Tests of emotionality conducted outside the home-cage show that rodents from standard laboratory housing are more anxious than animals from enriched housing; however, it is not known if this also indicates increased anxiety within the home-cage. We used a novel method, recording the self-administration of a psychoactive anxiolytic, to examine home-cage anxiety levels of laboratory mice (three per cage) in Standard (n = 10 cages), Unpredictable (n = 10 cages) and Enriched (n = 6 cages) housing. The mice were given a choice of drinking either non-drugged water or a solution of the benzodiazepine Midazolam. There was a significant effect of housing on the proportion of total fluid consumed from the bottle containing Midazolam solution (P = 0.02). Mice from Standard and Unpredictable cages drank a greater proportion than mice from Enriched cages. This indicates that mice from the Standard and Unpredictable laboratory caging may have been experiencing greater anxiety than mice from the Enriched cages. There was also a significant effect of bottle position (P = 0.002). Mice from the Standard and Unpredictable cages drank a greater proportion of total fluid from the bottle containing Midazolam solution when this was toward the rear of the cage rack, ie in a location that was less susceptible to extraneous disturbance. Monitoring self-administration of psychoactive drugs as a method of welfare assessment could be applied to a wide variety of housing conditions, husbandry practices, or experimental procedures that putatively induce negative mental states. The major finding is that standard cages for laboratory rodents may induce greater anxiety than enriched cages. This is discussed in terms of animal welfare and the validity of data from animals housed in minimalistic environments.