The Causal Markov Condition implies the generalized version of RCCP presented in Section 2. This is most easily seen using the d-separation version. If variables X and Y are correlated, then \(\{X\}\) and \(\{Y\}\) are not d-separated by the empty set. This means that at least one path between X and Y must be without a collider. If we further assume that neither variable is a cause of the other, then there is no directed path between them. It then follows that the collider-free path must conta