A large-scale randomised-controlled trial of reading tutoring in 80 schools in Scotland used the Paired Reading (PR) technique. On long-term evaluation, cross-age PR was significantly better than regular teaching, but same-age was not. On short-term evaluation, PR pupils did significantly better than control pupils in both years, and cross-age and same-age were similarly effective. Low socio-economic pupils, lower reading ability pupils, girls and reading with maths tutoring groups did significantly better. Implementation was good in parts, but some important aspects of technique were rare. Reading gains were significantly greater for those with mistakes about every 2 minutes and those who stopped reading to talk every 5 to 7 minutes. Significant gains in self-esteem were seen in same-age and cross-age groups, and for tutees and tutors, but not for controls. The relationship of achievement gain to quality of technique and socio-emotional gains is discussed, with implications for practice. (Contains 2 tables.)