Storing a new pattern in a palimpsest memory system comes at the cost of interfering with the memory traces of previously stored items. Knowing the age of a pattern thus becomes critical for recalling it faithfully. This implies that there should be a tight coupling between estimates of age, as a form of familiarity, and the neural dynamics of recollection, something which current theories omit. Using a normative model of autoassociative memory, we show that a dual memory system, consisting of two interacting modules for familiarity and recollection, has best performance for both recollection and recognition. This finding provides a new window onto actively contentious psychological and neural aspects of recognition memory.

Information transfer between brain structures located in the medial-temporal lobe and the neocortex is essential for learning. However, the neuronal underpinnings of this transfer are unknown. Doron et al. found that neurons located in the deep layers of the perirhinal cortex exhibit increased firing after microstimulation upon learning (see the Perspective by Donato). Learning was associated with the emergence of a small population of neurons in layer 5 of the somatosensory cortex that increased bursting upon stimulation. This increase in bursting was accompanied by an increase in dendritic activity, and silencing the perirhinal cortex to layer 1 projection effectively disrupted learning and its physiological correlates. During learning, perirhinal inputs thus act as a gate for the enhancement of cortico-cortical inputs, which are necessary for stimulus detection and are strengthened during learning. Science , this issue p. [eaaz3136][1]; see also p. [1410][2] ### INTRODUCTION Arguably one of the biggest mysteries in neuroscience is how the brain stores long-term memories. Since the 1950s, it has been well established that long-term memories reside in the neocortex but that their formation is dependent on the hippocampus and medial-temporal lobe structures. It is therefore remarkable that we still do not know what cellular mechanisms underlie long-term memory storage in the neocortex or exactly where they are located. ### RATIONALE The primary challenge for investigating the neural circuit underlying memory formation in the neocortex is the distributed nature of the resulting memory trace throughout the cortex. We therefore used a behavioral paradigm dependent on both the hippocampus and neocortex that enabled us to generate memory traces in a specific cortical location by training rodents to associate the direct electrical microstimulation of the cortex with a reward. This also allowed us to specifically examine the contribution of circuit elements in the defined anatomical location. Rodents learned to behaviorally report the microstimulation within only a few trials and improved over a 3-day period, during which we examined the evolution of learned neuronal responses in the stimulated area. We hypothesized that the influence of the hippocampus on memory formation in the neocortex occurs at the interface between these two structures. ### RESULTS We first confirmed that learning to associate microstimulation of the primary somatosensory cortex (S1) with a reward depends on hippocampal activity using suppression of action potentials (APs) with lidocaine in this brain area. Using retrograde and anterograde tracing methods, we found that the perirhinal cortex (the last station in the medial-temporal loop projecting to S1) predominantly targets layer 1 (L1), suggesting that important events relating to memory formation occur in neocortical L1. We tested this with targeted chemogenetic suppression of perirhinal input to L1 above the stimulated S1 region. Notably, this very specific and localized manipulation was sufficient to disrupt learning. The effect was learning-specific and had no influence in expert animals, which demonstrated that the perirhinal input did not alter the ability to perceive and behaviorally report the stimulus, per se. We found that the perirhinal cortex signaled information related to successful behavior during learning, gated the evolution of distinct firing, and enhanced burst responses in 11% of layer 5 (L5) pyramidal neurons in S1 (40% of neurons had reduced firing responses and 49% showed no change in firing). Apical dendritic excitability was correspondingly enhanced in a similar proportion of L5 pyramidal neurons. This suggested that the mechanism for memory formation had a dendritic origin. Consistent with this hypothesis, we found that both activation of γ-aminobutyric acid type B (GABAB) receptors—which disrupt apical dendritic calcium (Ca2+) activity—and activation of dendritic-targeting, somatostatin-positive interneurons impaired memory formation similarly to suppressing perirhinal input to L1. Finally, we found that after learning the microstimulation detection task, evoking a single burst of APs (but not the same number of low-frequency spikes) in a single L5 pyramidal neuron could trigger behavior. ### CONCLUSION We found that medial-temporal input to neocortical L1 gates the evolution of specific firing responses in subpopulations of L5 pyramidal neurons including up- and down-regulated firing patterns and an elevation in burstiness by means of a mechanism that is most likely related to apical dendritic activity. After learning, these neocortical responses become independent of the medial-temporal influence but continue to evoke behavior with bursts conveying higher saliency. We conclude that L1 is the locus for hippocampal-dependent associative learning in the neocortex, where memory engrams are established in subsets of pyramidal neurons by enhancing the sensitivity of tuft dendrites to contextual inputs and driving burst firing. ![Figure][3] Probing the influence of the medial-temporal lobe on memory formation. ( A ) The perirhinal cortex (PRh), part of the medial-temporal lobe (MTL) structures (blue box), targets L1 in S1. ( B ) We performed microstimulation (μStim) of L5 in the sensory cortex while chemogenetically suppressing the axonal projection to cortical L1 (red) from PRh (blue arrow). ( C ) Rodents learned to associate μStim with water reward. ( D to F ) Learning was suppressed by blocking MTL input [(D) and (E)] that otherwise evoked dendritic activity in L5 pyramidal neurons [(F), orange tuft], which allowed them to associate contextual input to L1 with the μStim. ( G ) More than 3 days of learning led to subpopulations of firing responses in L5 cells in S1 such that evoking a burst but not low-frequency spikes in single L5 cells retrieved learned behavior. Hippocampal output influences memory formation in the neocortex, but this process is poorly understood because the precise anatomical location and the underlying cellular mechanisms remain elusive. Here, we show that perirhinal input, predominantly to sensory cortical layer 1 (L1), controls hippocampal-dependent associative learning in rodents. This process was marked by the emergence of distinct firing responses in defined subpopulations of layer 5 (L5) pyramidal neurons whose tuft dendrites receive perirhinal inputs in L1. Learning correlated with burst firing and the enhancement of dendritic excitability, and it was suppressed by disruption of dendritic activity. Furthermore, bursts, but not regular spike trains, were sufficient to retrieve learned behavior. We conclude that hippocampal information arriving at L5 tuft dendrites in neocortical L1 mediates memory formation in the neocortex. [1]: /lookup/doi/10.1126/science.aaz3136 [2]: /lookup/doi/10.1126/science.abf4523 [3]: pending:yes

