The aim of this study was to evaluate the remaining dentine/cementum thickness using Gates-Glidden burs in serial and crown-down sequences and to observe which of the two sequences is the safest for preparing mesial roots of molars. Thirty-six left and right human mandibular first molars were selected. Standard access cavities were made and initially explored with Flexofiles sizes 10 and 15 until the tip was visible at the apex. The teeth were embedded in a muffle specially developed for this study using a PVC tube with two parallel metal rods in its lid. Each tooth-block was sectioned 3 mm apically to the furcation using a low-speed saw with a diamond disc. The tooth-block was examined under a microscope and an initial image was captured by a digital video system with 8 X and 12 X magnifications. Finally, the tooth-blocks were reassembled in the muffle so that the canals could be instrumented. After instrumentation the area of each mesial canal as well as the smallest distance to the root furcation were measured again. The mesio-buccal canals (crown-down order) and the mesio-lingual canals (serial sequence) presented an average area of 0.46 +/- 0.16 mm(2) and 0.88 +/- 0.27 mm(2) (P < 0.01), respectively. The mean values of the smallest distance to the furcation for the mesio-buccal and mesio-lingual canals were 0.66 +/- 0.19 mm and 0.39 +/- 0.13 mm (P < 0.01), respectively. The remaining dentine/cementum thickness using Gates-Glidden burs was greater in the crown-down sequence than in the serial sequence.