Results
This research concerns the extent to which the students learned from participating in the
curriculum projects. Recall that the tests had two component scores, science content and process,
and a total score that was the sum of the two components. For the first set of analyses, we
aggregated across teachers and computed within-subject t-tests for the two component scores and
the total score. The results of these analyses are reported in Table 2.
All the analyses, with the exception of the process score on the 1998–1999 water unit, showed
statistically reliable gains. The effect sizes for these gains were stronger for content scores than
process scores. Moreover, the weighted average effect sizes for total, content and process scores
(calculated across curriculum projects) grew stronger across the 3 years (Figure 1). Specifically,
Figure 1 shows that for all three scores, the effect sizes were more robust in the second and third
years. For the content scores, the average student on the posttest in 1998–1999 achieved a
percentile score of 73 compared with the pretest distribution, but in 1999–2000 the posttest
percentile score was 88 and in 2000–2001 it was 91. For the process scores, the average student in
1998–1999 achieved a percentile score of 62 on the posttest, but in 1999–2000 the percentile
score was 69 and in 2000–2001 it was 72.
All the tests were constructed to have items at three cognitive levels: low, medium, and high.
We analyzed gains from pre- to posttests for all the projects on these three item types (Table 3). All
the gains were statistically reliable at p