Aim: The number of fish taxa that occur exclusively in marine biomes is approximately equal to the number that occur in freshwater biomes. Both the geographic area and habitable volume of the marine realm are vastly greater than for Earth's freshwater ecosystems, suggesting that the density of marine species is proportionately much lower in the oceans. Because freshwater lineages are relatively recently derived from older marine lineages, this difference in species density suggests that speciation rates might be elevated in freshwater systems. I tested whether speciation rates differ systematically between freshwater and marine habitats.

Location: Aquatic ecosystems worldwide

Taxon: Ray-finned fishes (Actinopterygii)

Methods: Marine-freshwater transitions were tabulated from literature survey and from ancestral state reconstruction. I tested for repeated effects of salinity transitions on speciation rate using formal state-dependent diversification methods (STRAPP, FiSSE). Using maximum likelihood, I then tested for absolute (unreplicated) differences in speciation rate between marine and freshwater lineages.

Results: Ray-finned fishes have undergone numerous transitions from marine to freshwater systems, but the vast majority of freshwater species richness has resulted from a handful of freshwater colonization events. Speciation rates in freshwater lineages are substantially faster on average than those of marine lineages, but transitions to freshwaters do not lead to elevated rates of speciation in general. This paradox of state-dependent diversification arises because of the disproportionate effect of several freshwater clades with high species richness and fast rates of speciation. 

Main Conclusions: Transitions to freshwater do not cause faster rates of speciation, but freshwater ecosystems worldwide are dominated by several clades with relatively fast rates of speciation. There is no evidence that invasion of a novel habitat (freshwater) is generally sufficient to trigger a burst of speciation in colonizing lineages. These results raise an important conceptual problem for the interpretation of state-dependent diversification analyses.