This study investigated short-term memory for odors using a four-alternative, forced-choiced recognition paradigm. Stimuli were the odors of 36 common food substances. Twelve subjects were tested in each offour conditions, which differed in the activity performed during the retention interval. Recognition performance was poorest when subjects free associated to an additional odorant presented during the retention interval. Thus, interference from interpolated events does occur in odor memory. Recognition performance was best when the subjects free associated to the name of the target odorant during the retention interval. Thus, the memory code for odors may incorporate semantic information. Remembering odors appears, therefore, to be governed by the same principles as remembering stimuli in other modalities.