Phototherapy is regularly used in the treatment of seasonal affective disorder. There is evidence that it is also useful in the treatment of non-seasonal depression, but relevant controlled experiments are difficult to design. In this study we randomly assigned depressed in-patients to high and low levels of artificial light, the high levels exceeding those most commonly used in earlier reported trials. Both unipolar and bipolar depressions responded when phototherapy was used as an adjunct to pharmacotherapy, and mood improvement was related to the intensity of illumination, that is, patients treated with high levels of illumination improved significantly more than those who received low levels (P < 0.02). Our findings suggest that light therapy is generally applicable to depressive illnesses, and that the light intensities commonly used are suboptimal.