Examined the effect of varying the retention interval in a recognition task on the features used by 192 undergraduates to remember the stimuli. 4 groups of Ss were tested using retention intervals of 6, 20, and 80 sec., or 20 min. The recognition test contained words that correctly matched original list (OL) words, and 3 types of nonmatching words: neutral, acoustically similar to OL words, and semantically similar to OL words. The latency of correct responses increased over retention intervals, and was significantly longer for the acoustically and semantically similar words than for the neutral words, with no Items * Intervals interaction. Confidence ratings and error data show constant confusion effects, suggesting that memory stages should not be defined by mode of stimulus representation. (PsycINFO Database Record (c) 2006 APA, all rights reserved)