The purpose of this preregistered cluster randomized trial was to examine the effectiveness of a universal social-emotional learning program when implemented under routine conditions in second-grade classrooms. Thirty-nine teachers and 332 students from 13 elementary schools participated in the trial. Teachers randomly assigned to the treatment condition taught self-selected units from the SSIS Social-Emotional Learning Classwide Intervention Program (SSIS SEL CIP; Gresham & Elliott, 2017) while control teachers continued with business as usual. Teacher-completed rating scales and independent observations were used to assess students' positive and negative social behaviors, and a three-level random intercepts model was used to analyze the data. Results did not yield any statistically significant main effects, and most effect sizes were small in magnitude. Although direct observations of lesson implementation indicated a relatively high level of fidelity, the number of complete units and lessons varied across classrooms. Training and support also varied depending on each school's typical practices. Findings suggest that typical implementation practices and approaches to support delivery of universal SEL programs are less likely to yield student outcomes similar to those observed when such programs are implemented with higher levels of training, support, dosage, and adherence. [This paper will be published in "School Psychology."]