MEMORAT · Continual learning and memory for sounds in rat

How much can we remember and for how long? Decades of key advances in our understanding of synaptic plasticity resulted in a puzzle. On the one hand, memories can be highly stable and organisms are capable of adding new memories without destroying old ones. On the other hand, data and models support neither stability nor lifelong (‘continual’) learning. Many experiments suggest that responses of neuronal populations slowly but inexorably drift (“representational drift”), while synapses spontaneously turnover. At the same time, models predict that memory traces of new experiences tend to interfere destructively with old traces. Here, I will study auditory continual learning in the behaving rat. I will develop a novel behavioral paradigm that will train rats to recognize two new sounds every few days, memorizing up to 100 sounds over 3-4 months of continuous training, and check the stability of memory during that time. My lab will apply state-of-the-art recording techniques in order to document the neural activity during memory acquisition and between training sessions while the rats sleep and consolidate newly acquired memories. I will repeat these studies in a rat model for Alzheimer’s disease, which expresses early memory deficits. These results will be compared with predictions of models for memory in the auditory cortex. At one end, I will test the predictions of a model, based on recently published results, that posits the existence of a few neurons that carry the memory of each learned sound by intrinsic bursting. At the opposite end, I will test predictions of models in which memory resides in the coordinated activity of large, distributed ensembles of neurons. Overall, this project will address the deep puzzles of memory when memory load is high, a regime that has not been studied in animal models.