PURPOSE: To investigate the Roll-Over Shape (RoS) of foot and ankle while walking with and without Ankle-Foot Orthoses (AFOs) with different degrees of stiffness in children with Cerebral Palsy (CP) walking with excessive knee flexion. METHODS: 15 children with CP walking with excessive knee flexion were prescribed a ventral shell AFO with a rigid footplate and integrated ankle hinge of which stiffness could be varied. All participants walked barefoot (BF), with shoes only (Sh) and with AFO in a rigid (rAFO), stiff (sAFO) and flexible (fAFO) setting. Centre of pressure (CoP) and shank kinematics of 3 steps of the most affected leg were collected. RoS was determined during single support, based on a circular fit of CoP data in the local reference frame of the shank. The radius (R) and arc length (Alength) were calculated and normalized for shank length. RESULTS: Circular RoS fitting was possible for all subjects in BF, while in the other conditions some data could not be fitted, especially for sAFO (n=7) and rAFO (n=9). Compared to BF and Sh, mean(SD) R and Alength increased significantly with an AFO (R: BF=0.32(0.18); Sh=0.31(0.19); fAFO=0.51(0.31); sAFO=0.46(0.15); rAFO=0.41(0.32), Alength: BF=0.22(0.10); Sh=0.21(0.07); fAFO=0.24(0.08); sAFO=0.26(0.06); rAFO=0.31(0.06)). RoS was translated backward in all conditions relative to the barefoot condition. CONCLUSIONS: The backward shift of the RoS and increased R and Alength while walking with AFOs is related to a decreased inclination of the shank. A larger R could thus indicate better shank alignment during walking. However, RoS could not be fitted in all cases, suggesting an absence of a circular RoS. As a result of the rigid footplate and restricted ankle range of motion, the second rocker motion and roll-over of the foot is lost in some cases with an AFO. Our data thus suggests that RoS analysis could be informative of AFO behaviour and alignment in children with CP.