We address the controversy in the literature concerning the definition of holobionts and the apparent constraints on their evolution using concepts from community population genetics. The genetics of holobionts, consisting of a host and diverse microbial symbionts, has been neglected in many discussions of the topic, and, where it has been discussed, a gene-centric, species-centric view, based in genomic conflict, has been predominant. Because coevolution takes place between traits or genes in two or more species and not, strictly speaking, between (...) species, it may affect some traits but not others in either host or symbiont. Moreover, when interacting species pairs are embedded in a larger community, indirect ecological effects can alter the expected pairwise dynamics. Mode of symbiont transmission and the degree of host inbreeding both affect the extent of microbial mixing across host lineages and thereby the degree to which selection on one trait of either partner affects other aspects of a holobiont phenotype. We discuss several potential defining criteria for holobionts using community genetics and population genetics models, suggesting their application and limitations. Using community genetics models, we show how conflict between genomes can be self-limiting, while cooperation and mutualism tend to be self-accelerating. It is likely that this bias in the evolutionary dynamics of interaction between hosts and symbionts is an important feature of holobionts. This bias in the evolutionary dynamic could contribute to explaining the absence of cheaters from natural mutualisms, although cheaters are predicted by gene-centered conflict theory to cause the evolutionary instability of mutualisms. Additionally, it may help explain the more frequent origin of mutualisms from parasitic than from free-living systems, an evolutionary trajectory opposite to that predicted by genome conflict theory. (shrink)