It is argued that in the absence of causal hypotheses the conditionalisation rule makes little sense. This claim is illustrated with some examples, in particular with an analysis of the car-ride paradox. It is demonstrated that in some situations Bayesian probability estimates in the context of inquiry appear absurd, even though mathematically they are perfectly correct. In other circumstances, however, the same estimates may be meaningful. It is argued that what makes the difference is the relevance of the evidence one conditionalises on. The relevance, in turn, depends on the presupposed causal mechanisms of the process in question, rather than statistical correlations. On the one hand, the analysis on offer supports the explanationist approach against the Bayesian and the compatibilist. On the other hand, it does not rebut conditionalisation altogether. Rather, it points to severe limitations of the method that is hardly applicable to universal, theoretical hypotheses. Consequently, it is suggested that the method be confined to calculating the probability estimates of predictions made on a specified body of hypotheses, evaluating statistical hypotheses that need not involve causal explanations, or some specific hypotheses of limited scope.