Sex-allocation models predict that the evolution of self-fertilization
should result in a reduced allocation to male function and pollinator
attraction in plants. The evolution of sex allocation may be constrai
ned by both functional and genetic factors, however. We studied sex al
location and genetic variation for floral sex ratio and other reproduc
tive traits in a Costa Rica population of the monoecious, highly selfi
ng annual Begonia semiovata. Data on biomass of floral structures, flo
wer sex ratios, and fruit set in the source population were used to ca
lculate the average proportion of reproductive allocation invested in
male function. Genetic variation and genetic correlations for floral s
ex ratio and for floral traits related to male and female function wer
e estimated from the greenhouse-grown progeny of field-collected mater
nal families. The proportion of reproductive biomass invested in male
function was low (0.34 at flowering, and 0.07 for total reproductive a
llocation). Significant among-family variation was detected in the siz
e (mass) of individual male and female flowers, in the proportion of m
ale Rowers produced, and in the proportion of total flower mass invest
ed in male flowers. Significant among-family variation was also found
in flower number per inflorescence, petal length of male and female fl
owers, and petal number of female flowers. Except for female petal len
gth, we found no difference in the mean value of these characters betw
een selfed and outcrossed progeny, indicating that, with the possible
exception of female petal length, the among-family variation detected
was not the result of variation among families in the level of inbreed
ing. Significant positive phenotypic and broad-sense genetic correlati
ons were detected between the mass of individual male and female flowe
rs, between male and female petal length, and between number of male a
nd number of female flowers per inflorescence. The ratio of stamen-to-
pistil mass (0.33) was low compared to published data for autogamous s
pecies with hermaphroditic flowers, suggesting that highly efficient s
elfing mechanisms may evolve in monoecious species. Our results indica
te that the study population harbors substantial genetic variation for
reproductive characters. The positive genetic correlation between inv
estment in male and female flowers may reflect selection for maximum p
ollination efficiency, because in this self-pollinating species.