Singh, Samara, & Wonnacott (pre-print). Statistical and explicit learning of graphotactic patterns with no phonological counterpart: Evidence from an artificial lexicon study with 7–8-year-olds and adults

Abstract

Children are powerful “statistical spellers”, showing sensitivity to untaught probabilistic orthographic patterns. Children and adults can also learn novel spelling patterns from artificial lexicons via statistical learning processes, akin to those established for spoken language acquisition, at least when the patterns have phonological counterparts. It is not clear whether learning written (graphotactic) patterns is possible when these are unconfounded from correlated phonotactics. We address this question by inducing learning of purely graphotactic patterns of varying complexity and also compare their learning under implicit versus explicit conditions. Across three experiments, we expose children and adults to CVC letter strings ending either with single or double letters (e.g., s, ss) that shared the same pronunciation. Critically, preceding vowel context predicted single/double letter usage. Results from two post-tests showed that children and adults incidentally generalized over two novel graphotactic constraints, regardless of complexity. Explicit instruction benefitted pattern generalization. Exploratory correlations suggested a relationship with measures of literacy only for explicit learning. This is a first demonstration that statistical learning processes suffice for purely graphotactic generalization. Explicit instruction is helpful, supporting practice of teaching spelling patterns early in literacy instruction.

Bayes factors for logistic (mixed effect) models

“…We now demonstrate how to apply the motivated-maximum approach to calculate a Bayes factor, using data and analyses drawn from Singh et al (2020). All analysis code used in this case study is available on the Language Learning Lab's GitHub repository: https://n400peanuts.github.io/languagelearninglab/usefulFunc tions-explained.html.…”