With species found throughout both marine and fresh waters, the diatom order Thalassiosirales is one of the most phylogenetically and ecologically diverse lineages of planktonic diatoms. A clear understanding of the timescale of Thalassiosirales evolution would provide novel insights into the rates and patterns of species diversification associated with major habitat shifts, as well as provide valuable context for understanding the age and evolutionary history of the model species, Cyclotella nana (= Thalassiosira pseudonana). The freshwater fossil record for Thalassiosirales is extensive, well characterized, and generally supportive of a Miocene origin for the major freshwater lineages. The marine record is, by comparison, more sparse and in many cases, unverified. The discovery of freshwater thalassiosiroids in Eocene sediments pushed the freshwater fossil record considerably further back in time, highlighting an apparent gap of some 30 million years. An alternative interpretation is that the Miocene and Eocene reports represent competing hypotheses. In the absence of additional independent and decisive fossil data, I explored the relative plausibility of these two scenarios with Bayesian relaxed molecular clock methods under a range of fossil calibration schemes. Although I found no support for the Eocene fossil dates, the two major freshwater colonization events probably occurred much earlier than previously thought—as early as the Paleocene for Cyclotella, followed by an Eocene origin for the cyclostephanoid lineage. Much of the extant freshwater diversity in both lineages traces back to the Miocene, however, giving the impression of a single Miocene origin. Efforts to infer the timescale of Thalassiosirales evolution more accurately would benefit from a systematic reevaluation of the marine fossil record and formal integration of fossil species into existing phylogenetic hypotheses.