Does flower size follow the ‘island rule’? A commentary on ‘Flower size evolution in the Southwest Pacific’

Islands are considered as natural laboratories for evolution. In particular, the study of evolution of plant traits after island colonization has a long tradition, such as the ‘Baker law’ for the evolution of mating systems (Baker, 1955; Stebbins, 1957), the predictions about secondary evolution of woodiness (Zizka et al., 2022) or the evolution of lizard pollination (Olesen and Valido, 2003). In recent times, an island plant syndrome has been suggested (Burns, 2019; Whittaker et al., 2023) that is defined by particular changes in growth habit, plant size, seed size, defence against herbivores and the so-called floral island syndrome. The floral island syndrome entails secondary evolution of selfing or dioecy, as well as inconspicuous, actinomorphic flowers accessible to many kinds of pollinators (Ciarle and Burns, 2024). Despite continued attention to changes in body size related to island colonization, patterns of evolution of flower size on islands have received little attention. In a new study, Ciarle et al. (2025) address whether flower size follows the island rule, i.e. whether big flowers become smaller and small flowers become bigger after colonization of islands. In addition, they assessed whether the floral display increased for animal-pollinated species, compared with continental species.

For this study, Ciarle et al. (2025) compiled information from 129 colonization events involving 145 species from 43 families, in ten archipelagos of the Southwest Pacific, which included both subtropical and temperate/subantarctic climates. Species were chosen so that Australia or New Zealand were the sources of colonization and pairs of sister taxa were identified, using either phylogenies or, when phylogenies were not available, vegetative morphology. The use of sister species departs from other approaches to study the evolution of traits on islands, where phylogenetic relationships have been used (Zizka et al., 2022), in that it allows isolation of cases of in situ evolution of the derived trait stage (Ciarle and Burns, in press) (Fig. 1). Another elegant methodological detail was the inclusion of biological covariates to ensure the robustness of patterns found. In particular, to assess whether trends in modification of flower size were a mere product of allometry, i.e. differences in plant size (Niklas, 1994), a complementary test of the changes in leaf area or seed mass in the colonization events was performed. In addition, pollination syndrome (animal- or wind-pollination), sexual system (monomorphic or dimorphic), lineage (family), continental source (Australia or New Zealand) and floral morphology (actinomorphy with free petals, actinomorphy with fused petals or zygomorphy) were entered as covariates. This is a comprehensive dataset: future comparisons of island-continent sister taxa could also consider time since the split of the taxa. Such information is relevant for a distinction between the processes of environmental filtering and in situ evolution (Fig. 1), which are expected to occur at contrasting paces.

Fig. 1.

Two scenarios of evolution of flower size in islands, illustrated with dwarfism. Left, environmental filtering of continental variation in flower size, such that only the small flowers get established on islands. Right, in situ evolution of smaller size from initial unfiltered colonization of all kinds of flower sizes.

Actually, covariates were key to making sense of the patterns found, particularly between animal- and wind-pollinated species. While animal-pollinated species followed the island rule, wind-pollinated island flowers showed a consistent gigantism compared with their continental sister species. Furthermore, the pattern for animal-pollinated species was dependent on their sexual system. Interestingly, flowers of dimorphic (including dioecious and gynodioecious sexual systems) island species became larger than their continental sister species, while no size change was found for monomorphic (including hermaphroditic, monoecious, andromonoecious and gynomonoecious) species. No trend was found for the floral display when island species were compared with their continental sister species. These results were robust to the influence of other covariates and to the exclusion of those species for which the phylogeny was not known. Finally, it was reassuring that these patterns departed from the trends found in the two vegetative traits used as surrogates of plant size: leaf area and seed mass.

This study strengthens the evidence that an island floral syndrome can be reconciled with the island rule, although the generality for flower size in islands of other geographical areas, in particular tropical latitudes, remains to be explored. More importantly, mechanisms underlying these patterns remain speculative. Ciarle et al. (2025) suggest that in animal-pollinated species the modification of floral size can be a consequence of the need to maintain a match between e.g. flower and pollinator size. As island pollinators are not very diverse and usually of small size (Kaiser-Bunbury et al., 2010), flowers will convergently evolve towards intermediate sizes (Burns, 2022). However, Ciarle et al. (2025) admit that the importance of a size match with pollinators is doubtful, because island flowers are mainly generalist and actinomorphic. For wind-pollinated plants, Ciarle et al. (2025) suggest that larger flowers could be the result of the lower dispersibility of the larger pollen expected for allometric reasons, or of the longer stigmas, which can capture more pollen. Now that the pattern is on the table, there is plenty of room for the study of its underlying mechanisms, either in particular species or in a comparative framework. An obvious first step is to compare the pollinator fauna of sister animal-pollinated species, as well as pollen and stigma size for sister wind-pollinated species. As a final thought, a scarcity of pollinators occurs not only on islands, but also at higher altitudes and latitudes. Similarities and differences in selective filters between these contrasting ecological scenarios can shed light on how flower size evolves in environments with an impoverished pollinator fauna.