Response strategy in the dual-solution plus maze is regarded as a form of stimulus-response learning. In this study, by using an outcome devaluation procedure, we show that it can be based on both action-outcome and stimulus-response habit learning, depending on the amount of training that the animals receive. Furthermore, we show that deactivation of the dorso-medial and the dorso-lateral striatum with "Botulinum neurotoxin A", mimicked or abolished, respectively, the effects of practice on the sensitivity of the response strategy to outcome devaluation. These findings have relevant implications for the understanding of the learning mechanisms underlying different overt behaviors in this widely used maze task.