After products were selected, the study team began recruiting school districts to participate. The team focused on school districts that had low student achievement and large proportions of students in poverty, but these were general guidelines rather than strict eligibility criteria. The study sought districts and schools that did not already use products like those in the study so that there would be a contrast between the use of technology in treatment and control classrooms. Product vendors suggested many of the districts that ultimately participated in the study. Others had previously participated in studies with MPR or learned of the study from news articles and contacted MPR to express interest.
Interested districts identified schools for the study that fell within the guidelines. Generally, schools were identified by senior district staff based on broad considerations, such as whether schools had adequate technology infrastructure and whether schools were participating in other initiatives. By September 2004, the study had recruited 33 districts and 132 schools to participate. Five districts elected to implement products in two or more grade levels, and one district decided to implement a product in all four grade levels, resulting in 45 combinations of districts and product implementations. Districts and schools in the study had higher-than-average poverty levels and minority student populations (see Table 1).
To implement the experimental design, the study team randomly assigned volunteering teachers in participating schools to use products (the "treatment group") or not (the "control group"). Because of the experimental design, teachers in the treatment and control groups were expected to be equivalent, on average, except that one group is using one of the study's technology products. Aspects of teaching that are difficult or impossible to observe, such as a teacher's ability to motivate students to learn, are "controlled" by the experimental design because teachers were randomly assigned, and therefore should be the same in both groups, on average. The study also used statistical methods to adjust for remaining differences in measured characteristics of schools, teachers, and students, which arise because of sampling variability.
The experimental design provides a basis for understanding whether software products improve achievement. Teachers in the treatment group were to implement a designated product as part of their reading or math instruction. Teachers in the control group were to teach reading or math as they would have normally, possibly using technology products already available to them. Because the only difference on average between groups is whether teachers were assigned to use study products, test-score differences could be attributed to being assigned to use a product, after allowing for sampling variability.
Because the study implemented products in real schools and with teachers who had not used the products, the findings provide a sense of product effectiveness under real-world conditions of use. While the study worked to ensure that teachers received appropriate training on using products and that technology infrastructures were adequate, vendors rather than the study team were responsible for providing technical assistance and for working with schools and teachers to encourage them to use products more or use them differently. Teachers could decide to stop using products if they believed products were ineffective or difficult to use, or could use products in ways that vendors may not have intended. Because of this feature of the study, the results relate to conditions of use that schools and districts would face if they were purchasing products on their own.