Measuring Statistical Learning and its relationship to reading ability without Measuring Statistical Learning: A Structural Equation Modelling Approach

Statistical learning is proposed to affect important cognitive skills, including reading ability. Empirically, however, the link between statistical learning ability and reading is contentious. One issue is that it is difficult to design a psychometrically sound task measuring statistical learning ability, especially for children. Here, we circumvent this issue by assessing the correlation between reading-related tasks that are proposed to rely on a statistical learning mechanism. Both morphological awareness (“this is one wug – these are two …?”) and sensitivity to graphotactic regularities (“fubb looks like a word, ffub does not”) rely on the extraction of print-related regularities. As such, they should correlate with each other, and with reading ability. Using data from an online study with 2,624 German primary school children, we built a Structural Equation Model including sensitivity to graphotactic regularities and morphological awareness as predictors of reading ability, alongside other established predictors. We found no evidence for an independent effect of morphological awareness and graphotactic sensitivity on reading: rather, these affect reading indirectly via whole-word knowledge or reflect co-variation with age. However, morphological awareness and graphotactic sensitivity correlated with each other. This suggests that a common learning mechanism, such as statistical learning, may underlie the learning of different types of print-related regularities.