2024年7月5日 Friday Lunch Seminar (英語で開催します)
12:15 〜 13:00
CiNet棟大会議室にて開催
演題:Context-dependent changes in neural representation of working memory
大阪大学
大学院 生命機能研究科
助教 内村 元昭
担当PI : 北澤 茂
Abstract:
Working memory is central to cognition, allowing one to hold items ‘in mind’. However, despite its importance, working memory is faulty – stimuli are imperfectly encoded in memory and memories can degrade over time. Previous psychophysical studies suggest the accuracy of memories depends on the context. Memories are inaccurate when there is a large range of stimuli that could be remembered on any given trial. However, when the range of stimuli is smaller, memories are more accurate. This suggests working memory representations are adaptable, changing to efficiently store information in working memory for the current context.
To understand the neural mechanisms that support this adaptability, we trained two monkeys to perform a working memory task that required them to remember two colored squares and, then, after a memory delay, report the color of the ‘selected’ memory. Initially, the color of each stimulus was randomly selected from the full color wheel (i.e., the entire rainbow). However, after a few hundred trials, the stimuli colors were drawn from one half color space (arbitrarily chosen). Consistent with previous work, the monkeys were more accurate when colors were drawn from a smaller range of possible values (the half color space). In addition, we found the monkey’s response was attracted towards the center of restricted range, reducing the accrual of errors over time.
To understand the neural basis of these improvements, we simultaneously recorded from prefrontal and parietal cortex as the monkeys performed the task. The distance in neural space of the representation of nearby colors was greater when the color range was restricted. Furthermore, the geometry of the color representation changed from semi circular to triangular, allowing nearby colors to be more finely discriminated and, thus, improving working memory accuracy. Our results provide insight into how neural representations adapt to the context in order to improve working memory accuracy.