No simple, well-accepted procedures exist for comparing the relative performance of genetic strains of fish in aquaculture environments. The major technical problem is providing adequate statistical replication of ponds or cages. Inclusion of an "internal reference" population of standard fish in each pond can reduce replicate error caused by variation in the physical environment, food supply and population density. This statistical procedure is feasible only if there is no biological interaction (interference competition) between the strain and the reference fish. An experimental design is described in which the replicates contain different proportions of reference fish relative to the test strains. This permits the effects of interstrain interference to be estimated. Data from growth of tilapia Oreochromis mossambicus in crowded aquaria are used to illustrate the effect of interference and other nonlinearities in the experimental design.