Homonyms with one pronunciation (e.g., sal) but two spellings and meanings (sail, sale) were presented aurally and then written under a disguised pretext. The spelling was assumed to reflect the dominant encoding. Initial encodings were either prompted by semantically selected context words or were not prompted. After intervals of 1-8 min, encodings were observed again, either with prompting context or without. Initially prompted encodings were less stable than unprompted ones, and the instability increased with increasing intervals. A prompting context at the second encoding opportunity increased the amount of recoding.This paper is based on a thesis submitted to the University of Iowa in partial fulfillment of the requirements for the PhD. Appreciation is due especially to Rudolph W. Schulz and also to Sue R. Rosner, James V. Hinrichs, and an anonymous reviewer for advice and encouragement, and to Ronald Kellog and Alice Sanders for assistance.