Inference processes in causal analogies

In recent papers, Lee & Holyoak (2007, 2008a, 2008b) argue that extant models of analogy fail to explain how people draw inferences from causal analogies. In the current research, we argue that structure-mapping theory sufficiently explains the analogical inferences drawn from these causal analogies, and that, contrary to L&H‘s claims, the effect inference can indeed be evaluated by a post-analogical causal reasoning process. In Study 1, we present evidence that – consistent with SMT (Gentner, 1983), and counter to L&H – when relational inferences are considered, the inductive strength of these causal analogies matches their similarity. In Study 2, we provide evidence that, by analogical mapping, the base analog makes two contributions to the reasoner‘s knowledge about the causal system in the target, and argue that this analogically-constructed causal model is subsequently used to determine the presence of the effect. In an SME (Falkenhainer et al., 1989) simulation, we show that ―outsourcing‖ the effect inference to a simple post-analogical calculation can match L&H‘s human data very closely. In short, although we agree with Lee & Holyoak that analogy is important for learning about causal systems, we maintain that analogy is a domain-general process. Models of analogical processing need not—and should not—subsume causal inferencing processes.