Thr Fraction of pollen that is carried over from one flower to the nex
t as a pollinator visits a sequence of flowers has a strong effect on
patterns of mating in plant populations. Although most studies have us
ed simple exponential or geometric functions to represent carryover, p
ollen deposition patterns often show longer-than-geometric tails, sugg
esting a systematic increase in the carryover fraction. We introduce a
changing carryover model that allows the carryover fraction to increa
se or decrease during a sequence of visits. Maximum Likelihood methods
were used to fit constant and changing carryover models to twelve pre
viously published data sets. A model with an increasing carryover frac
tion provided a better fit than a constant carryover model in 75% of t
he cases examined. We found only partial evidence for one proposed cau
se of increasing carryover, the burial of donor pollen beneath pollen
collected from recipient flowers. Finally, we use the re suits for Ipo
mopsis aggregata, a hummingbird-pollinated plant that presents many fl
owers simultaneously, to show that the changing carryover model predic
ts less pollen will be transferred among flowers on the same plant (ge
itonogamy) than does a constant carryover model fit to the same experi
mental data.