Host-parasite coevolution is often described as a process of reciprocal adaptation and counter adaptation, driven by frequency-dependent selection. This requires that different parasite genotypes perform differently on different host genotypes. Such genotype-by-genotype interactions arise if adaptation to one host (or parasite) genotype reduces performance on others. These direct costs of adaptation can maintain genetic polymorphism and generate geographic patterns of local host or parasite adaptation. Fixation of all-resistant (or all-infective) genotypes is further prevented if adaptation trades off with other host (or parasite) life-history traits. For the host, such indirect costs of resistance refer to reduced fitness of resistant genotypes in the absence of parasites. We studied (co)evolution in experimental microcosms of several clones of the freshwater protozoan Paramecium caudatum, infected with the bacterial parasite Holospora undulata. After two and a half years of culture, inoculation of evolved and naive (never exposed to the parasite) hosts with evolved and founder parasites revealed an increase in host resistance, but not in parasite infectivity. A cross-infection experiment showed significant host clone-by-parasite isolate interactions, and evolved hosts tended to be more resistant to their own (local) parasites than to parasites from other hosts. Compared to naive clones, evolved host clones had lower division rates in the absence of the parasite. Thus, our study indicates de novo evolution of host resistance, associated with both direct and indirect costs. This illustrates how interactions with parasites can lead to the genetic divergence of initially identical populations.