The reliability of a paralleling instrument for dental radiographs was retrospectively evaluated in the course of a 6-month clinical study. The angular variation between successive exposures was precisely quantified. Ninety-one percent of the angular variations were below a 1.4 degree threshold, but some paired images were found unsuitable for quantitative analysis. A mathematic procedure to assess the global angular projection error of any similar coupled x-ray cone-film holder device is proposed. This mathematic procedure should be routinely used to check superimposability of serial radiographs before density quantification or subtraction.