From mazes to automation: Modernizing working memory research in animal models

Working memory (WM) is a core cognitive mechanism necessary for adaptive behavior. In the last few decades, scientists have studied WM using rodent models through traditional and time-consuming approaches, such as the Radial Arm Maze and the T-Maze. While these traditional tools have presented fundamental understanding, their dependence on manual operations restrains experimental precision and scalability. Here, we refine how emerging automated technologies-such as touchscreens, virtual reality (VR), and artificial intelligence (AI)-are inspiring this field by allowing high-throughput testing with improved precision. Further, we present a new framework to evaluate both classic and modern tasks based on their Scalability, Precision, and Neural Compatibility. This evaluation underlines how automation allows the emergence of modern paradigms, such as the Pulse-Based Accumulation Task and the Trial-Unique Nonmatching-to-Location (TUNL) task, offering more precise assessments of WM. Such technological progressions not only boost data quality and mitigate the efforts involved in data collection but also make way for a more unified understanding of the neural processes that underlie working memory.