The effect of low back pain on the size of the lumbar multifidus muscle was examined using real-time ultrasound imaging. Bilateral scans were performed in 26 patients with acute unilateral low back pain (LBP) symptoms (aged 17–46 years) and 51 normal subjects (aged 19–32 years). In all patients, multifidus cross-sectional area (CSA) was measured from the 2nd to the 5th lumbar vertebrae (L2–5) and in six patients, that of S1 was also measured. In all normal subjects, CSA was measured at L4 and in 10 subjects measurements were made from L2–5. Marked asymmetry of multifidus CSA was seen in patients with the smaller muscle being on the side ipsilateral to symptoms (between-side difference 31 ± 8%), but this was confined to one vertebral level. Above and below this level of wasting, mean CSA differences were < 6%. In normal subjects, the mean differences were < 5% at all vertebral levels. The site of wasting in patients corresponded to the clinically determined level of symptoms in 24 of the 26 patients, but there was no correlations between the degree of asymetry and severity of symptoms. Patients had rounder muscles than normal subjects (measured by a shape ratio index), perhaps indicating muscle spasm. Linear measurements of multifidus cross-section were highly correlated with CSA in normal muscles but less so in wasted muscles, so CSA measurements are more accurate than linear dimensions. The fact that reduced CSA, i.e. wasting, was unilateral and isolated to one level suggests that the mechanism of wasting was not generalized disuse atrophy or spinal reflex inhibition. Inhibition due to perceived pain, via a ling loop reflex, which targeted the vertebral level of pathology to protect the damaged tissues was the likely mechanism of wasting. The lack of correlation between severity of wasting and symptoms highlights the need for objective measurement in the assessment of the paraspinal muscles in LBP.