We studied variation in flowering phenology, fruit and seed set, and the abundance of the pollinators
of four species of night-blooming Sonoran Desert columnar cacti for up to eight years at one site in
Mexico and one year at one site in Arizona. We determined how spatiotemporal variation in plant–pollinator
interactions affects the evolution of generalized pollination systems. We conducted pollinator exclusion and
hand pollination experiments to document annual variability in pollinator reliability and to determine whether
pollination systems were redundant (different species are partially or totally substitutable) or complementary
(different species have an additive effect on fruit set). The cacti we studied included three species with generalized
pollination systems involving bats, birds, and bees (cardon, Pachycereus pringlei; saguaro, Carnegiea gigantea;
and organ pipe, Stenocereus thurberi) and one specialized moth-pollinated species (senita, Lophocereus schottii).
We predicted that the migratory lesser long-nosed bat, Leptonycteris curasoae, is a less reliable pollinator than
birds and bees, and that cacti with generalized pollination systems have more variable flowering phenologies
than the specialized species.
Annual time of peak flowering and mean size of flower crops were relatively invariant in saguaro and organ
pipe. Time of peak flowering in cardon varied by as much as six weeks, and mean flower crop size varied threefold
over six years. In senita, peak flowering varied by as much as 5–8 wk among years. Peak numbers of the
nectar bat L. curasoae varied among years, and bat density (0.9/ha) was an order of magnitude lower than that
of cactus-visiting birds at both study sites. The abundance of migratory hummingbirds was also highly variable
Pollinator exclusion experiments indicated that bats were major pollinators of cardon, whereas diurnal visitors
accounted for most fruit set in saguaro (except in 1995 when bats were most important) and organ pipe at our
Mexican site; honeybees accounted for 64–87% of diurnal fruit set in these species. Annual variation in the
contribution to fruit set by bats was substantially higher than that of diurnal pollinators in saguaro and organ
pipe, but not in cardon. There was little geographic variation in the relative importance of nocturnal vs. diurnal
pollinators in saguaro and senita, but bats were much more important for fruit set in organ pipe in Arizona than
in Mexico. We generally detected no effect of different pollinators on number of seeds per fruit in any species.
Annual variation in fruit set was lowest in saguaro, the species with the most diurnal pollination system,
and highest in organ pipe, the species with the most generalized pollination system. Fruit set was strongly pollen
limited only in females of cardon (a trioecious species) and in organ pipe (at both sites). The ‘‘missing’’ pollinators
in both species are likely Leptonycteris bats. The pollination systems of saguaro and cardon were partially
redundant, whereas that of organ pipe was complementary.
The four species of cactus that we studied occur at the northern geographic limits of Mexican columnar cacti
where many vertebrate pollinators are seasonal migrants. In the Sonoran Desert, variation in rainfall and spring
temperatures affects timing of flowering and the extent of competition between cacti for pollinator visits and
causes the relative importance of particular pollinators, especially Leptonycteris bats, for fruit set to vary annually.
Under such conditions, selection has favored generalized pollination systems (as seen in organ pipe) or shifts
from reliance primarily on nocturnal pollinators (as seen in cardon) to reliance primarily on diurnal pollinators
(as seen in saguaro). Nonetheless, as exemplified by the senita–senita moth system, highly specialized pollination
mutualisms can also evolve in this habitat in plants that rely on sedentary insects rather than migratory bats
and birds for pollination