There was little improvement with age across the explicit tasks used in Studies 1A1, 1A2, 1B1 and 1B2 (see Horizontal Motion and Object Fall). No age effects were detected in Study 1A1 and Study 1B2, and age-related improvements were limited to deceleration on a baize surface in Study 1A2 and fall from rest in Study 1B1. Yet across the age range covered in the studies, children receive teaching that is intended to support explicit understanding of precisely the same topics. Thus, there is a need for new approaches to instruction, and this was one rationale behind the software that was developed and evaluated in Studies 2A and 2B. Indeed, educational considerations motivated: a) the contrast in the studies between using the software under adult guidance (if necessary, demanding of teachers) and collaboratively with a classmate (if helpful, readily established in classrooms); b) the focus upon promoting explicit understanding, both the prediction of outcomes and the understanding of why outcomes occur.

Study 2A used the billiards scenarios from Studies 1A1 and 1A2 (see Figure 1), and Study 2B used the hot air balloon scenarios from Studies 1B1 and 1B2 (see Figure 4). The problems associated with each scenario began with what, in effect, was the explicit task in the 1A and 1B studies, i.e. the scenario froze at the point of white and red ball impact (Study 2A) or ball release from the balloon (Study 2B). The task was to predict the subsequent direction or speed, with 8 problems in each study addressing direction and 8 addressing speed. Of the contextual factors covered in the 1A studies, the scenarios in Study 2A presented all combinations of on- vs. off-centre impact, baize vs. glass surface, and questions about red vs. white ball. This time however force was always gentle. Like the 1B studies, the scenarios in Study 2B presented all combinations of fall from rest vs. fall after motion, and fall through air vs. fall through air plus water, but this time they were restricted to two of the three balls used in the earlier studies.

Once predictions had been made, the problems in each study proceeded as follows:

• If the prediction with a given scenario was correct, positive feedback was given, i.e. ‘Well done! You are correct’. This was accompanied with an invitation to see what happens, whereupon the relevant tacit correct scenario from the 1A and 1B studies was activated.
• If the prediction with a given scenario was incorrect, negative feedback was given, i.e. ‘You are incorrect’. This was accompanied with an invitation to see what would happen if the ball moved as predicted, whereupon the relevant tacit incorrect scenario from the 1A and 1B studies was activated. Upon completion, there was an invitation to view the correct motion, and compare with the incorrect, predicted motion. Correct and incorrect motion could be replayed indefinitely.

Studies 2A and 2B began with pre-tests to classes of 8- to 9-year-olds, 9- to 10-year-olds, and 10- to 12-year-olds (but only one 12-year-old in each study). In both studies, the tests involved 16 problems, all 16 requiring predictions of outcome but 8 also requiring identification of reasons for predictions (through selection from multiple-choice options – see Figures 7 and 8). The problems were presented to the whole class via PowerPoint slides supported with oral instruction. The children indicated their answers in booklets that reproduced the PowerPoint slides. The Study 2A test items covered bowls, golf, marbles and croquet as well as billiards, and the Study 2B items used helicopters and trains as carriers as well as balloons. A few days after the pre-test, about 33% of children from each class worked through the teaching software under adult guidance, about 33% worked collaboratively with a randomly chosen classmate, and about 33% did not use the software. (Because the aim of these particular studies was to establish whether the software in general was effective, ‘non-tutored’ controls were appropriate; future research will examine whether explicit-tacit engagement contributed specifically to the results through comparison with ‘differently taught’ controls). Several weeks later, the children were post-tested employing the same items as were used in the pre-test.

167 children (78 girls; Mean age = 9.51 years) completed all stages in Study 2A, and 139 children (77 girls; Mean age = 9.73 years) did this in Study 2B. At pre-test, the children in the three conditions (+adult; + peer; control) were equivalent. At post-test:

• In Study 2A, the +adult and +peer children made superior predictions to the controls, and the factors selected by the +adult children to justify their predictions contained fewer scientifically irrelevant options.
• In Study 2B, the +adult and +peer children made superior predictions to the controls, with the +peer children surpassing the +adult.

Figure 7: Sample item (Study 2A pre/post-test)

Figure 8: Sample item (Study 2B pre/post-test)