Motion of the wrist bones is complicated and difficult to measure. Noninvasive measurement of carpal kinematics using medical images has become popular. This technique is difficult and most investigators employ custom software. The objective of this paper is to describe a validated methodology for measuring carpal kinematics from computed tomography (CT) scans using commercial software. Four cadaveric wrists were CT imaged in neutral, full flexion, and full extension. A registration block was attached to the distal radius and used to align the data sets from each position. From the CT data, triangulated surface models of the radius, lunate, and capitate bones were generated using commercial software. The surface models from each wrist position were read into engineering design software that was used to calculate the centroid (position) and principal mass moments of inertia (orientation) of (1) the capitate and lunate relative to the fixed radius and (2) the capitate relative to the lunate. These data were used to calculate the helical axis kinematics for the motions from neutral to extension and neutral to flexion. The kinematics were plotted in three dimensions using a data visualization software package. The accuracy of the method was quantified in a separate set of experiments in which an isolated capitate bone was subjected to two different known rotation/translation motions for ten trials each. For comparison to in vivo techniques, the error in distal radius surface matching was determined using the block technique as a gold standard. The motion that the lunate and capitate underwent was half that of the overall wrist flexion-extension range of motion. Individually, the capitate relative to the lunate and the lunate relative to the radius generally flexed or extended about 30 deg, while the entire wrist (capitate relative to radius) typically flexed or extended about 60 deg. Helical axis translations were small, ranging from 0.6 mm to 1.8 mm across all motions. The accuracy of the method was found to be within 1.4 mm and 0.5 deg (95% confidence intervals). The mean error in distal radius surface matching was 2.4 mm and 1.2 deg compared to the use of a registration block. Carpal kinematics measured using the described methodology were accurate, reproducible, and similar to findings of previous investigators. The use of commercially available software should broaden the access of researchers interested in measuring carpal kinematics using medical imaging.