A learning curve for returning to conventional total knee arthroplasty (TKA) after using computer-assisted (CAS) TKA has not yet been established. In this study, the postoperative mechanical axes of the first 30 consecutive CAS TKAs performed by a single surgeon were compared to his subsequent 120 conventionally performed TKAs. A ''re-learning curve'' of 30 conventional TKAs was necessary to attain an average postoperative mechanical axis statistically indistinguishable from the average CAS mechanical axis (1.99 ). This is a trend of which surgeons should be aware when converting from CAS TKA to conventional TKA. As a secondary goal, the authors identify the first clinical parameter, preoperative deviation from neutral mechanical axis, that may potentially serve as a guide for the selective use of CAS in TKA.