Species trees in (A) show the history of substitutions for loci in two genetic networks that differ in molecular evolution (case on left with e.g. fewer genes, stronger purifying selection, less adaptation, greater modularity, or lower pleiotropic effects) and in the potential for DMIs (solid purple lines = potential DMI for derived-derived substitutions between species, dashed lines = potential derived-ancestral DMIs; lowercase = ancestral alleles, uppercase = derived substitutions unique to one lineage). The number of potential DMIs (purple) scales faster than linear with number of substitutions (red and blue hashes); faster evolving genetic networks may be more likely to experience this ‘snowball effect’ of reproductive isolation (Orr, 1995). DSD arises when the outward phenotype remains constant despite molecular divergence between descendant species. Panel (B) shows how two loci (a and b) that diverge can potentially create a DMI upon formation of F1 hybrids between descendant species. Panel (C) illustrates with a Fisher’s geometric model visualization, for two traits with a shared genetic architecture, how adaptive evolution with respect to one trait (Trait 1) can generate DSD in another (Trait 2). Concentric circles represent lines of equal fitness; filled dots (black = ancestor, red and blue = descendant species) indicate genotypes (letters as in A) that evolve via three substitutions (arrows) toward the fitness optimum at the center. Note the DSD in Trait 2 due to no net phenotypic difference relative to the ancestor at the end of the adaptive walk for both species, despite underlying genetic changes. Panel (D) shows evolution along ridges of equal fitness in a fitness landscape comprised of a genetic architecture with many genes. Genotypic paths evolve independently in different species (ancestral black to derived red and blue species), similarly to DSD, such that hybrids between them (purple) occupy a portion of genotype space with low fitness (‘holes’).