Simultaneous learning of semantic and graphotactic regularities in spelling: An artificial orthography learning experiment

Orthographies such as English represent various levels of regularity such that spelling patterns are conditioned to varying extents on phonological, graphotactic and semantic cues. Previous research has shown that people learn graphotactic regularities (i.e. spelling rules governing possible combinations of graphemes) implicitly through exposure, and that this knowledge generalises to new forms. Using an artificial orthography learning task, we investigated whether adults could simultaneously learn both graphotactic regularities and semantic regularities (where possible spellings depend on grammatical word class). Results across three experiments showed that adults learned spelling patterns conditioned on semantics, especially when semantic information was more salient with nouns and verbs rather than adjectives and adverbs. This learning was associated with the ability to verbalize the patterns at post-test. In contrast to earlier work, there was no evidence of graphotactic learning. This suggests that the two types of nonphonological regularities simultaneously may not be possible given the short exposure to the artificial lexicon in our paradigm. These findings are discussed in terms of a statistical learning account of orthographic learning.