Opening the species box: what parsimonious microscopic models of speciation have to say about macroevolution

In the last two decades, lineage-based models of diversification, where species are viewed as particles that can divide (speciate) or die (become extinct) at rates depending on some evolving trait, have been very popular tools to study macroevolutionary processes. Here, we argue that this approach cannot be used to break down the inner workings of species diversification and that "opening the species box" is necessary to understand the causes of macroevolution, but that too detailed speciation models also fail to make robust macroevolutionary predictions. We set up a general framework for parsimonious models of speciation that rely on a minimal number of mechanistic principles: (a) reproductive isolation is caused by excessive dissimilarity between genotypes; (b) dissimilarity results from a balance between differentiation processes and homogenizing processes; and (c) dissimilarity can feed back on these processes by decelerating homogenization. We classify such models according to the main homogenizing process: (a) clonal evolution models (ecological drift), (b) models of genetic isolation (gene flow), and (c) models of isolation by distance (spatial drift). We review these models and their specific predictions on macroscopic variables such as species abundances, speciation rates, interfertility relationships, or phylogenetic tree structure. We propose new avenues of research by displaying conceptual questions remaining to be solved and new models to address them: the failure of speciation at secondary contact, the feedback of dissimilarity on homogenization, and the emergence in space of breeding barriers.