Hallucinations and delusions are the most prominent symptoms of schizophrenia and characterized by impaired reality testing. Representation-mediated taste aversion (RMTA) has been proposed as a potential behavioral assessment of reality testing and has been applied to a neurodevelopmental rat model of schizophrenia. However, the theory underlying this approach has not been generalized yet with any demonstration of impaired reality testing in other animal models of schizophrenia, such as genetically-modified mice. We devised a RMTA procedure for mice that combines a Pavlovian association protocol pairing odor conditioned stimulus (CS) with sugar reward unconditioned stimulus (US), and a conditioned taste aversion (CTA) method. In this RMTA paradigm, we compared performances of wild-type (PLCβ1+/+) mice and phospholipase C β1 knock-out (PLCβ1-/-) mice which are known as one of the genetic models for schizophrenia. With a minimal amount of initial odor-sugar associative training, both PLCβ1+/+ and PLCβ1-/- mice were able to form an aversion to the sugar reward when the odor CS predicting sugar was paired with nausea. With an extended initial training, however, only PLCβ1-/- mice could form a RMTA. This persistent RMTA displayed by PLCβ1-/- mice shows their inability to distinguish real sugar from the CS-evoked representation of sugar at a stage in associative learning where wild-type mice normally could differentiate the two. These results demonstrate an impaired reality testing first observed in a genetic mouse model of schizophrenia, and suggest that RMTA paradigm may, with general applicability, allow diverse biological approaches to impaired reality testing.