The Bath ankylosing spondylitis metrology index (BASMI; range 0-10) has gained widespread use in daily clinical practice as an objective measure of spinal stiffness not only in patients with ankylosing spondylitis (AS) but also in patients with other spondylarthropathies (SpA). We examined intra-rater and inter-rater reproducibility of BASMI scoring in 30 Danish patients with SpA (median age 40 years, range 22-56 years) fulfilling the European Spondylarthropathy Study Group criteria, 25 of them satisfying the modified New York Criteria for AS. Measurements were performed twice on two different days (median interval 7 days, range 4-11) by a trained physiotherapist (PT) and by an untrained nurse who had undergone a single 1-h training session with the PT. The median BASMI score obtained by the PT on the two test days was 3.5 (range 1-8) and 3.0 (range 1-8), respectively (NS). Test-retest BASMI scores from the PT were significantly correlated (r(s) = 0.95, p < 0.0001). The 95% likely range for the difference between a patient's BASMI scores from two tests was +/-1.4 corresponding to a minimal detectable difference of +/-2 in the individual patient as the scale consists of intervals of 1. Similar results were achieved by the nurse. BASMI scores obtained by the two raters were significantly inter-correlated (r(s) = 0.95, p < 0.0001). The mean difference between paired BASMI scores obtained by the nurse and the PT on test day 1 was -0.2 with a minimal detectable difference of +/-2. A similar result was found using data from test day 2. In conclusion, a change in BASMI less than 2 may be due solely to expected random measurement error. A single 1-h training session allowed an untrained nurse to obtain BASMI results almost identical to those of an experienced PT.