Theories set forth by Pötzl and Fisher were subjected to replication under standardized conditions with a group of twenty-four medical students. They were shown tachistoscopic images for one one-hundredth of a second and it was then suggested that they dream. The testing of the hypothesis was much more rigorous and rigid than that done by Pötzl or Fisher. Judges using the dream complement hypothesis as a guide tried to pair the dream drawing with the original images for each of the subjects. The pairings occurred no more often than might have been expected on the basic random selection. An alternative hypothesis was tested that there would be a similar rather than a complementary relationship between the drawings and the dream report drawing. Again, pairing was on a chance basis. The authors discuss the possible sources of the disparities between the Pötzl and the Fisher findings and re-examine some of the implicit assumptions made by both the earlier papers and some of the unavoidable difficulties inherent in such research.