Abstract
We are studying familiarity memory (knowing that I have seen this image somewhere. sometime before) and identification memory (knowing where and when I saw it, and perhaps its name). We now address memory decay in these conditions, including cases when forgetting is desirable. We introduce a multiple-item memory task, where subjects report repetition of any stimulus in a sequence. Two primate groups participated: Group TB (n=2) trained with a fixed set of 16 images, with many between-trial image repetitions. Group DL (n=2) were trained with an unlimited number of novel images, except for some "catch" trials with previously-seen images. Performance reflects two types of error: Misses, not reporting a repetition, and False Positives (FP), erroneously responding to an image which is not a within-trial repetition. FPs often occur for images which were present in a previous trial - as occurs frequently with the fixed set. Still, group TB had few FPs, and they drop off quickly with number of trials since the original presentation (mean: 1-trial-back: 9%; 2-trials-back: 3%). We suggest presence of an inter-trial reset mechanism which "forgets" (most) seen images. With unlimited novel images, however, group DL had many FPs (1-trial-back: 80%; 2-trials-back: 66%), suggesting they used familiarity memory. Group DL did not experience inter-trial repetitions during training, so perhaps they didn't develop the inter-trial reset. When the tasks were switched, group TB utilized its previously-learned reset also for novel images, and group DL now established reset. Furthermore, group DL, originally without reset, afforded a unique opportunity to estimate memory decay without reset. Memory gradually decreases, with ~20% forgotten each trial (~5s), so that ~10% of images remain in memory and produce FP errors following 10 trials. We conclude that primates easily remember dozens of images, with a fixed gradual loss of image recognition over many minutes.
Meeting abstract presented at VSS 2012