Singh, F. D. , Samara, A. & Wonnacott, E. .
University College London (UCL)
Memorisation and explicit rule learning are insufficient for spelling development in languages such as English. Recent work suggests that developing spellers also exploit frequency-based properties of their orthography using the same statistical learning processes that operate in spoken language. Samara and Caravolas (2014) demonstrated incidental learning of novel graphotactic constraints on permissible letter combinations (e.g., to is permissible, te is not) that were embedded both in word beginnings and ends. We replicate this work and also ask: are both word contexts equally beneficial to learners? Over two brief sessions, we exposed 7-year-olds to word-like stimuli exemplifying novel constraints on letter distributions either in word beginnings or word ends (e.g., strings cannot begin with de vs. strings cannot end with el). At test, children were asked to make legality judgements (can this word exist in the language?) for novel strings that were/were not permissible. Overall discrimination was above chance, suggesting learning of the graphotactic constraints. There was no evidence of greater learning for word-final than word-initial constraints, in contrast to theories predicting higher learning in word-final positions (Treiman et al., 1995). Ongoing experiments explore whether graphotactic learning is underpinned by phonotactic learning by manipulating these factors in written and/or spoken stimuli.