Identifying the Neural Correlates of Mental Simulation in Multi-Step Planning

Overview

Planning is the ability to think ahead by considering possible future actions and their consequences. This research study aims to understand how the brain supports multi-step planning by testing whether people simulate promising future move sequences while deciding what to do next. Healthy adult volunteers will learn and play a strategy game called "Four-in-a-Row" (similar to Connect Four). Participants will complete two sessions on successive days: an online behavioral training/playing session and an in-person brain-recording session at New York University. During the brain-recording session, participants will view mid-game board positions and choose the best move while the study team records brain activity (using magnetoencephalography \[MEG\] or functional MRI \[fMRI\]) and eye movements. Data from the game and eye tracking will also be used to fit computational models of planning that help interpret the neural measurements.

This is a human neuroimaging study consisting of two related experiments designed to characterize the neural correlates of mental simulation during multi-step planning in the "Four-in-a-Row" game. Planning is modeled as a feature-based heuristic evaluation combined with look-ahead (tree search) that evaluates candidate actions by simulating future states and outcomes. Participants complete two sessions on successive days. Session 1 is a \~60-minute online behavioral session in which participants learn the rules of Four-in-a-Row (including a comprehension/quiz check) and play multiple games against computer opponents spanning difficulty levels. Behavioral data from Session 1 are used to fit a computational model of planning for each participant. Session 2 is an in-person neuroimaging session with simultaneous eye tracking. In the MEG experiment, participants complete a feature localizer followed by a primary planning task in which they evaluate mid-game board positions with a fixed decision window (e.g., 15 seconds) to encourage planning. B2 In the fMRI experiment, participants complete a planning task while BOLD activity and eye movements are recorded, using a trial structure designed to dissociate model-derived quantities such as myopic value and tree-search value. The main analyses will test where (fMRI) and when (MEG) the brain represents simulated future states, their values, and the evolving decision process, guided by participant-specific computational-model predictions.