and generalization go hand-in-hand. All fields in psychology dealing
with learning and inference have also explored generalization. We study
both dimensional generalization, or how learning about a stimulus is
transferred to new stimuli that differ from the original along
continuous dimensions, and compound generalization, or how learning
about one stimulus is transferred to new compounds comprising that
In the past, these two forms of generalization have been
studied largely independently, and researchers have shown little
interest in developing a unified theoretical framework to understand
both. We have recently developed just such a unified framework, by
extending Shepard’s rational theory of dimensional generalization to
the explanation of compound generalization. The model explains many
results from the literature on causal and associative learning.
However, to do so, it assumes a particular hypothesis about what
distinguishes separable from non-separable (integral) dimensions, which
is different from the explanation originally favored by Shepard. We
have recently tested these two hypotheses in causal learning
experiments and found that the assumptions of our model are correct.
- Soto, F. A., Gershman, S. J., & Niv, Y.
(2014). Explaining compound
generalization in associative and causal learning through rational
principles of dimensional generalization. Psychological Review, 121(3),
- Soto, F. A., Quintana, G. R., Ponce, F.
P., Perez, A. M., Vogel, E. H.
(2015). Why are some dimensions
integral? Testing two hypotheses through causal learning experiments. Cognition,