Premise: The relative per‐flower production of ovules and pollen varies broadly with angiosperm mating systems, with outcrossing types commonly producing more pollen grains per ovule than selfing types. The evolutionary causes of this variation are contentious, especially the relevance of pollination risk. Resolution of this debate may have been hampered by its focus on pollen:ovule (P:O) ratios rather than on the evolution of pollen and ovule numbers per se.

Methods: Using published mean ovule and pollen counts, we analyzed associations with the proportion of removed pollen that reaches stigmas (pollen‐transfer efficiency) and differences between pollinator‐dependent and autogamous forms within and among species. Analyses involved Bayesian methods that simultaneously considered variation in pollen and ovule numbers and accounted for phylogenetic relatedness. We also assessed the utility of P:O ratios as mating‐system proxies and their association with female outcrossing rates.

Results: Median pollen number declined consistently with pollen‐transfer efficiency among species, whereas median ovule number did not. Similarly, in both intraspecific and interspecific analyses, pollinator‐dependent plants produced more pollen than autogamous plants, whereas ovule production did not differ statistically. Distributions of P:O ratios overlapped extensively for self‐incompatible and self‐compatible species and for different mating‐system classes, and P:O ratios correlated weakly with outcrossing rate.

Conclusions: Our findings demonstrate that pollinator dependence and pollination efficiency commonly influence the evolution of pollen number per flower but have more limited effects on ovule number. P:O ratios provide ambiguous, possibly misleading, information about mating systems, especially when compared among clades.

Data compilation

We searched for papers that reported pollen and ovule numbers using three complementary approaches: by searching the Web of Science (WoS) for papers that included the key words “pollen:ovule”, “pollen-ovule” and “pollen/ovule”; by using Erbar and Langlotz’s (2005, Botanische Jahrbucher 126: 71-132) compilation of published pollen:ovule ratios to identify older papers not abstracted by WoS; and by using the WoS citation information to identify papers that had cited Cruden's (1977, Evolution 31: 32-46) seminal paper on pollen:ovule ratios. We also recorded information concerning species characteristics, of which those relevant for this study are pollen-vector dependence, life history (annual, biennial, or monocarpic or iteroparous perennial), female outcrossing rate and pollen-transfer efficiency. If the source reference of the pollen and ovule records did not provide this information, we sought it from supplementary sources.

Phylogeny construction

Most analyses accounted explicitly for the phylogenetic relatedness of species. As described in more detail, phylogenetic trees were constructed using the online version of OneTwoTree (http://onetwotree.tau.ac.il/; Drori et al., 2018, Molecular Ecology Resources 18: 1492-1499). To facilitate taxon recognition by OneTwoTree, the list of species for an analysis (including subspecies and varieties) named in the original data sources was submitted to the Global Biodiversity Information Facility (https://www.gbif.org/) to retrieve currently accepted taxonomic names. 

All data files can be opened with any text editor.