The form of angiosperm flowers is thought to have evolved in part via
selection that excludes ineffective pollinators and increases the effi
ciency of pollen transfer by effective ones. In previous studies with
the montane plant Ipomopsis aggregata, we documented pollinator-mediat
ed selection on several aspects of floral form, including flower width
. This character varies continuously within natural populations of I.
aggregata, and individuals with wider flowers export more pollen per f
lower to surrounding plants. We previously showed that this component
of phenotypic selection is due primarily to the per-visit effectivenes
s of hummingbirds, the most important pollinator in our study populati
ons. Here we investigate mechanisms of differential visit effectivenes
s in greater detail. First, counts of pollen in unvisited flowers show
ed that pollen production itself increases with width. In aviary exper
iments, hummingbirds also removed a greater proportion of available po
llen as width increased. We next videotaped visits to flowers that var
ied in width either naturally, or by experimental treatment, and found
that hummingbirds inserted their bills more deeply into wider flowers
. Finally, we directly manipulated how deeply birds could insert their
bills, and found that more pollen was removed after deep insertion. T
hus, several mechanisms appear to underlie selection on corolla width
via visit effectiveness in pollen export. One involves a phenotypic co
rrelation with pollen production; this underscores the value of experi
ments for untangling indirect from direct selection. Another mechanism
involves direct selection due to the depth of insertion of a hummingb
ird's bill; this is rare evidence for one form of selection involving
the ''fit'' between pollinator and flower.