Magnetic resonance imaging was used to determine the cross-sectional areas and the T2-weighted and proton density-weighted signal intensities of the paraspinal muscles in a group of 128 men, aged 35-63, who had varied histories of occupational and leisure-time physical activities. These measures, and the isokinetic lifting, psychophysical lifting, and static back muscle endurance tests were examined as predictors of low back pain over 12 months of follow-up, in the 43 men who reported no low back pain in the year preceding testing. None of the imaging measures or the muscle function tests was useful as a predictor of future low back pain. Associations with the frequency of low back pain before testing were investigated in the larger group. Smaller total cross-sectional area of the paraspinal muscles and greater signal intensities had weak but significant correlations with more frequent low back pain in the previous year, possibly due to muscle atrophy.