A bstract Common models of speciation with gene flow consider constant migration or admixture on secondary contact, but earth’s recent climatic history suggests many populations have experienced cycles of isolation and contact over the last million years. How does this process impact the rate of speciation, and how much can we learn about its dynamics by analyzing the genomes of modern populations? Here we develop a simple model of speciation through Bateson-Dobzhansky-Muller incompatibilities in the face of periodic gene flow and validate our model with forward time simulations. We then use empirical atmospheric CO 2 concentration data from the Vostok Ice Cores to simulate cycles of isolation and secondary contact in a tropical montane landscape, and ask whether they can be distinguished from a standard isolation-with-migration model by summary statistics or joint site frequency spectrum-based demographic inference. We find speciation occurs much faster under periodic than constant gene flow with equivalent effective migration rates ( Nm ). These processes can be distinguished through combinations of summary statistics or demographic inference from the site frequency spectrum, but parameter estimates appear to have little resolution beyond the most recent cycle of isolation and migration. Our results suggest speciation with periodic gene flow is a common force in generating species diversity through Pleistocene climate cycles, and highlight the limits of current inference techniques for demographic models mimicking the complexity of earth’s recent climatic history.