“Pollination syndromes” are specific combinations of floral traits that are proposed to evolve convergently across angiosperm lineages in response to different types of animal pollinators. In spite of their long history, pollination syndromes have not been tested adequately–they rarely have been examined critically to determine how well they describe floral trait diversity or predict pollinators. In a recent meta-analysis of data from the literature, Rosas-Guerrero et al. (2014) provide a welcome test that draws on insights from past studies. At the same time, their study illustrates several difficulties of meta-analysis approaches in general, and for pollination biology in particular. Here we discuss those difficulties and propose some solutions. We first consider how to gather studies from the literature without introducing unintended bias, such as the old-fashioned method of working backward from cited literature. We next consider how to deal with difficulties that invariably arise when extracting and analyzing often-incomplete information from heterogeneous studies. Finally we discuss issues of interpreting and presenting the results in the most informative manner. We conclude that although Rosas-Guerrero et al. (2014) and other studies such as Ollerton et al. (2009) have arrived at different conclusions about the utility of pollination syndromes, their results are not necessarily incompatible.
