It is shown that the before-before (or Suarez-Scarani) experiment refutes hidden variable models with a deterministic (“realistic”) nonlocal part, whereas experiments violating Leggett-type inequalities refute models with biased random local part. Therefore the claim that Gröblacher et al. (Nature 446:871–875, 2007) present “an experimental test of nonlocal realism” is misleading, and Marek Żukowski’s (Found. Phys. 38:1070, 2008) comment misses the point. A new experiment is suggested.