Hi! Here is the code, stimuli & presentation scripts, and results from our forthcoming paper on SL. Feel free to look around, and send me an email if you have any questions. The data we will make public as soon as we get confirmation from our local ethics board.

Online and offline measures of statistical learning reflect sensitivity to pairwise relationships, but not chunking

Abstract

Statistical learning allows individuals to rapidly extract regularities in the sensory environment. We replicated previous findings showing participants become sensitive to the implicit structure in a continuous speech stream of repeating tri-syllabic pseudowords, as measured by an online target detection task and offline word recognition task. Consistent with previous findings, we found only a weak correlation between these two measures of learning, leading us to question whether there is overlap between the information captured by these two tasks. Representational similarity analysis on reaction time data from the target detection task revealed that reaction time data reflects sensitivity to transitional probability, ordinal position, and pairwise groupings, but not triplet (pseudoword) groupings. Furthermore, individual performance on the word recognition task was only significantly predicted by their sensitivity to transitional probability. We conclude that these canonical SL tasks do not reliably measure chunking of embedded units (e.g. tri-syllabic pseudowords), but more simply a sensitivity to transitional probability and pairwise relationships between stimulus items (e.g. syllables).

Sneak Peak Figs.

Fig. 1. Implicit learning of a pseudo-language leads to rapid reactiom time facilitation for target syllables in predictable locations.

Note that the Experiments 1 and 2 in the manuscript refer to "exp 3" and "exp 4", respectively, in the code.