The aim of this study was to investigate the reliability of hand-held dynamometry for measuring isometric lower-limb muscle strength in children with cerebral palsy (CP). Twenty-three children (14 males, nine females) with CP (spastic diplegia; Gross Motor Function Classification System Levels I-III) aged 5 years 7 months to 14 years 5 months (mean 9 y 6 mo [SD 2 y 8 mo]) attended two test sessions 1 week apart. A 'make' test, using a gradual build-up of force to a maximum isometric contraction, was employed and peak values were normalized to body weight for analyses. Within-session reliability was high with an intraclass correlation coefficient (ICC) of 1,1>0.79 for all muscle groups, and there was acceptable between-session reliability ICC>0.70 and measurement errors for hip flexors and extensors (measured in supine), knee flexors and extensors, and ankle dorsiflexors (with stabilization). Within- and between-session reliability was poor (ICC<0.70) for hip extensors (in prone), knee extensors (20 degrees flexion), ankle dorsiflexors (without stabilization), and ankle plantarflexors. Measurement error differed in each test and across limbs, with stabilization producing inconsistent reliability outcomes. Changes in strength measurements in children with CP should take into account measurement error for particular muscle groups. Changes should be determined for separate muscle groups and limbs, and reported relative to body weight. Different testing positions may be required for greater reliability.