Ultrasound imaging is widely used to measure architectural features of human skeletal muscles in vivo. We systematically reviewed studies of the reliability and validity of two-dimensional ultrasound measurement of muscle fascicle lengths or pennation angles in human skeletal muscles. A comprehensive search was conducted in June 2011. Thirty-six reliability studies and six validity studies met the inclusion criteria. Data from these studies indicate that ultrasound measurements of muscle fascicle lengths are reliable across a broad range of experimental conditions [intraclass correlation coefficient (ICC) and r values were always > 0.6, and coefficient of variation values were always < 10%]. The reliability of measurements of pennation angles is broadly similar (ICC and r values were always > 0.5 and coefficient of variation values were always < 14%). Data on validity are less extensive and probably less robust, but suggest that measurement of fascicle lengths and pennation angles are accurate (ICC > 0.7) under certain conditions, such as when large limb muscles are imaged in a relaxed state and the limb or joint remains stationary. Future studies on validity should consider ways to test for the validity of two-dimensional ultrasound imaging in contracted or moving muscles and the best method of probe alignment.