BACKGROUND Pediatric pedestrian injuries are a major health care concern, specifically in urban centers. An educational program (WalkSafe), given one time during the school year, has been shown to improve childhood pedestrian safety. We examined whether this program could create similar long-term cognitive and behavioral changes in our school-aged children. METHODS An established pediatric pedestrian curriculum was modified slightly for use in our area. Students K-fourth grade were exposed to the program once annually for 2 years. The program was carried out weekly for 3 consecutive weeks. The first and third sessions consisted of an educational module given by the classroom teacher. The second week consisted of an interactive assembly that allowed the children to demonstrate good pedestrian safety using a simulated street. Short- and intermediate-term cognitive knowledge was evaluated using standardized pre-, post- and 3-month follow-up tests. Long-term knowledge was assessed by comparing scores as students advanced in grade from year 1 to 2 of the program (K to first, first to second, etc.). At six schools during year 2, pedestrian behavior was measured through direct observation of children on city streets before and after administering the program. The project was approved by university and school board institutional review boards. RESULTS During the 2 years, 1,564 students from nine schools were educated. In both years of the program, students in all grades had a significant gain in test scores immediately after and at 3 months compared with baseline knowledge. In contrast, only students moving from grade 3 to 4 demonstrated long-term retention (K→1: 7.7 vs. 6.7; grade 1→2: 7.8 vs. 6.7; grade 2→3: 7.3 vs. 6.8; grade 3→4: 7.1 vs. 8.0; all p < 0.05 year 2 pretest vs. year 1 3-month posttest; analysis of variance and generalized linear model). Only 30% of children walk with an adult. Direct observation showed 64% of children stopped at the curb but only 8% looked left-right-left. Children walking alone were more likely to cross mid-block compared with those walking with an adult (12% vs. 3%; p < 0.001) and also tend to look left-right-left significantly more than those walking with an adult (67% vs. 20%; p < 0.0001). CONCLUSIONS A one-time annual educational program resulted in long-term knowledge retention between grades 3 and 4 only. In contrast, scores in younger grades reverted to baseline pretest values seen in year 1. Short- and intermediate-term knowledge gains were seen in all grades for both years. Because older children more often walk alone, we postulate that the improved retention may be the result of repeated exposure and practice as a pedestrian. Cognitive knowledge did not appear to translate into improved pedestrian behavior. Walking with an adult also had a negative impact on observed pedestrian safety behavior. The efficacy and impact of a one-time educational program may be insufficient to change long-term behavior and must be reevaluated.