The problem of late mandibular incisor crowding is a well established phenomenon, the cause of which has been the substance of considerable debate over the years. A central issue is the possible role of the third molars though no definitive conclusions have been consistently drawn. This prospective study was designed to investigate the effects of randomly assigned early extraction of third molars on late crowding of the mandibular incisors. One-hundred-and-sixty-four patients entered the study from 1984 following completion of retention after orthodontic treatment. Seventy-seven patients (47%) returned for records up to a mean of 66 months later, and their start and finish study casts were digitized on a reflex microscope to determine Little's index of irregularity, intercanine width and arch length. Forty-four of the patients had been randomized to have third molars removed. There was no evidence of responder bias. Where third molars were extracted the mean increase in lower labial segment irregularity was reduced by 1.1 mm from a mean of 2.1 mm for the group where third molars were retained (P = 0.15, not statistically significant). This difference was also not considered to be clinically significant. The principal conclusion drawn from this randomized prospective study is that the removal of third molars to reduce or prevent late incisor crowding cannot be justified.