The third evaluation is reported with much more detail. It was based on over 20,000 children, grades 1-9, from 76 schools. The average gain was five NCE's, but nearly all of this was in the early grades: "Grades one through four had the highest growth (29,9,4, and 5 percentiles respectively). Grades five through nine had less growth (0,-2,2,1, and 2 percentiles respectively)" (p. 2).
As was the case in earlier evaluations, students in schools in which teachers had more AR training made better gains. But those in schools in which teachers had more training also read more.
Some students had AR for three years (n = 401). They gained seven NCE's, but the average grade level of the books they read increased only from 3.1 to 4.5, that is, 1.4 years in three years (tables 13, 14). Some of these students were in schools that had "less than full implementation" of AR (p. 12).
Renaissance provides data on the average time spent reading for each grade level, as well as their estimate of the percentage of time spent reading that is above the average for that grade level. For example, first graders averaged 17 minutes of reading per day, while fifth graders averaged 24 minutes. But that 17 minutes was more than double the time the average first grader reads (240%) while the 24 minutes for the fifth grade was 88% of the average for fifth graders. I calculated that the number of minutes spent reading did not predict gains (r = -.051) but the percentage above average was a strong predictor of gains (r = .911). This makes sense when one hypothesizes that gains are clearly to related to amount read; norms on tests are based on average performance at that grade level.
Table 7 summarizes the results of studies using standardized tests without control groups
Table 7: Summary of studies using standardized tests, no comparison group