Ss (n = 100) generated verbal mediators for 25 low-, and 25 highassociability CVC pairs. They returned after periods of from one hour to four weeks to repeat the procedure. Two judges rated independently whether the mediators given by the Ss had changed. The results indicated that the high-associability pairs were less likely to change, but that there was substantial change for even the high-associability items after only one hour. The data were seen as consistent with Bower’s theory of stimulus encoding variability, and highlight the dynamic, immediate, and continuing nature of that process.