Most optical character recognition systems are not error-free. Thus, some characters are rejected or substituted when an off-line document is processed. Alternatives of a rejected character can be provided to help a human user to select the correct one while proofreading the processed document. This study investigated several ways of presenting alternatives for rejected characters and the user strategy of proofreading documents with rejected and substituted characters. Design factors examined included rejection and substitution rates, number of alternatives, and the availability of the original document for proofreading. Proofrading speed and accuracy were measured. The results were analyzed statistically and analytically to come up with human factors design guidelines. Detailed design strategies and recommendations were suggested based upon the results. Relevance to industry. Since most optical character recognition systems are not perfect, the documents processed by the systems often require proofreading by a human user to correct substituted and rejected characters. This study provides guidelines for designing an off-line character recognition system to optimize the human proofreading performance such as proofreading speed and accuracy.