The application of photogrammetry to the reconstruction of automotive accidents has been well documented and many techniques are available to the reconstructionist. However, the accuracy that can be expected from a photogrammetric project can vary by method and by application. Greater understanding of the factors affecting the accuracy of a photogrammetric reconstruction is desirable. Specific variables affecting the accuracy of 3-D, non-metric, analytical photogrammetry were studied using the commercially available software package, Eos Systems’ PhotoModeler. The 3-D co-ordinates of targets on a vehicle were first surveyed and then photographed with standard 35 mm camera equipment. The knowledge of camera properties, the method of image generation, photograph cropping, use of fiducial markings, and the number of control points were investigated to determine their relative effects on the accuracy of the solved coordinates. Of the factors studied, the number of control points and the amount of image cropping had the largest individual effects on the overall accuracy. The most accurate cases were those with many control points, no image cropping, camera calibration, use of Kodak Photo CD image generation, and use of fiducial markings.