This study investigated the reliability of the modified Tardieu scale in the assessment of biceps spasticity in the upper limbs of children with hemiplegic cerebral palsy (CP). Ten children, with hemiplegic CP participated in the study: six males (mean age 9 years, SD 4 years) and four females (mean age 12 years, SD 3 years). Blinded, duplicate measures of dynamic elbow extension were performed on the hemiplegic arm at time 0 and 7 days later, using the three angular velocities described in the Tardieu scale (V1, slow; V2, speed of gravity; V3, as fast as possible). The resulting elbow joint angles were defined as R1, the angle of catch following a fast velocity stretch at either V2 or V3; and R2, the passive range of movement achieved following a slow velocity stretch at V1. Both elbow joint angle and movement angular velocity were measured by three‐dimensional kinematics. Median error in measured elbow joint angle within one session ranged from 3 to 5°. Between sessions median absolute differences in measured elbow joint angle ranged from 4 to 13°, with measurement errors of up to 25 to 30° in some participants at the fastest velocity (V3). The therapist was able to apply three significantly different angular velocities as required for the Tardieu scale (p < 0.001). However, the ranges of the three angular velocities overlapped, with fast velocities for some participants being equivalent to slow velocities for other participants. Three out of 10 participants had an intersessional difference in their R2‐R1 score of more than 20°. From this study, we concluded that the R2‐R1 value determined from the modified Tardieu scale may be of limited value in assessing biceps spasticity the upper limbs in children with hemiplegic CP.