Traditional approaches to teaching and learning in large undergraduate STEM lectures has not kept pace with the massification and diversification of higher education. Efforts to alleviate learning obstacles call for improved instruction, stipulating the utility of specific pedagogical techniques delineated as "promising practices"; however, little evidence supports their effectiveness. In this study, a quasi-experimental method is applied to a large panel dataset containing course observations and institutional records to investigate the effect of promising practices on student learning. Additionally, differential effects are examined across several important nontraditional and historically under-served populations, such as Hispanic, first-generation, low-income and lower prior ability students. Results suggest that two information relaying techniques--the use of prior content and reviewing exam content--were positively and significantly associated with student outcomes, and estimates were found to be stable across subgroups. Implications for higher education teaching and learning, as well as empirical research in these areas, are discussed.