The problem of catastrophic forgetting can be traced back to the 1980s, but it has not been completely solved. Since human brains are good at continual lifelong learning, brain-inspired methods may provide solutions to this problem. The end result of learning different objects in different categories is the formation of concepts in the brain. Experiments showed that concepts are likely encoded by concept cells in the medial temporal lobe (MTL) of the human brain. Furthermore, concept cells encode concepts sparsely and are responsive to multi-modal stimuli. However, it is unknown how concepts are formed in the MTL. Here we assume that the integration of audio and visual perceptual information in the MTL during learning is a crucial step to form concepts and make continual learning possible, and we propose a biological plausible audio-visual integration model (AVIM), which is a spiking neural network with multi-compartmental neuron model and a calcium based synaptic tagging and capture plasticity model, as a possible mechanism of concept formation. We then build such a model and run on different datasets to test its ability of continual learning. Our simulation results show that the AVIM not only achieves state-of-the-art performance compared with other advanced methods but also the output of AVIM for each concept has stable representations during the continual learning process. These results support our assumption that concept formation is essential for continuous lifelong learning, and suggest the AVIM we propose here is a possible mechanism of concept formation, and hence is a brain-like solution to the problem of catastrophic forgetting.

Seen it, seen it, seen it, seen it, seen it. Most of us instinctively know whether objects are familiar or unfamiliar. Now we may know how we know. It turns out monkeys have a cluster of neurons in their brains that decides whether or not they have seen objects before. The primary visual area, at the back of the brain, does most of the early work in perceiving an object, especially its physical attributes, such as what direction it is moving.

Storing a new pattern in a palimpsest memory system comes at the cost of interfering with the memory traces of previously stored items. Knowing the age of a pattern thus becomes critical for recalling it faithfully. This implies that there should be a tight coupling between estimates of age, as a form of familiarity, and the neural dynamics of recollection, something which current theories omit. Using a normative model of autoassociative memory, we show that a dual memory system, consisting of two interacting modules for familiarity and recollection, has best performance for both recollection and recognition. This finding provides a new window onto actively contentious psychological and neural aspects of recognition memory.