In this paper, we use a unique matched worker–workplace data set to estimate the effect on own earnings of co-workers’ education. Our results, using the 1998 GB Workplace Employee Relations Survey, show significant effects. An independent, significantly positive effect from average workplace education is evident; own earnings premia from years of education fall only slightly when controlling for workplace education. This result suggests that the social returns to education are strongly positive – working with colleagues who each had 1.2 years (1 standard deviation) of more education than the average worker, boosts own earnings by 11.1%. An additional year of any single co-worker's education is worth about 3.2% of an additional own year of education. We also test for interactions between own and co-worker education levels and for ‘skills incompatibility’ when worker education levels are heterogeneous. The interactions appear negative: own education is not much valued at workplaces where co-workers’ education levels are already high. There is no evidence that workplace heterogeneity in worker education levels adversely affects own earnings. This result runs counter to theoretical predictions, and suggests that workers compete in tournaments for high-paying jobs.