Abstract Measuring the implementation fidelity (IF) or integrity of interventions is extremely important, since without it a positive or negative outcome cannot be interpreted. However, IF is actually measured relatively rarely. Direct and indirect methods of measurement have been used in the past, but tend to over-emphasize teacher behaviour. This paper focuses on student behaviour collated through computers - an interesting alternative. It deals with the reading of real books and reading achievement, for which variables a very large amount of computerised data was available – on 852,295 students in 3243 schools. Reading achievement was measured pre-post with STAR Reading, a computerised item-banked adaptive norm-referenced test of reading comprehension. IF came from the Accelerated Reader (AR), which measures understanding of independent reading of real books the student has chosen by a quiz. Results showed higher IF was related to higher achievement. Neither IF nor reading achievement related to socio-economic status. Primary (elementary) schools had higher IF and achievement than secondary (high) schools. Females had higher IF and achievement than males. Students of higher reading ability implemented AR at a higher level, but did not gain in reading at a higher level. However, this computerised method of measuring IF with book reading showed limited reliability, no greater than methods emphasising teacher behaviour. In future, IF measures emphasising student response and those emphasising teacher behaviour need to be blended, although the latter will never generate the sample size of the former. This may be true of implementation fidelity in areas other than book reading.