Forty-eight adults were trained on monosyllabic pseudowords and their meanings and then tested in vocal spelling. The orthographic inconsistency of the rime (e.g. "orn, awn" for "glorn") and the number of learning trials affected accuracy and response latency in the vocal spelling test. In addition, orthographic typicality as assessed by neighbour statistics predicted item spelling accuracy. Spelling accuracy on orthographically consistent items significantly increased with training, suggesting that unfamiliar monosyllables are not necessarily spelled by reliance on sound-spelling correspondences at the rime level. Analysis of spelling errors revealed that good spellers made more spelling errors containing alternative rime spellings and fewer errors that were phonologically inappropriate than poor spellers.