Peer-led small-group learning has been used quite extensively in the US to enhance performance and retention of undergraduate students in science, math, and engineering classes. This study presents the results from an evaluation of a peer-led small-group programme at a research university in the US over a 10-year period across five disciplines (biology, chemistry, engineering, physics, and mathematics) and seven courses. Data suggest the programme had a positive impact on participants' grades in five of the seven courses and on retention in the four courses that require students to take a course sequence. Effects of the programme were investigated across gender and ethnic groups. Participants benefited from the programme regardless of their gender or ethnicity. However, effect sizes were often larger for students from underrepresented groups. This was particularly true for course retention, where effect sizes for females were larger than those for males in four courses.