The Morris water maze is frequently used to screen mutant mice generated by gene targeting. Targeted ES-cells are often derived from 129/Sv or BALB/c mice, known as poor swimming navigation learners. After mating the founders with C57BL/6 mice, the F2 or F3 hybrid generation is typically used for behavioral testing. In hybrid 129/Sv x C57BL/6 mice, a modification of the betaAPP gene entails impaired swimming navigation learning. This is readily detected despite behavioral variability, because wild-type 129/Sv x C57BL/6 hybrids outperform either of the parental strains and provide a control sample with good baseline performance. However, after backcrossing to the 129/Sv(ev) strain, the mutation effects are no longer detectable, masked by the very poor performance of wild-type 129/Sv(ev) mice. We conclude that F2 and F3 generations of 129/Sv x C57BL/6 crosses provide a suitable genetic background for behavioral testing of transgenic mice, provided that the samples are large enough to compensate for genetic and epigenetic variability and provided that normal performance in the control group is verified by comparison against a large database of mice tested under identical conditions. Creating congenic lines by backcrossing to an inbred strain is unlikely to enhance the sensitivity of the Morris test. Backcrossing to 129/Sv(ev) may even reduce it.