To evaluate the "learning curve" effect on feasibility and reproducibility of laparoscopic liver resection (LLR). LLR is currently limited to few centers and to few procedures. Its reproducibility is still debated. : Patients undergoing LLR between 1996 and 2008 were included. Indications and type of hepatectomies were compared with those of open resections performed in the same period, considering 3 periods (1996-1999, 2000-2003, and 2004-2008). LLRs were divided into 3 equal groups of 58 cases and technical data and outcomes were compared. Risk-adjusted Cumulative Sum model was used for determining the learning curve based on the need for conversion. Of 782, 174 (22.3%) patients underwent LLR. Proportion of LLR progressively increased (17.5%, 22.4%, and 24.2%), such as hepatocellular carcinoma (17.6%, 25.6%, and 39.4%, P < 0.05), colorectal metastases (0%, 6.5%, and 13.1%, P < 0.05), major hepatectomies (1.1%, 9.1%, 8.5%, P < 0.05), and right hepatectomies (0%, 13.2%, and 13.1%, P < 0.05). Comparing groups, results of LLR significantly improved in terms of conversion rate (15.5%, 10.3%, and 3.4%, P < 0.05), operative time (210, 180, and 150 minutes, P < 0.05), blood loss (300, 200, and 200 mL, P < 0.05), and morbidity (17.2%, 22.4%, and 3.4%, P < 0.05). Pedicle clamping was less used over time (77.6%, 62.1%, and 17.2%, P < 0.05) and for shorter durations (45, 30, and 20 minutes, P < 0.05). Having adjusted for case-mix, the Cumulative Sum analysis demonstrated a learning curve for laparoscopic hepatectomies of 60 cases. A slow but constant evolution of LLR occurred: indications and magnitude of procedures increased and technical outcomes improved. The learning curve demonstrated in this study suggests that LLR is reproducible in liver units but specific training to advanced laparoscopy is required.