Omeract IV started a discussion on the development of radiological response criteria in rheumatoid arthritis (RA). Such criteria depend on the definition of what constitutes the minimum clinically important progression of damage. Because such a definition is currently not available, as a first step we have used the concept of random measurement error to determine what is the smallest detectable difference (SDD) in radiological progression between 2 radiographs of a particular patient. Baseline and 12 month radiographs (hands, wrists, feet) of 52 patients representative of the spectrum of radiological progression were selected from a randomized controlled trial of early rheumatoid arthritis (COBRA study) and were read paired and chronologically by 2 observers using the van der Heijde modified Sharp method (0-448 scale) and another 2 observers using the Scott modified Larsen method (0-200). The measurement error of progression was determined using the metric 95% limits of agreement method of Bland and Altman. In the setting of early RA the SDD is 11 modified Sharp score units and 8 modified Larsen score units if there is an equal distribution of baseline damage and progression in the sample and the mean score of the same trained observers is always used. The SDD is 15.5 modified Sharp score units and 11 modified Larsen score units if there is an equal distribution of baseline damage and progression in the sample and the mean score of any 2 trained observers is used. Other SDD were determined depending on the context of measurement. Although this exercise needs repetition in other settings, the SDD is a useful starting point in the development of radiological response criteria.