(A) Definition. If a perturbation in can result in a change in future values of , then causes . This definition does not require that any perturbation in will perturb . For example, if the effect of on has saturated, then a further increase in will not affect . In this article, causality is represented by a hollow arrow. To embody probabilistic thinking (e.g. drunk driving increases the chance of car accidents; Pearl, 2000), and are depicted as histograms. Sometimes, perturbations in one variable can change the current value of another variable if, for example, the two variables are linked by a conservation law (e.g. conservation of energy). Some have argued that these are also causal relationships (Woodward, 2016). (B) Direct versus indirect causality. The direct causers of are given by the minimal set of variables such that once the entire set is fixed, no other variables can cause . Here, three variables , , and activate . The set constitutes the direct causers of (or ’s ‘parents‘ [Hausman and Woodward, 1999; Pearl, 2000]), since if we fix both and , then becomes independent of . If a causer is not direct, we say that it is indirect. Whether a causer is direct or indirect can depend on the scope of included variables. For example, suppose that yeast releases acetate, and acetate inhibits the growth of bacteria. If acetate is not in our scope, then yeast density has a direct causal effect on bacterial density. Conversely, if acetate is included in our scope, then acetate (but not yeast) is the direct causer of bacterial density (since fixing acetate concentration would fix bacterial growth regardless of yeast density). When we draw interaction networks with more than two variables, hollow arrows between variables denote direct causation.