Sensitivity and specificity of lumbar spine radiography in the assessment of facet joint osteoarthritis were evaluated, with computed tomography (CT) as the standard. Two independent radiologists used a four-point scale to blindly grade facet joint osteoarthritis on oblique radiographs and transaxial CT scans obtained within an 8-month period in 50 consecutive patients with pain in the lower back. The L-3 to L-4, L-4 to L-5, and L-5 to S-1 facet joints were evaluated, and 68% appeared abnormal on CT scans, with 28% exhibiting moderate or severe disease. Interobserver agreement was high for conventional radiography (perfect agreement in 57% and agreement to within one grade in 39%) and still higher for CT (perfect in 63% and to within one grade in 35%). Receiver operating characteristic curve analysis indicated that oblique radiography was most accurate (55% sensitivity, 69% specificity) in distinguishing the presence from the absence of disease; in distinguishing absent or mild from moderate or severe disease, the specificity of oblique radiography was higher, at 94%, but its sensitivity was much lower, at 23%. Conventional radiography is a useful technique in screening for facet joint osteoarthritis but is insensitive compared with CT.