Purchase this article with an account.
Shaul Hochstein, Volodya Yakovlev, Sandro Romani, Daniel Amit; Memory mechanisms for familiarity recognition and identification. Journal of Vision 2008;8(6):693. doi: https://doi.org/10.1167/8.6.693.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Encountering someone on the street, we can often say whether their face is familiar, though it may be difficult to identify the person: Familiarity memory may differ essentially from identification, which embeds and generates contextual information. We studied multiple stimulus memory in macaque monkeys using a modified delayed-match-to-sample task. In each trial, a sequence of 1–6 different images served as samples, followed by a match stimulus - a repetition of any of the samples (the cue). Sequence length and cue position were randomized. Repetition recognition entailed holding all sequence images in memory. For Experiment 1, we used 16 images. Miss errors, (failures to detect repetitions), increased with cue-match separation. False positives (FP; erroneous reports of a repetition), increased with trial duration. Overall, Hits reached 83% with 3.2% FP. For Experiment 2, we used 12,000 novel images; all stimuli were “Novel”, except the final match and a number of images introduced as “catch” trials. Unexpectedly, performance was better than with well-trained stimuli, reaching 91% Hits, 2.6% FP. We suggest that monkeys use delay-activity working-memory for well-trained stimuli and a new strategy, based on modulated responses to repeated images, for Novel stimuli. The low FP rate suggests a between-trial memory reset mechanism. These results imply extremely effective one-shot learning of familiarity, resembling Standing's (1973) finding that people detect familiarity for 10,000 once-seen pictures (with 80% accuracy). How is this immense memory established and maintained? We present a generic, simple neural network model, with coupled neural and synaptic dynamics, quantitatively simulating these behaviors, based on the same conservative Hebbian synaptic plasticity that generates delay-activity identification memory. Familiarity becomes the first step toward establishing identification. An inter-trial reset limits FP for previous-trial images. The model relates repetition and familiarity recognition with enhanced neural activity, as experimentally observed in most differentially responding prefrontal cortex neurons.
This PDF is available to Subscribers Only