In a four-year study of the effect of class size on pupil outcomes in a sample of 36 primary schools in Hong Kong, it has been found that there are few positive differences in attainment between classes set at less than 25 pupils and those of normal size averaging 38. Three cohorts of pupils were studied. In Cohort 1 pupils spent 3 years in small classes (from P1 to P3) while in Cohort 2 they spent two years (P1 and P2) before reverting to normal size in P3. Cohort 3 began in normal classes in P1 and continued to do so in the following P2 year. A series of multi-level models have been used to explore the effect on academic performance. Four different analyses are presented. In the first, pupils who spent three years in a normal class were compared with pupils from the same schools who spent two years in a small class before reverting to a normal class in Primary 3. The second analysis compared pupils in their third year of small class teaching with those in the same age group in normal classes. In the third comparison being in a small class for two years was compared with being in a normal class for a similar period. The final comparison was between being in a small class for 3 years and being in a small class for only two years before reverting to a normal class in P3. Overall, the analysis showed no significant, consistent average effect either against sustained small class teaching or for it. However, in some schools remaining in small classes for three years led to a gain in attainment. The corollary to this is that for a similar number of schools there was a negative gain when classes return to normal in the third year of primary school. In all analyses boys tended to perform less well than girls. Motivation towards the study of Chinese, English and mathematics was always positively associated with attainment. (Contains 10 tables.)