PURPOSE: To evaluate the short-term and long-term reproducibility of lung tumor position for scans acquired using an active breathing control (ABC) device. METHODS AND MATERIALS: Ten patients with lung cancer were scanned over three sessions during the course of treatment. For each session, two scans were acquired at deep inhale, and one scan each at half of deep inhale and at exhale. Long-term reproducibility was evaluated by comparing the same breathing state scans from two sessions, with setup variation removed by skeletal alignment. Tumor alignment was based on intensity matching of a small volume around the tumor. For short-term reproducibility, the two inhale volumes from the same session were compared. RESULTS: For the short-term reproducibility, the mean and the standard deviation (SD) of the displacement of the center of tumor were 0.0 (1.5) mm in anteroposterior (AP), 0.3 (1.4) mm in superior/inferior (SI), and 0.2 (0.7) mm in right/left (RL) directions. For long-term reproducibility, the mean (SD) were -1.3 (3.1) mm AP, -0.5 (3.8) mm SI, and 0.3 (1.6) mm RL for inhale and -0.2 (2.8) mm AP, 0.2 (2.1) mm SI, and -0.7 (1.1) mm RL for exhale. CONCLUSION: The ABC device demonstrates very good short-term and long-term reproducibility. Increased long-term variability in position, primarily in the SI and AP directions, indicates the role of tumor-directed localization in combination with breath-held immobilization.