The hypothesis that the parasite Toxoplasma gondii manipulates the behaviour of its intermediate rat host in order to increase its chance of being predated specifically by its feline definitive host, rather than a non-definitive host predator species, was tested. The impact of a range of therapeutic drugs, previously demonstrated to be effective in preventing the development of T. gondii-associated behavioural and cognitive alterations in rats, on definitive-host predator specificity was also tested. Using a Y-shaped maze design, we demonstrated that T. gondii-associated behavioural changes, apparently aimed to increase predation rate, do appear to be specific to that of the feline definitive host – there were significant and consistent differences between the (untreated) infected and uninfected rats groups where T. gondii-infected rats tended to choose the definitive host feline-predator-associated maze arm and nest-box significantly more often than a maze arm or nest-box treated with non-definitive host predator (mink) odour. Drug treatment of infected rats prevented any such host-specificity from being displayed. We discuss our results in terms of their potential implications both for T. gondii epidemiology and the evolution of parasite-altered behaviour.