Most geometric morphometric studies are underpinned by sets of photographs of specimens. The camera lens distorts the images it takes, and the extent of the distortion will depend on factors such as the make and model of the lens and camera and user-controlled variation such as the zoom of the lens. Any study that uses populations of geometric data digitized from photographs will have shape variation introduced into the data set simply by the photographic process. We illustrate the nature and magnitude of this error using a 30-specimen data set of Recent New Zealand Mactridae (Mollusca: Bivalvia), using only a single camera and camera lens with four different photographic setups. We then illustrate the use of retrodeformation in Adobe Photoshop and test the magnitude of the variation in the data set using multivariate Procrustes analysis of variance. The effect of photographic method on the variance in the data set is significant, systematic, and predictable and, if not accounted for, could lead to misleading results, suggest clustering of specimens in ordinations that has no biological basis, or induce artificial oversplitting of taxa. Recommendations to minimize and quantify distortion include: (1) that studies avoid mixing data sets from different cameras, lenses, or photographic setups; (2) that studies avoid placing specimens or scale bars near the edges of the photographs; (3) that the same camera settings are maintained (as much as practical) for every image in a data set; (4) that care is taken when using full-frame cameras; and (5) that a reference grid is used to correct for or quantify distortion.