Aim To evaluate the relative performance of ProTaper nickel-titanium (Ni-Ti) instruments shaping root canals of varying preoperative canal geometry. Methodology Extracted human maxillary molars were scanned, before and after shaping with ProTaper, employing micro computed tomography (muCT) at a resolution of 36 mum. Canals were three-dimensionally reconstructed and evaluated for volume, surface area, 'thickness' (diameter), canal transportation and prepared surface. Based on median canal volume, canals were divided into 'wide' and 'constricted' groups. Comparisons were made between mesiobuccal (mb), distobuccal (db) and palatal (p), as well as 'wide' and 'constricted' canals, using repeated-measures anova and Scheffe posthoc tests. Results Volume and surface area increased significantly and similarly in mb, db and p canals, and gross preparation errors were found infrequently. Root canal diameters, 5-mm coronal to the apex, increased from 0.38 to 0.65 mm, 0.42 to 0.66 mm and 0.57 to 0.79 mm for mb, db and p canals, respectively. Apical canal transportation ranged from 0.02 to 0.40 mm and was independent of canal type; 'wide' canals had a significantly higher (P < 0.05) proportion of unprepared surfaces than 'constricted' canals. Conclusions Canals in maxillary molars were prepared in vitro using ProTaper instruments without major procedural errors. These instruments may be more effective in shaping narrow canals than wider, immature ones.