Hands on what? The relative effectiveness of physical versus virtual materials in an engineering design project by middle school children

56 seventh- and eighth-grade students

US middle school students (grades 7-8, ages ~12-14); engineering design task ('mousetrap cars').

Randomised assignment of individual students to physical vs virtual materials, holding the TASK, the instruction, the time on task and the outcome measures constant. Students designed, built and tested mousetrap cars to maximise distance travelled, then were tested on (a) design performance, (b) knowledge of which car features matter, and (c) transfer to a new design problem. The physical condition used real purchased components; the virtual condition used a deliberately sparse on-screen equivalent. This is a direct test of PHYSICALITY, not of practical work vs lecture — the alternative to hands-on here is hands-on-a-mouse, not sitting still. Small n (56) means only moderate-to-large differences would be detectable, so 'no difference' should be read as 'no large difference'.

The cleanest K-12 test of whether the PHYSICAL part of hands-on work is doing the work, and the answer is no. 56 grade 7-8 students were randomly assigned to build and test mousetrap cars with real components or with an intentionally impoverished on-screen simulation. Physical and virtual materials were equally effective on every measured outcome: design performance, knowledge of which features mattered, transfer, and students' own confidence and interest. Klahr's summary across this study and its elementary predecessor is that 'physical and virtual materials were equally effective' in every condition and phase. This matters enormously for the cost argument: if the learning survives replacing real apparatus with a sparse simulation, then the equipment, the technicians, the consumables and the purpose-built laboratory are buying something other than the measured learning. Caveats: small sample, single session, immediate outcomes, researcher-designed measures, and the task is engineering design rather than school science content.

Low — random assignment of individuals.

Replicates and extends Triona & Klahr 2003 (elementary, control-of-variables) upward to middle school and to an open engineering-design task. Further replicated at kindergarten age by Zacharia, Loizou & Papaevripidou 2012 and, in the higher-education literature, by Zacharia & Olympiou 2011 and the 50-of-56 tally in Brinson 2015. This is one of the better-replicated null results in science education.