Abstract: Chemical communication is critical in establishing angiosperm-pollinator mutualisms. However, our understanding of how chemical communication shapes coevolution remains limited. Here, we integrated information theory to model three coevolutionary scenarios (I?III), where the pollinator fitness is always optimized by the highest certainty of chemical information provided by plants, but plant fitness is determined by (I) the certainty of chemical information attracting pollinators, (II) the uncertainty of chemical information confusing antagonists, or (III) both aspects. We found that the statistical properties of empirical plant volatiles from 45 pairs of fig-pollinator mutualisms were best explained by the selection from both pollinators and antagonists (scenario III). Under this scenario, plant-pollinator mutualisms evolve to be specialized and as few as two volatile chemicals could supply sufficient information for pollinators’ host identification. Our study provides new insights into plant-pollinator coevolution and will facilitate further studies on the evolution and diversification in specialized plant-pollinator-herbivore systems.

Key words: coevolution, conditional entropy, fig-pollinator mutualism, fitness, information theory, mutual information, plant-insect interaction