Generalization Cannot Predict Abstract-Concept Learning

Previously thought to be unique to humans, abstract-concept learning has been demonstrated in a variety of species spread across the phylogenetic tree. A parameter important to abstract-concept learning is the number of training exemplars. For numerous species, increasing the number of training exemplars of the concept facilitates full transfer to novel stimuli. However, the number of training exemplars required to fully acquire the concept can vary between species. These findings have offered support for Darwin's assertion that cognitive differences between species are matters of degree rather than matters of kind. In light of these findings, some concern has grown over whether stimulus generalization can account for the functional relationship between training exemplars and transfer performance. The current chapter examined the viability of the generalization hypothesis for solving the non-matching-to-sample task. We tested the generalization hypothesis using a pre-existing model to simulate rates of acquisition (Wright & Katz, 2007). Our results indicate generalization cannot account for acquisition or transfer of non-matching abstract-concept learning. These findings further demonstrate the importance of training exemplars on concept formation.