A method is described for reconstructing the 3-dimensional coordinates of anatomical landmarks from multiple video images. This method consists of 2 stages: (1) separate reconstructions of landmark coordinates visible in dorsal and ventral orientations using Direct Linear Transformation (Abdel-Aziz & Karara, 1972), and (2) least-squares alignment of the partial reconstructions using landmarks present in both. The repeatability of the overall reconstructions and of individual landmarks were assessed by performing multiple independent reconstructions of landmark coordinates for 2 canid specimens. Results indicate that the repeatability of interlandmark distances based on stereophotogrammetric reconstructions from standard-resolution video images is comparable to that of caliper measurement, and slightly superior to that of electronic 3-dimensional digitisation. At least 4 images should be available per landmark for optimal repeatability, but additional images allow only a slight increase in repeatability.