Since the pioneer work of Darwin on the evolution of sexual systems in plants (Darwin, 1877a, 1877b) no sexual strategy in recent years has attracted as much attention as dioecy. First, a number of population genetic models were developed in the 1970s to trace the evolution of dioecy via different pathways (Lloyd, 1975, 1976, 1979; Ross, 1970, 1978, 1980, 1982; Charlesworth & Charlesworth, 1978a, 1978b). Almost concurrently, field studies highlighted the ecological consequences of dioecy (Bawa & Opler, 1975; Lloyd & Webb, 1977; Wallace & Rundel, 1979; Meagher, 1980, 1981; Bullock & Bawa, 1981). Then a resurgence of interest in the sexual selection theory led to a search for new selective pressures driving the evolution of dioecy (e.g., Willson, 1979). The finding that dioecy is associated with certain pollination and seed-dispersal syndromes further eroded the traditional view that outcrossing is the main selective force in the evolution of dioecy (Bawa & Opler, 1975; Bawa, 1980a; Givnish, 1980; Beach, 1981; but see Thomson & Barrett, 1981; Lloyd, 1982; for a balanced review, see Charnov, 1982). Here, I briefly consider the major unresolved problems in the evolution of dioecy, including some already discussed at length by the contributors to this symposium. First, a fundamental problem concerns the extent to which sex expression in dioecious species is constant. Freeman et al. (this symposium) document in detail substantial sex reversals in Atriplex canescens. On the other extreme the dioecious lily, Chamaelirium luteum, studied by Meagher (this symposium) exhibits no change in sex expression. Furthermore, the two sexes in C. luteum show remarkable ecological divergence. Sexual dimorphism in many other dioecious species is also pronounced (Lloyd & Webb, 1977; Bawa, 1980b; Bawa et al., 1 982; Bullock & Bawa, 1981; Bullock, 1982; Bullock et al., 1983). If indeed there is no constancy in sex expression, then we need models to explain how sex-linked divergence in morphological, behavioral, physiological, and biochemical traits might have evolved. Freeman et al. (1980, this symposium) mention many other species that presumably change sex, but as pointed out by Lloyd and Bawa (1984), patterns of gender modification in plants are varied and complex. In order to understand the origin of these complex patterns and their adaptive significance, it is necessary to distinguish, for example, extremes such as "sex choosers" (e.g., Arisaema triphyllum) and "sex adjustors" (e.g., many dioecious species, see Lloyd & Bawa, 1984). Only a precise quantitative description of gender may allow the resolution of various patterns of gender modification. For many species that are assumed to change sex, such information is simply not available (Lloyd & Bawa, 1984). Second, the study of evolutionary pathways to dioecism remains an area of major importance. Dioecism has been presumed to have evolved via five distinct routes directly from hermaphroditism and via androdioecy, gynodioecy, monoecy, and heterostyly (Bawa, 1980a; Ross, 1982). It is not known if the ecological pressures favoring the evolution of dioecy are the same in each pathway. However, the population genetic models for almost all pathways assume selective pressure against inbreeding as the major driving force (Lloyd, 1982; Ross, 1982 and references therein). Field studies for specific taxa are badly needed to test the models. Another major problem in the understanding of evolutionary pathways is the uncertainty about the frequency with which dioecy has evolved directly from hermaphroditism or via androdioecy. In fact, the evolution and occurrence of androdioecy itself has been questioned (Charlesworth & Charlesworth, 1978a, 1978b, pers. comm.; and see Haber & Bawa, this symposium). Systems such as those in Actinidia chinensis (Schmid, 1978), Saurauia spp. (Haber & Bawa, this symposium) and Solanum spp. (Anderson, 1979) may prove to be useful in the search for general models for the evolution of dioecism via androdioecy. Third, the importance of selection against inbreeding depression (see e.g., Willson, 1979; Bawa, 1980a, 1982a; Givnish, 1980, 1982; Thomson & Barrett, 1981; Beach, 1981; Char-