Abstract -- Vertebrate animals reproducing without genetic recombination typically are hybrids, which have large ranges, are locally abundant, and live in disturbed or harsh habitats. This holds for the hemiclonal hybridogenetic frog Rana esculenta: it is widespread in Europe and commonly is found in disturbed habitats such as gravel pits. We hypothesize that its widespread occurrence may either be the result of natural selection for a single hemiclone acting as a broadly adapted "general-purpose" genotype, or of interclonal selection, which maintains multiple hemiclones that each are relatively narrowly adapted and perform differently across environments, that is, the Frozen Niche Variation model. We tested these competing hypotheses using 1000-L outdoor artificial ponds to rear tadpoles of the parental species (Rana lessonae [LL] and Rana ridibunda [RR]) alone, and each of three hemiclones of Rana esculenta (GUT1, GUT2, GUT3) alone, and in mixed hemiclonal populations from hatching to metamorphosis. Tadpoles of three coexisting hemiclones from a single natural population (near GM-|tighausen, Switzerland) were reared in both two- and three-way mixtures in equal total numbers at high and low density. For each species and hemiclone, the proportion of tadples metamorphosing decreased as the density of tadpoles increased, with the threee hemiclones spanning the range of values exhibited by the two parental species. LL and GUT1 tadpoles produced the highest proportion of metamorphs, whereas tadpoles of RR produced the fewest metamorphs at both densities. GUT1 tadpoles also produced the largest metamorphs at low density, GUT2 and GUT3 tadpoles produced smaller metamorphs than did GUT1 tadpoles at the low density, but the three hemiclones did not differ from each other at high density. The parental species (LL and RR) were intermediate in metamorphic size to the hemiclones at low density, but all genotypes converged on a similar size at high density. Length of the larval period was also affected by density, but its effect was dependent on genotype. GUT1 tadpoles had the shortest larval period at the low density, but larval period was longer and not different between GUT1, GUT3, and LL at high density. RR tadpoles had the longest larval period at both densities. The most dramatic results were that three genotypes (GUT1, GUT2, and RR) maintained rank order and increased days to metamorphosis from low to high density, whereas two genotypes (GUT3 and LL) changed rank order and decreased days to metamorphosis from low to high density. Mixtures of hemiclones in two- and three-way combinations facilitated the proportion of tadpoles metamorphosing for GUT1 and GUT2 at both densities, but only at the low density for GUT3 tadpoles. Results from this experiment are incompatible with the General-Purpose Genotype model as a global explanation of hybrid abundance in these frogs. Alternatively, the Frozen Niche Variation prediction of general performance superiority of clonal mixtures relative to single clone populations is strongly supported. The data confirm that fitness advantages of hemiclones change, depending on the environment, such that in temporally and spatially heterogeneous habitats like ponds, frequency-dependent selection among hemiclones may promote coexistence in hemiclonal assemblages. Yet, differential dispersal or colonization ability and historical factors affecting hemiclone distribution may also be important in shaping patterns of clonal coexistence.