Clinical measurement, reliability study. To investigate the improvements in precision when averaging multiple measurements of percent change in muscle thickness of the transversus abdominis (TrA) and lumbar multifidus (LM) muscles. Although the reliability of TrA and LM muscle thickness measurements using rehabilitative ultrasound imaging (RUSI) is good, measurement error is often large relative to mean muscle thickness. Additionally, percent thickness change measures incorporate measurement error from both resting and contracted conditions. Thirty volunteers with nonspecific low back pain participated. Thickness measurements of the TrA and LM muscles were obtained using RUSI at rest and during standardized tasks. Percent thickness change was calculated with the formula thickness(contracted) - thickness(rest)/thickness(rest). Standard error of measurement (SEM) quantified precision when using 1 or a mean of 2 to 6 consecutive measurements. Compared to when using a single measurement, SEM of both the TrA and LM decreased by nearly 25% when using a mean of 2 measures, and by 50% when using the mean of 3 measures. Little precision was gained by averaging more than 3 measurements. When using RUSI to determine percent change in TrA and LM muscle thickness, intra examiner measurement precision appears to be optimized by using an average of 3 consecutive measurements.