Most published brain studies are performed with too few participants to yield reliable findings, researchers suggest, in Nature.