We report a consecutive series of 105 extractions of impacted lower third molars, in each of which operative difficulty was predicted preoperatively using the Pederson scale (radiographic appearance of the anatomical position of the lower third molar), and postoperative difficulty was scored with a modified version of the Parant scale (operative manoeuvres that were needed for extraction of the third molar). Preoperative classification as "difficult" on the Pederson scale was not an accurate predictor of true difficulty (postoperative classification as "difficult" on the modified Parant scale). There was no significant association between the Pederson score and duration of operation, but high Parant scores were significantly associated with longer operations. We suggest that scales for the prediction of operative difficulty in the extraction of impacted lower third molars should take into account factors other than the anatomical position of the tooth.