A Connectionist Semantic Network Modeling the Influence of Category Member Distance on Induction Strength


We present a model for inductive inference when both the premises and the conclusion are categorical. The phenomenon under investigation is that less similar categories in the premises lead to stronger conclusions. The model is based on the Rumelhart semantic connectionist network (Rogers & McClelland, 2004, 2008). Simulations addressed the main phenomenon and nine additional non-trivial phenomena of categorical induction (from Osherson, Smith, Wilkie, Lopez, & Shafir, 1990), providing support to the majority of the hypotheses.