The complementary incompatibility system, characterized by co-operative control of a single S specificity by alleles of two or more distinct S genes, has raised interesting questions regarding the origin and evolutionary significance of this system. What were the factors which led to the appearance of the complementary system of self-incompatibility. Is complementary incompatibility a primary or secondary development?Lundqvist and Γsterbye have suggested that the fundamental characteristic of this system - lack of dominance and competitive interactions between alleles of the same and different series - developed once, early in the evolution of angiosperms, at a stage when self-incompatibility was controlled polygenically. In one line of development, where two or more such incompatibility genes were strengthened by specific modifiers at the expense of the rest, co-operation among loci was favoured to promote increased interplant compatibility in the population. In this evolutionary line, allelic or intergenic interactions were excluded. In the other line of development, where only one incompatibility locus was strengthened, there was no need for such selection and alleles of this locus retained the property of allelic interaction in the pollen.In this article an alternative hypothesis has been proposed for the origin of complementary incompatibility. It is suggested that this type of incompatibility system, conforming with generally held views on the evolution of self-incompatibility systems, developed secondarily, and polyphyletically, after the breakdown of the original one-locus, multiallelic, gametophytic system. In the re-revolution of self-incompatibility through introgression with a related self-incompatible taxon, the essential action of the presumed physiologically integrated self-compatible complex led to the exclusion of allelic or intergenic interaction as a prerequisite for evolution of complementary control. According to this hypothesis, breakdown of the original self-incompatibility and re-evolution of self-incompatibility, in the manner suggested above, could have occurred many times in the evolution of angiosperms and such systems might therefore be expected to occur scattered among different phylogenetic lines.