Share (1995) proposed "phonological recoding" (the translation of letters into sounds) as a self-teaching mechanism through which readers establish complete lexical representations. More recently, McKague et al. (2008) proposed a similar role for "orthographic recoding", that is, feedback from sounds to letters, in building and refining lexical representations. We reasoned that an interaction between feedback consistency measures and spelling ability in a spelling decision experiment would lend support to this hypothesis. In a linear mixed effects logistic regression of accuracy data, this interaction was significant. Better spellers but not poorer spellers were immune to feedback effects in deciding if a word is spelled correctly, which is consistent with McKague et al.'s prediction that the impact of phonological feedback on word recognition will diminish when the orthographic representation for an item is fully specified. The study demonstrates the importance of considering individual differences when investigating the role of phonology in reading.