We investigated the impact of enhanced mechanical transparency during practice on closed-loop performance as well as on the acquisition of an internal representation of the visuo-motor transformation of a lever. Three groups of participants controlled a cursor on a monitor by moving the effort arm of a sliding two-sided lever. The level of mechanical transparency was manipulated by varying the amount of conceptual knowledge about the tool the participants were using as well as by the visualization of the effective part of the tool on the monitor. While for groups cursor− and cursor+ only a cursor representing the tip of the load arm of the lever was visible, group shaft was additionally provided with a representation of the load arm on the monitor. The cursor groups differed in conceptual knowledge about the tool, with only group cursor+ being informed that they were controlling a two-sided lever. Enhanced mechanical transparency, resulting from the visibility of the lever arm, boosted performance in terms of faster and straighter movements when concurrent visual feedback was available, and therefore closed-loop control was possible. In contrast, a detrimental effect of enhanced mechanical transparency during practice on the accuracy of the internal representation of the visuo-motor transformation was observed. When assessing open-loop control in test trials without visual feedback, participants who had previously practiced the tool transformation with the visible lever arm showed larger end point errors than those who had practiced with the cursor only. These findings support the notion that enhanced mechanical transparency of a tool is a type of environmental support which boosts closed-loop performance, but impedes open-loop performance, most likely by serving as a substitute for an internal representation. When practice conditions facilitate closed-loop control, the acquisition of an accurate internal representation of a novel visuo-motor transformation seems to be hampered.