Objective: Common practice in the statistical comparison of imaging instruments with limited reproducibility consists in the seperate estimation of the instruments' reliabilities. However, as soon as one of the imaging methods under consideration underlies item-specific bias (which has to be expected for many clinical imaging procedures), this approach will end in severe errors in reliability computation and the corresponding erroneous clinical conclusions. This paper seeks to point out this effects and to illustrate a more appropriate model for the comparison of instrumental reliabilities. Material and Methods: A reliability model is adjusted for item--specific bias. Its application is illustrated by the comparison of twice repeated x-ray based cephalometry versus twice repeated non-ivasive orthodontic video imaging in 50 children; the posterior face height is used for illustration. Results: The proposed model allows to detect remarkable inferiority of the video-based imaging system as compared to the x-ray based standard. Analysis using seperate estimation would result in even adverse conclusions and falsely establish video imaging as a more reliable and thus superior diagnostic method. Conclusion: The reliabilities of clinical imaging methods have to be adjusted for possible item-specific bias to avoid the result of erroneous superiority of a diagnostic novum. Keywords: digital dental radiography, CCD sensor, storage phosphor, dose reduction