Methods for detecting recruitment patterns of the lumbar muscles during exercise in patients with chronic low back pain are limited. This article discusses the use of magnetic resonance imaging with Roman chair extension exercise to examine lumbar muscle usage in five normal volunteers, five chronic low back pain patients without surgery, and five chronic low back pain patients with surgery. Changes in signal intensities of psoas, multifidus, and longissimus/iliocostalis with graded exercise were measured at three lumbar disc levels. At rest, there was a difference between multifidus and longissimus/iliocostalis signal intensity in chronic low back pain subjects without surgery (P = 0.0162) and in chronic low back pain subjects with surgery (P = 0.0036), but not in normal subjects. At peak exercise, there was a difference in signal intensities between multifidus and longissimus/iliocostalis in all groups (normal volunteers, P = 0.0069; chronic low back pain patients without surgery, P = 0.0125; chronic low back pain patients with surgery, P = 0.0060). The exercise response was attenuated in chronic low back pain patients with surgery. Thus, MRI demonstrates static and dynamic differences in lumbar paraspinal musculature in chronic low back pain subjects compared to normal subjects.