The goal of this study was to test the item-based assumption of the self-teaching hypothesis by reanalyzing Cunningham's (2006) data. In Cunningham's study, 37 first graders participated in a self-teaching experiment, and their orthographic learning was measured by a spelling task and an orthographic choice task. A 2-level logistic regression model was used to analyze accuracy of phonological decoding of a target word and spelling acquisition of that word. The findings suggest that the probability of spelling a target word correctly increases as the decoding accuracy of the target word increases. In contrast, this relation was not found for the orthographic choice task. The discrepancy in the results between the 2 dependent measures might be due to the difference in cognitive processes required to complete a production task as opposed to a recognition task. The current findings provide item-based evidence for the accuracy of phonological decoding on participants' performance on a spelling task.