Evolutionary biologists have long hypothesized that the diversity of flower
colours we see is in part a strategy to promote memorization by pollinator
s, pollinator constancy and therefore, a directed and efficient pollen tran
sfer between plants. However, this hypothesis has never been tested against
a biologically realistic null model, nor were colours assessed in the way
pollinators see them. Our intent here is to fill these gaps. Throughout one
year, we sampled floral species compositions at five ecologically distinct
sites near Berlin, Germany. Bee-subjective colours were quantified for all
168 species. A model of colour vision was used to predict how similar the
colours of sympatric and simultaneously blooming flowers were for bees. We
then compared flower colour differences in the real habitats with those of
random plant communities. We did not find pronounced deviations from chance
when we considered common plants. When we examined rare plants, however, w
e found significant divergence in two of the five plant communities. At one
site, similarly coloured species were found to be more frequent than expec
ted, and at the other two locations, flower colours were indistinguishable
from a random distribution. These results fit theoretical considerations th
at rare plants are under stronger selective pressure to secure pollination
than common plants. Our study illustrates the power of linking such distinc
t biological traditions as community ecology and the neuroethology of bee v
ision.