In two experiments, the authors tested the effect of two types of reading on the spelling memory of strange or sound-spelling inconsistent words in Dutch students with and without learning disabilities: standard reading and regularized reading. Standard reading refers to reading the word the way it has to be read. Regularized reading refers to reading a sound-spelling inconsistent word as if it is sound-spelling consistent. In Experiment 1, both groups showed a short-term effect. Shortly after training, all students who participated in the regularized-reading condition showed better spelling performance than students who took part in a standard-reading condition. One week after training, spelling knowledge of students without learning disabilities appeared to be more stable than that of students with learning disabilities. In Experiment 2, only students with learning disabilities participated. The results reveal that more training substantially enhanced spelling performance in the long term, and repetition of the regularized word was even more effective. A recurrent network account served as the guiding principle to explain the results.