This study examined learning gains among humanities and science students using assessment scores from a large sample of over 17,000 students. The researchers found that science students showed more variance in initial achievement and learning gains compared to social science students. Socio-demographic factors like gender, ethnicity and prior education explained more variance in learning gains for science students. The study also compared learning gains across universities, finding substantial differences at the departmental level, suggesting that aggregate estimates can be misleading. Multilevel modeling provided a more accurate method for estimating learning gains compared to simple linear models.