1. The inferior temporal (IT) cortex plays an important role in both short- and long-term memory for visual patterns. Most previous studies of IT neurons have tested their responses in recency memory tasks, which require that the memory lasts only the length of a single behavioral trial, which may be < 1 s. To determine the role of IT neurons in longer lasting memories, we measured their responses to initially novel stimuli as the stimuli gradually became familiar to the animal. 2. Two rhesus monkeys were trained on a delayed matching to sample (DMS) task with several intervening stimuli between the sample and the final matching stimulus on each trial. The purpose of the task was to ensure that the animal attended to the stimuli and held them in memory, at least temporarily. Unlike in several previous studies, the focus was not on within-trial effects but rather on the incidental memories that built up across trials as the stimuli became familiar. Each cell was tested with a set of 20 novel stimuli (digitized pictures of objects) that the monkey had not seen before. These stimuli were used in a fixed order over the course of an hour-long recording session, and the number of intervening trials between repetitions of a given sample stimulus was varied. 3. The responses of about one-third of the cells recorded in anterior-ventral IT cortex declined systematically as the novel stimuli became familiar. After six to eight repetitions, responses reached a plateau that was approximately 40% of the peak response. Virtually all of these cells also showed selectivity for particular visual stimuli and thus were not ''novelty detectors'' in the sense of cells that respond to any novel stimulus. Rather, the responses of these cells were a joint function of familiarity and specific object features such as shape and color. A few cells showed increasing responses with repetition over the recording session, but these changes were accompanied by changes in baseline firing rate, suggesting that they were caused by nonspecific effects. 4. The decrement in response with familiarity was stimulus specific and bridged > 150 presentations of other stimuli, the maximum tested. For some cells the maximum decrement in response occurred for those stimuli that initially elicited the largest response. There was no significant change in response to stimuli that were already familiar. 5. The same cells that showed familiarity effects also showed reduced responses to the matching stimuli at the end of each trial, compared with the responses to the samples. The responses to these matching stimuli declined with familiarity in parallel with the decline in responses to the samples over the session. Recency and familiarity effects appear to summate within IT cortex. 6. We examined the time course of responses to novel and familiar stimuli. The population of cells took 100 ms after response onset to distinguish between a novel stimulus and the same stimulus seen once previously, possibly reflecting feedback to IT cortex. However, after a single additional presentation, cells distinguished between novel and familiar stimuli within 10 ms of response onset, that is, by nearly the first action potential. Thus there is virtually no time for the effect of familiarity on the initial phase of the response to be caused either by lengthy temporal processing with IT cortex or by feedback from other structures. 7. The results support the proposal that a subpopulation of IT cells functions as ''adaptive mnemonic filters'' for both short- and long-term memories. A high level of activation in IT cortex may provide a feedback signal to orienting systems that the current stimulus is new and deserving of attention. Because increased contact with a stimulus drives down activity in IT cortex, this feedback between memory and attentional systems may result in an organism driven to seek out contact with new stimuli.
