Letter coding in visual word recognition: The impact of embedded words

•Lexical decisions to nonwords are delayed by the presence of an embedded word.•Such interference is greater for initial than final embedding.•An initially embedded word only interferes when it has a consonantal coda.•The results oppose letter-coding models that ignore consonant/vowel structure.

Nonword classification responses are examined in this study to establish the amount of interference arising from the presence of an embedded word. In Experiment 1, greater interference is found from an initial embedding (e.g., furb vs lurb, cf. fur) than a final embedding (e.g., clid vs clig, cf. lid). In addition, an “outer” embedding (e.g., jomb vs vomb, cf. job) generates interference that is no greater than for an initial embedding. These results are inconsistent with the idea of left-to-right parsing, while accounts of word recognition that center on open bigrams or the spatial coding of letters require additional processes. Instead, the results are interpreted within a model of word parsing and lexical access that incorporates subsyllabic structures; an account that is supported in Experiments 2 and 3 by the critical finding that an initially embedded word interferes more when it ends in a consonant (e.g., the fur of furb, the shadow of shadowl) than with a vowel (e.g., the tea of teaf, the coffee of coffeep).