Humans can operate a variety of modern tools, which are often associated with diferent visuomotor transformations. Studies investigating this ability have shown that separate motor memories can be acquired implicitly when diferent sensorimotor transformations are associated with distinct (intended) postures or explicitly when abstract contextual cues are leveraged by aiming strategies. It still remains unclear how diferent transformations are remembered implicitly when postures are similar. We investigated whether features of planning to manipulate a visual tool, such as its visual identity or the environmental efect intended by its use (i.e. action efect) would enable implicit learning of opposing visuomotor rotations. Results show that neither contextual cue led to distinct implicit motor memories, but that cues only afected implicit adaptation indirectly through generalization around explicit strategies. In contrast, a control experiment where participants practiced opposing transformations with diferent hands did result in contextualized afterefects difering between hands across generalization targets. It appears that diferent (intended) body states are necessary for separate afterefects to emerge, suggesting that the role of sensory prediction error-based adaptation may be limited to the recalibration of a body model, whereas establishing separate tool models may proceed along a diferent route.