Two experiments examined false recall for lists of semantically and phonologically associated words as a function of presentation duration. Veridical recall increased with long exposure durations for all lists. For semantically associated lists, false recall increased from 20–250 ms, then decreased. There was a high level of false recall with 20 ms durations for phonologically associated lists , which declined as duration increased. In Experiment 2, for lists presented at 20 and 50 ms rates, false recall given zero correct recall (...) was observed frequently, suggesting that conscious recollection of studied words was not necessary for phonological false memory. Differences between phonologically and semantically associated lists were consistent with a feature integration model based on automatic initial processing of phonetic features of words. (shrink)