Objective: To identify factors associated with long-term improvement in gait in children after selective dorsal rhizotomy (SDR). Design: Retrospective cohort study. Setting: University medical center. Participants: Children (N=36) (age 4-13y) with spastic diplegia of Gross Motor Function Classification System (GMFCS) level I (n=14), II (n=15), and III (n=7) were included retrospectively from the database of our hospital. Children underwent SDR between January 1999 and May 2011. Patients were included if they received clinical gait analysis before and 5 years post-SDR, age >4 years at time of SDR and if brain magnetic resonance imaging (MRI) scan was available. Intervention: Selective dorsal rhizotomy. Main Outcome Measures: Overall gait quality was assessed with Edinburgh visual gait score (EVGS), before and 5 years after SDR. In addition, knee and ankle angles at initial contact and midstance were evaluated. To identify predictors for gait improvement, several factors were evaluated including functional mobility level GMFCS, presence of white matter abnormalities on brain MRI, and selective motor control during gait (synergy analysis). Results: Overall gait quality improved after SDR, with a large variation between patients. Multiple linear regression analysis revealed that worse score on EVGS and better GMFCS were independently related to gait improvement. Gait improved more in children with GMFCS I and II compared to III. No differences were observed between children with or without white matter abnormalities on brain MRI. Selective motor control during gait was predictive for improvement of knee angle at initial contact and midstance, but not for EVGS. Conclusion: Functional mobility level and baseline gait quality are both important factors to predict gait outcomes after SDR. If candidates are well selected, SDR can be a successful intervention to improve gait both in children with brain MRI abnormalities as well as other causes of spastic diplegia.