The existence of a continuous array of sympatric biotypes - from polymorphisms, through ecological or host races with increasing reproductive isolation, to good species - can provide strong evidence for a continuous route to sympatric speciation via natural selection. Host races in plant-feeding insects, in particular, have often been used as evidence for the probability of sympatric speciation. Here, we provide verifiable criteria to distinguish host races from other biotypes: in brief, host races are genetically differentiated, sympatric populations of parasites that use different hosts and between which there is appreciable gene flow. We recognize host races as kinds of species that regularly exchange genes with other species at a rate of more than ca. 1% per generation, rather than as fundamentally distinct taxa. Host races provide a convenient, although admittedly somewhat arbitrary intermediate stage along the speciation continuum. They are a heuristic device to aid in evaluating the probability of speciation by natural selection, particularly in sympatry. Speciation is thereby envisaged as having two phases: (i) the evolution of host races from within polymorphic, panmictic populations; and (ii) further reduction of gene flow between host races until the diverging populations can become generally accepted as species. We apply this criterion to 21 putative host race systems. Of these, only three are unambiguously classified as host races, but a further eight are strong candidates that merely lack accurate information on rates of hybridization or gene flow. Thus, over one-half of the cases that we review are probably or certainly host races, under our definition. Our review of the data favours the idea of sympatric speciation via host shift for three major reasons: (i) the evolution of assortative mating as a pleiotropic by-product of adaptation to a new host seems likely, even in cases where mating occurs away from the host; (ii) stable genetic differences in half of the cases attest to the power of natural selection to maintain multilocus polymorphisms with substantial linkage disequilibrium, in spite of probable gene flow; and (iii) this linkage disequilibrium should permit additional host adaptation, leading to further reproductive isolation via pleiotropy, and also provides conditions suitable for adaptive evolution of mate choice (reinforcement) to cause still further reductions in gene flow. Current data are too sparse to rule out a cryptic discontinuity in the apparently stable sympatric route from host-associated polymorphism to host-associated species, but such a hiatus seems unlikely on present evidence. Finally, we discuss applications of an understanding of host races in conservation and in managing adaptation by pests to control strategies, including those involving biological control or transgenic parasite-resistant plants.