Chapter 5 introduces asymmetry properties of the independence assumption of predictors and errors obtained from causally competing models, and illustrates that, under non-Gaussianity of variables, reverse causation biases (e.g., erroneously entertaining a model of the form y → x instead of x → y) lead to systematic non-independence of model predictors and errors. Therefore, the degree of independence observed in causally competing models holds promise to derive conclusions concerning the causal structure of variable associations. Decision rules and methods for statistical inference are presented, and results of two Monte-Carlo simulation experiments are reported that illustrate the statistical behavior of independence statistics in the direction dependence context. Artificial and real-world data examples are used to demonstrate the application of independence-based DDA model selection.