Testing Statistical Learning Implicitly: A Novel Chunk-based Measure of Statistical Learning

Isbilen, E. S. ¹ , McCauley, S. M. ² , Kidd, E. ³ & Christiansen, M. H. ¹

1 Cornell University
2 University of Liverpool
3 The Australian National University

Connecting individual differences in statistical learning with broader aspects of cognition has received considerable attention, yet has yielded mixed results. This may in part arise from how statistical learning is typically tested, using the two-alternative forced choice (2AFC) task. As a meta-cognitive task that relies on explicit familiarity judgments, 2AFC may not accurately capture implicitly formed statistical computations, nor does it elucidate how these computations result in word-level representations. To address these issues, we adapt the classic serial recall memory paradigm to implicitly test statistical learning in the statistically-induced chunking recall (SICR) task. We hypothesized that artificial language exposure would lead participants to chunk recurring statistical patterns into words, which should facilitate recall of the words from the input. In Experiment 1, we compared the efficacy of SICR to 2AFC in capturing statistical learning behavior, and show that SICR offers more fine-grained insight into individual differences in learning. In Experiment 2, we establish the test-retest reliability of SICR, which demonstrates higher reliability than 2AFC. Furthermore, preliminary data suggests that SICR predicts individual differences in language comprehension, whereas 2AFC does not. Thus, SICR offers a more sensitive measure of individual differences, and suggests that basic chunking abilities may explain statistical learning.