A NOVEL CONTINUOUS-REPORT SPATIAL WORKING MEMORY TASK REVEALS STRATEGY-DEPENDENT CORTICAL DYNAMICS

Understanding the neural mechanisms that support working memory (WM) remains a central challenge in neuroscience. Multiple competing models have been proposed to explain WM-related neural dynamics, yet discriminating between these models has proven difficult, in part due to existing behavioral paradigms being limited. Many rely on discrete stimuli, binary choices and accuracy-based measures, which obscure the format, precision and evolution of internal memory representations. Here, we introduce a novel spatial working memory task in rats designed to generate trial-by-trial behavioral signatures that can distinguish between competing models of WM. The task is high-throughput and touchscreen-based, enabling efficient training and testing of large cohorts of animals, while capturing individual differences in strategy and biases, revealing the internal models used in solving the task. Unlike previous WM paradigms that rely on discrete or binary stimuli and / or choices, our task uses parametric spatial stimuli combined with continuous report on each trial. This design yields rich behavioral error distributions, allowing quantitative, model-specific predictions about the format, precision, and evolution of memory representations over time. Behavioral modelling reveals that rats adopt distinct internal strategies to solve the task. We then simultaneously recorded neural population activity from medial prefrontal cortex (mPFC) in rats performing the task. These recordings reveal a rich diversity of structured neural dynamics that covary with behavioural and inferred internal representations. Together, this new behavioral and neurophysiological framework provides a powerful platform for probing cortical population dynamics underlying spatial working memory and for empirically testing competing theoretical models of WM.