Using students (especially undergrads) is a typical problem of programmer productivity studies. N=49 and 27h of experimentation time per subject isn't a lot either. Doing a study that can draw actually useful conclusions may just be prohibitievely expensive.