This study investigated the effect of class practices on students' learning gains in reading comprehension in the 5th grade. A sample of 4,344 students in 283 classes in 176 schools was studied. Several class practices that have previously been demonstrated to be effective were tested while controlling for student characteristics and socioeconomic and ethnic class composition. Differential effects were tested to identify class practices that can contribute to narrowing the achievement gap between high- and low-risk students. Most class practices turned out to have a similar effect for both low- and high-risk students. However, "discovery learning" and "well-organised and attractive instruction" appeared to be more beneficial for low-risk than for high-risk students. Group composition in terms of social and ethnic background turned out to have no significant effect on learning gains in reading comprehension.