Zettersten, M. 1 , Potter, C. 2 & Saffran, J. 1
1 University of Wisconsin-Madison
2 Princeton University
Non-adjacent dependencies are ubiquitous in language, but difficult to learn. Previous research has shown that the presence of high variability between dependent items facilitates learning. Yet what allows learning of non-adjacent dependencies even without high variability in intervening elements? One possibility is that learning non-adjacent dependencies highlights similar structures, allowing people to learn new non-adjacent dependencies that are otherwise difficult. Two studies (N=138) tested whether exposing participants to non-adjacent dependencies presented with high variability subsequently helps learners discover novel non-adjacent dependencies they would otherwise fail to detect. In the Experimental Condition, we first exposed participants to three non-adjacent dependencies with high variability in the intervening elements (items with the structure aXb, cXd and eXf, with 24 different X elements). The Control Condition contained identical elements, but inconsistent non-adjacent information. All participants then encountered three new non-adjacent dependencies with low variability in X elements (gXh, iXj and kXl, with three different X elements) that are difficult to learn. Participants in the Experimental Condition showed better learning of the novel non-adjacent dependencies (M=.60, 95% CI=[0.56, .65]) than Control participants (M=.53, 95% CI=[0.51, 0.56]), t(136)=2.52, p=.01, d=.42), suggesting that exposure to learnable non-adjacent dependency structure helped participants learn challenging new non-adjacent dependencies.