Conducted 3 experiments with 179 undergraduates and faculty members. In Exp I, Ss indicated which of 21 pairs of odors had been in the original group of 48 odors at presentation-test intervals of up to 1 mo. Exp II (n = 68) Ss were presented with 20 odors and asked to write labels for each; recognition was tested after 3 mo. In Exp III (n = 74), Ss assigned labels to each of 20 odors and were tested for recognition 3 mo later. Immediate recognition tests produced numerous errors, but there was little further retention loss for periods up to 3 mo. Neither verbal labeling nor odor familiarity aided memory, while long-term retention held up even when there were no instructions to memorize. Use of similar odors as test distractors impaired recognition significantly, but it remained well above chance after 3 mo. It is suggested that odors are coded as unitary perceptual events with little attribute redundancy; this leads to poor immediate retention but great subsequent resistance to distortion of immediately retained odors. (PsycINFO Database Record (c) 2012 APA, all rights reserved)