The one-to-one constraint in analogical mapping and inference

Theories of analogical reasoning have assumed that a 1-to-1 constraint discourages reasoners from mapping a single element in 1 analog to multiple elements in another. Empirical evidence suggests that reasoners sometimes appear to violate the 1-to-1 constraint when asked to generate mappings, yet virtually never violate it when asked to generate analogical inferences. However, few studies have examined analogical inferences based on nonisomorphic analogs, and their conclusions are suspect due to methodological problems. We sought to elicit mixed inferences that could result from combining information from 2 possible mappings. Participants generated 2-to-1 correspondences when asked for explicit mappings, but did not produce mixed inferences. Multiple correspondences appear to arise from multiple isomorphic mappings, rather than from a single homomorphic mapping.