Jet lag degrades performance and operational readiness of recently deployed military personnel and other travelers. The objective of the studies reported here was to determine, using a narrow bandwidth light tower (500 nm), the optimum timing of light treatment to hasten adaptive circadian phase advance and delay. Three counterbalanced treatment order, repeated measures studies were conducted to compare melatonin suppression and phase shift across multiple light treatment timings. In Experiment 1, 14 normal healthy volunteers (8 men/6 women) aged 34.9+/-8.2 yrs (mean+/-SD) underwent light treatment at the following times: A) 06:00 to 07:00 h, B) 05:30 to 07:30 h, and C) 09:00 to 10:00 h (active control). In Experiment 2, 13 normal healthy subjects (7 men/6 women) aged 35.6+/-6.9 yrs, underwent light treatment at each of the following times: A) 06:00 to 07:00 h, B) 07:00 to 08:00 h, C) 08:00 to 09:00 h, and a no-light control session (D) from 07:00 to 08:00 h. In Experiment 3, 10 normal healthy subjects (6 men/4 women) aged 37.0+/-7.7 yrs underwent light treatment at the following times: A) 02:00 to 03:00 h, B) 02:30 to 03:30 h, and C) 03:00 to 04:00 h, with a no-light control (D) from 02:30 to 03:30 h. Dim light melatonin onset (DLMO) was established by two methods: when salivary melatonin levels exceeded a 1.0 pg/ml threshold, and when salivary melatonin levels exceeded three times the 0.9 pg/ml sensitivity of the radioimmunoasssy. Using the 1.0 pg/ml DLMO, significant phase advances were found in Experiment 1 for conditions A (p < .028) and B (p < 0.004). Experiment 2 showed significant phase advances in conditions A (p < 0.018) and B (p < 0.003) but not C (p < 0.23), relative to condition D. In Experiment 3, only condition B (p < 0.035) provided a significant phase delay relative to condition D. Similar but generally smaller phase shifts were found with the 2.7 pg/ml DLMO method. This threshold was used to analyze phase shifts against circadian time of the start of light treatment for all three experiments. The best fit curve applied to these data (R(2) = 0.94) provided a partial phase-response curve with maximum advance at approximately 9-11 h and maximum delay at approximately 5-6 h following DLMO. These data suggest largest phase advances will result when light treatment is started between 06:00 and 08:00 h, and greatest phase delays will result from light treatment started between 02:00 to 03:00 h in entrained subjects with a regular sleep wake cycle (23:00 to 07:00 h).