We extend the result from Bossert and Sprumont (2013) that every single-valued choice function is backwards-induction rationalizable via strict preferences to the case of choice correspondences via weak preferences.