We investigated whether low intensity dawn-dusk simulation (DDS), a 'naturalistic' form of light therapy designed to embed sleep in its accustomed phase, could improve the disturbed circadian rest-activity cycle, nocturnal sleep and and/or cognitive functions in dementia. A protocol of 3 weeks each of baseline, treatment and follow-up was completed by 13 patients (85yr old+/-5yr, MMSE 14+/-5; n=9 DDS versus n=4 'placebo' dim red light) who wore an activity/lux monitor throughout. There were no significant changes in clinical or cognitive status, nor modification of circadian stability or amplitude characteristics of the rest-activity cycle. However, two aspects of sleep responded to DDS but not to dim red light. The main sleep episode was 1:14h earlier during treatment (p=0.03) compared with before and after DDS. With respect to actimetry-determined sleep variables, the DDS group tended to have shortened 'sleep latency', longer 'sleep duration', more nocturnal immobility and less nocturnal activity than the dim red group (p<0.1). In parallel, nighttime light exposure tended to be reduced (p=0.07). These promising findings-after only 3 weeks of light treatment in elderly patients with advanced dementia-suggest that the circadian timing system remains functionally responsive even to low intensity DDS light. Increasing zeitgeber strength is an important strategy for improving sleep quality and timing in dementia, and DDS light therapy may provide one of the appropriate means to do so.