Why We Dive Deep: A Momentum-Based Account of Depth-First Persistence in Hierarchically Organized Goals

Humans can flexibly organize their behavior to pursue hierarchically structured goals, dynamically updating their commitment to goals and subgoals in response to changing environmental contingencies. Using a novel hierarchical goal-selection paradigm, we investigated patterns of persistence during goal and subgoal pursuit under varying subgoal prospects. Participants exhibited a depth-first pattern of commitment, tending to persist with both goals and subgoals rather than maintaining commitment to goals alone without a corresponding subgoal commitment. Greater depth-first commitment was associated with improved task performance and larger switching costs when changing goals. We further found that depth-first persistence varied systematically with progress at both the goal and subgoal levels. To characterize these behavioral patterns, we evaluated a hierarchical momentum model alongside alternative computational accounts. Our proposed model captured key features of participants' goal and subgoal commitment patterns and generally provided a better fit than the alternative models considered. Moreover, model-derived momentum signals predicted persistence patterns above and beyond measures related to dynamic switching costs. Taken together, these findings are consistent with the hypothesis that hierarchical momentum contributes to persistence in complex, long-horizon goal pursuit and provide a computational framework for understanding behavior in hierarchically structured goals.