Statistical and explicit learning of graphotactic patterns with no phonological counterpart: Evidence from an artificial lexicon study with 7-year-olds and adults

Statistical and explicit learning of graphotactic patterns with no phonological counterpart: Evidence from an artificial lexicon study with 7-year-olds and adults

First Author: Anna Samara -- University of Greenwich
Additional authors/chairs: 
Daniela Singh; Elizabeth Wonnacott
Keywords: Statistical learning, Graphotactic sensitivity, Spelling, literacy acquisition, Orthographic Knowledge
Abstract / Summary: 

Purpose. Recent work suggests that children and adults are sensitive to untaught spelling patterns of their orthography and can learn similar artificial patterns from brief incidental exposure. Here, we investigate learning of graphotactic constraints which cannot be underpinned by phonotactic learning: We ask whether visual statistical learning processes suffice to support children’s and adults’ learning of two patterns of varying complexity; and compare children’s ability to learn “difficult” patterns under comparable incidental and explicit conditions.

Method. In three 2-day learning experiments, we exposed 7-year-olds (n = 85) and adults (n = 64) to CVC(C) strings ending either with single or double codas (e.g., s, ss) that share the same pronunciation. In Experiment 1, single consonants were always preceded by one of two word-medial vowels (e.g., duf, rut) and doublets were always preceded by the other (e.g., deff, rett). In Experiments 2/3, consonants doubled following one of two vowels (e.g., deff, rell) and occurred as singlets following the other (e.g., duf, rul). At test, participants (a) made legality judgements for novel strings that were/were not permissible and (b) performed a “fill-in-the-blank” task.

Results. Children and adults incidentally learned and generalized over the novel graphotactic constraints regardless of complexity. Explicit instruction benefitted learning performance.

Conclusions. We provide the first strong demonstration that visual statistical learning processes support the learning of purely graphotactic spelling patterns. The advantage for learning under explicit conditions suggests that explicit teaching of spelling patterns, even when these are hard to articulate, is helpful for typically developing spellers.