Haptic guidance has been shown to interfere with learning a novel visuo-motor rotation. Here, we ask whether this interference is specific to the learning of visuo-motor transformations or whether it is a more generalized phenomenon of learning spatial movement characteristics. Participants practiced to make movements at an angle of 75° relative to target directions with and without haptic path guidance. Test trials without guidance and without visual feedback revealed poorer performance after haptic guidance practice than in the no-guidance control group, even though during guided practice movement directions were more accurate than during unguided practice. The more precise directions during guided practice were associated with shorter movement latencies than in the control group, suggesting an incomplete rotation of the movement direction relative to the direction of the visual target. A detailed analysis of the first few millimeters of each movement revealed that the initial movement direction was indeed rotated too little, and the correct direction was adopted only gradually under the influence of haptic guidance. These findings support the perspective on motor learning as a process that is driven by consequent improvements of performance and that is compromised by factors that serve to improve performance even without learning. Haptic guidance is one such factor.