blance is maintained because of selection for similar-
ity (Willson and Agren 1989) and that the reproductive
success of the rewardless female flowers is a function
of their degree of resemblance to the reward-producing
male flowers (Agren and Schemske 1991).
Selection for resemblance could potentially take
two different forms: the most attractive female flower
could resemble either the average male flower in the
population, or the male with the highest reward level
(Agren and Schemske 1991, Schemske and Agren
1995). In the former situation, plants would receive
the most pollen and have the highest reproductive
success. In the latter, the highest seed production is
expected in plants producing female flowers that are
most similar to the most rewarding male flowers
(Schemske and Agren 1995). We found a positive
correlation between petal size and estimates of pol-
len production in male flowers of several Begonia
species, suggesting that flower visitors may use petal
size as an indicator of reward level (Schemske et al.
1996, J. Agren and D. Schemske, unpubl. data).
The perennial herb B. involucrata is native to Costa
Rica and Panama and is found along roadsides, trails,
and pastures in Monteverde (Fig. 8.17). This species
flowers during the dry season (January-March), and
male and female floral displays are very similar
(Agren and Schemske 1991, Schemske and Agren
1995). The yellow stigma of the female flower strongly
resembles the anthers of the male flower. In ultravio-
let light, both the stigma and the anthers appear as
dark targets in the center of the flower against the
lighter background of the petallike sepals. Both male
and female flowers have two petallike sepals of ap-
proximately the same size. Similar numbers of male-
and female-phase inflorescences open simultane-
ously; both produce a sweet fragrance.
The main pollinators of B. involucrata in Monte-
verde (bees Trigona grandipennis and Melipona fas-
data panamica) respond to the sexual difference in
reward-production by showing a strong preference for
male flowers (Fig. 8.18). When pollen was added by
hand, seed production was increased by about 40%
in the study population at Monteverde in 1990, indi-
cating that pollinator visitation may limit seed pro-
duction (Agren and Schemske 1991).
To examine how pollinator preference may shape
the evolution of female flower size, we constructed
artificial Begonia involucrata inflorescences that dif-
fered in flower size, and exposed these inflorescences
to natural pollinators. Both the number of pollinator
approaches and the number of pollinator visits in-
creased monotonically with petal size (Schemske and
Agren 1995; Fig. 8.19). In this species, individuals pro-
ducing the largest petals were at an advantage. This
is consistent with the hypothesis that pollinators use
the intensity of visual stimuli to identify flowers with
the greatest expected reward, and suggests that female
flowers which mimic the most rewarding male flow-
ers will have higher pollinator visitation and repro-
ductive success.Figure 8.17. Female-phase inflorescence of Begonia Involucrata in Monteverde. Photograph by
J. Agren.280 Plant-Animal Interactions