Carotenoids produce most of the brilliant orange and yellow colours seen in animals, but animals cannot synthesize these pigments and must rely on dietary sources. The idea that carotenoids make good signals because they are a scarce limiting resource was proposed two decades ago and has become the leading hypothesis for the role of carotenoids in animal communication. To our knowledge, until now, however, there has been no direct evidence that carotenoids are a limiting resource in nature. We showed that carotenoid availability in the wild limits the expression of sexual coloration in guppies (Poecilia reticulata), a species in which females prefer males with brighter orange carotenoid-containing spots. Further, the degree of carotenoid limitation varies geographically along a replicated environmental gradient (rainforest canopy cover), which opens new avenues for testing signal evolution theory.

Life history traits in guppies (Poecilia reticulata) vary geographically along a predator assemblage gradient, and field experiments have indicated that the association may be causal; guppies introduced from high predation sites to low predation sites have evolved the phenotype associated with low predation in as few as seven generations. It has long been recognized, however, that low predation sites tend to have greater forest canopy cover than high predation sites. Stream differences in canopy cover could translate into stream differences in resource availability, another theoretically potent agent of selection on life history traits. Moreover, new computer simulations indicate that the high predation phenotype would outcompete the low predation phenotype under both mortality regimes. Thus, predation alone may not be sufficient to explain the observed life history patterns. Here we show that food availability for guppies decreases as forest canopy cover increases, among six low predation streams in the Northern Range of Trinidad. Streams with less canopy cover received more photosynthetically active light and contained a larger standing crop of algae (the primary food of guppies), as measured by algal pigments (chlorophylls and carotenoids) on both natural cobble and artificial tile substrates, but did not contain a greater biomass of guppies (per square meter of streambed). Consequently, algae availability for guppies (in micrograms of algal pigments per milligram of guppy) increased with decreasing canopy cover. The biomass of guppies and algae both decreased after a series of floods, with no net effect on algae availability. Field mark-recapture studies revealed that female and juvenile guppies grew faster, and that the asymptotic size of mature males was larger, in streams with less canopy cover. Canopy cover explained 84% of the variation among streams in algae availability which, in turn, explained 93% of the variation in guppy growth rates. Laboratory "common garden" experiments indicated that the stream differences in growth and adult male size in the field were largely environmental (non-genetic). These results strongly suggest that stream differences in canopy cover result in consistent stream differences in food availability, independent of predation. Our preliminary data indicate that some life history traits (offspring size and litter size) vary genetically along the canopy cover gradient, among low predation streams, in the same direction as along the predation gradient. Another recent study shows that food availability is higher at high predation sites than at low predation sites, partly as an indirect effect of predators reducing guppy densities. Further research is required to disentangle the direct effects of predation from those of resource availability in the evolution of life histories.