University students made spelling accuracy judgments about correctly and incorrectly spelled words that had been seen incorrectly spelled (Experiments 1 and 2). In contrast to results for spelling production, studying a misspelling produced a small benefit in classification of the correct word at test. When the studied misspelling was re-presented at test, there was a substantial cost in accuracy. Testing spelling recognition in an old context had a biassing effect, but there was little evidence of context re-instatement effects for studied words. In Experiment 3 students decided whether a correctly spelled word was spelled the same way at study and test. Participants' poor performance with words studied misspelled supports a priming explanation of the benefit for correct words. The differential effects for correct and incorrect test words cannot be explained in terms of updating abstract lexical representations, and the limitations on participants' item and context memory challenge episodic accounts of lexical representations.