Deleterious impacts of major mutations can be ameliorated by stabilising selection acting on modifier genes. We hypothesise that a new hyperactive circling mouse (counterspin: Cr) arises when modifier genes inadvertently selected to ameliorate the negative impacts of a growth hormone transgenic insertion segregate into the normal genetic background that lacks the transgene. We hypothesise that such modifiers generate a phenotype ''mirror image'' to the transgenics on the otherwise normal background. We highlight this by testing a priori hypotheses that counterspin and transgenic growth hormone mice deviate oppositely from normal mice across a broad spectrum of characteristics. Results spanning growth, sensorimotor performance, cognition and striatal neurotransmitters provide strong circumstantial evidence for the hypothesis. In a more direct test for selection in the transgenic mice, we found that those examined in 2008 slept *3 h/d less than they did 14 years ago (P \ 0.0005). This is a profound change strongly supporting the reality of modifier selection in these mice. Our results highlight that modifiers may act powerfully on genetically engineered constructs given a genetically variable background. Furthermore, we suggest that modifier selection might provide a novel method for deriving genetic models, and specifically, models phenotypically opposite to engineered constructs or natural mutations.