Mechanisms underlying learning-related representational reorganization

点击翻译按钮获取中文摘要

As humans, we boast a seemingly unbounded capacity for new learning, but have only a finite amount of neural tissue to store information. Our brains accomplish this by partially reusing neurons to represent new information. This adaptive organization is critical and may be fundamental to developing expertise, remembering without interference, and generalizing across experiences. In the proposed work, we will combine behavioural experiments, neuroimaging, and computational methods to characterize how memories are adaptively reorganized as we learn and act in the world. The long-term aim of my research is to understand the basic cognitive mechanisms underlying adaptive learning and memory systems. Furthermore, I aim to quantify how learning-related memory reorganization impacts later performance, developing approaches for preventing interference and promoting learning. The three aims presented below contribute to addressing these broader goals. AIM1 When we experience our visual world, what we encode into memory is not a snapshot of our experience, but rather the result of a trajectory of hierarchical visual processing, wherein low-level orientations and colors are built into object identities, which are fed forward into our learning systems. To understand how representations reorganize themselves in memory, we must understand this path to encoding. In AIM1, we leverage recent advances in computer vision to map from model features to the human visual cortex, which will allow us to understand representational organization at various levels of visual complexity. AIM2 In recent pilot research, we demonstrated that learned representations differentiate from one another, or become more distinct, when there is moderate overlap, or competition between them. In contrast, when there is high overlap, learned representations integrate with one another. In AIM2 we will explore the alluring possibility that such competition-related differentiation is a domain-general learning mechanism, that occurs across tasks and stimulus types. AIM3 In recent work, I demonstrated that drawing can foster distinctive memories. Drawing, through creating a context-rich memory trace, might render memories more distinctive. In AIM3, we will test the hypothesis that competition between overlapping drawings drives differentiation, that primarily the most distinctive drawings are remembered, and that distinctive patterns of visuomotor connectivity leads to improved memory. Together, the impact of these research aims will propel the field forward by illuminating how memory representations are encoded and reorganized. The results will also yield potential applied approaches for improving education. Through this work, we will also develop and share cutting-edge tools capable of exerting control over the representational spaces instantiated in the brain. The findings and their applied potential will be of interest no only to researchers, but also to educators, and the general public.