Causal cognition is a key part of human learning, reasoning, and decision-making. In particular, people are capable of learning causal relations from data, and then reasoning and planning using those cognitive representations. While there has been significant normative work on the causal structures that ought to be learned from evidence, there has been relatively little on the functional forms that should (normatively) be used or learned for those qualitative causal relations. Moreover, empirical research on causal inference—learning causal relations from observations and interventions—has found support for multiple different functional forms for causal connections. This paper argues that a combination of conceptual and mathematical constraints leads to a privileged (default) functional form for causal relations. This privileged function is shown to provide a theoretical unification of the widely-used noisy-OR/AND models and linear models, thereby showing how they are complementary rather than competing. This unification thus helps to explain the diverse empirical results, as these different functional forms are “merely” special cases of the more general, more privileged function.