When the choice of outgroup species involves groups that are polymorphic for the character of interest (e.g., as is true of the presence or absence of distyly in the Boraginaceae, see below), it is possible that the use of several outgroups may help to resolve the ancestral character state in the ingroup (Maddison et al.
To explore the evolution of distyly in Amsinckia, the mating system character states were mapped onto the cpDNA phylogeny under a number of different assumptions about the evolution of distyly as a character and of the primitive character state of the mating system in the genus.
Apart from Amsinckia, many genera in the Boraginaceae contain distylous members, for example, Anchusa, Arnebia, Cordia, Cryptantha, Echioides, Lithodora, and Lithospermum (Ganders 1979), but distyly is unknown in other Boraginaceous genera.
Because polytomies interfere with the reconstruction of character evolution (Maddison 1989), it was necessary to explore the different possible resolutions of the polytomies with respect to the evolutionary history of distyly. This was done by randomly resolving the polytomies 1000 times using the algorithm described by Maddison (1989), implemented in MacClade, vers.
The Evolution and Breakdown of Distyly in Amsinckia
Figure 4 shows phylogenies for the combined phylogenetic analysis of the diploid and polyploid species, representing the most parsimonious reconstructions of the evolution of the mating system in Amsinckia given the cpDNA phylogeny, resolution of the polytomous branches, and four sets of assumptions about distyly as a character.
Figure 5 shows the most parsimonious reconstructions of the evolution of the mating system when the diploid species of Amsinckia are analyzed separately, given the cpDNA phylogeny obtained for them, resolution of the single polytomous branch, and the four sets of assumptions about distyly discussed above.
Viewed in the context of what is known about the mating system of other Amsinckia species (Table 1), the overall results of the mating system studies reported here fit the pattern of a close association of outcrossing with distyly, and of selfing with homostyly.
It is, however, possible to cite a number of arguments in support of the notion that heterostyly has broken down recently and repeatedly in the genus, as opposed to a pattern of repeated evolution of distyly from homostylous ancestors.