The concept of legal validity is regarded within the dominant—legal-positivistic—account of law as a nongradable concept: a legal rule is either valid or nonvalid. However, this account of validity is criticized by some scholars as being too strict and rigid. Apparently, an attractive alternative might be Alf Ross's probabilistic account of validity. Ross assumed that the stronger the predictions of judicial behavior a given rule generates, the greater the probability of their being valid. However, this account of legal validity is by no means uncontroversial. In this paper, four objections to it are formulated: the objection of the apparent gradability, the objection of the problematic ascertainability, the objection of the normative insignificance of probabilistic information, and the objection of neglecting the normativity of legal rules. These objections are treated in the paper as providing strong reasons for rejecting Ross's claim that predictions of judicial behavior formulated on the basis of rules are conceptually linked to their validity (i.e., they define their meaning); it is argued that they are merely a way of testing empirical hypotheses concerning the application (effectiveness) of legal rules.