whereby recessive alleles are more easily lost by drift than are dominant alleles. This situation may lead to a process of continuous evolution towards ever increasing dominance, as new mutant alleles will spread most readily if they are more dominant than the present alleles, such that all previous alleles will shift towards lower relative dominance levels. As alleles fall down this ‘relative dominance ladder’ during their lifespan, alleles of increasing age and recessivity tend to be preferentially lost from the population when they finally become the most recessive ones. Uyenoyama (2000c) studied a model with strict co-dominance in the pistil and two levels of dominance in pollen, with co-dominance between different alleles at the same level. This model was intended to mimic the situation described in self-incompatible Brassica species. Such a model does not allow evolution towards ever increasing dominance. She showed that several alleles from the dominant class were then expected to co-occur, all experiencing equivalent and regular turnover, while a single allele was expected to occur in the recessive class in most situations. This allele would impede the spread of other recessive alleles, so it is expected to have an extremely long lifespan. Hence, in this model, the equivalence assumption may hold only among alleles from the dominant class, not overall.
In spite of this lack of equivalence, expected allelic genealogies under these models are very similar to those for gametophytic SI systems (Schierup et al. 1998). A noteworthy difference, however, was expected for genealogies of gene copies of recessive alleles. Because recessive alleles have higher frequency and longer lifespan than dominant alleles, higher divergence among gene copies may be expected relative to dominant alleles. Again, this results in an expected difference among alleles in the level of nucleotide divergence among gene copies of the same allelic specificity.
Empirical studies aimed at determining dominance relationships between S-alleles in sporophytic systems have remained scarce. Evidence from several species of Brassicaceae indicates that co-dominant interactions between alleles are frequent, if not predominant, in the pollen or the stigma, that dominant/recessive interactions are expressed more often in the pollen than in the stigma, and that dominance in pollen and stigma are not necessarily correlated (Sampson 1964; Stevens & Kay 1989; Hatakeyama et al. 1998; Mable et al. 2003). Strict co-dominance in the stigma and either co-dominance or dominance in the pollen expression of incompatibility was reported for the species Corylus avellana (Betulaceae; Mehlenbacher 1997). In contrast, in species from Asteraceae (Samaha & Boyle 1989; Brennan et al. 2002) and from Convolvulaceae (Kowyama et al. 1994) dominance/recessive interactions were found between most alleles and were mostly consistent between pollen and stigma. Evidence for a higher frequency of the more recessive alleles has been found in several studies (Sampson 1964; Stevens & Kay 1989; Kowyama et al. 1994; Mable et al. 2003). In species of the genus Brassica, recessive and dominant alleles were found to constitute two distinct monophyletic groups of ancient origin based on their nucleotide sequence (Uyenoyama 1995), whereas no such evidence has been found in Arabidopsis lyrata (Mable et al. 2003).
As far as the limited number of empirical studies allows one to conclude, they roughly confirm theoretical predictions about the effect of dominance. The technical difficulty of precisely estimating dominance relationships seems to be the major impediment to testing the numerous theoretical predictions now available.