I propose a “covariate-informed parallel design” for identifying causal mechanisms. This new experimental design does not require the assumption of no treatment-by-mediator interaction required by the parallel designs and the assumption of no carry-over effect required by the cross-over designs that Kosuke Imai and his colleagues presented.
Hong, G. (2013). Covariate-informed parallel design: Discussion on “experimental designs for identifying causal mechanisms” by Imai, Tingley, and Yamamoto. Journal of the Royal Statistical Society, Serial A, 176, 35.