A neural circuit mechanism for abstract free choice

Making abstract free choices, such as arbitrary decisions between comparable options formed independently of specific motor plans, is central to everyday behavior. Yet, the neural mechanisms underlying free choice remain poorly understood. Using a delayed-response paradigm (N=80), we dissociated abstract goal selection (color choice) from later motor response mapping. We found that freely chosen goals elicit slower responses and lower goal-switching costs than instructed goals. We developed a hierarchical attractor network model in which symmetry-breaking attractor dynamics drive free goal selection, and conjunctive representations translate these abstract goals into motor output. Freely chosen goals converged on less extreme fixed points than instructed ones, accounting for their slower, more variable responses. Consistent with our model, human EEG activity (N=30) sequentially encoded goal, conjunctive, and response representations, with conjunctive representations specifically coupled to behavioral variability. Our findings offer plausible circuit-level mechanisms for how abstract free choices are made and translated into action.