Abstract
Decision-making is typically studied as a sequential process from the selection of what to attend to which actions to take based on the attended information. However, people often process information across these various levels in parallel. Here we scan participants while they simultaneously weigh how much to attend to two dynamic stimulus attributes and what response to give. Regions of the prefrontal cortex track information about the stimulus attributes in dissociable ways, related to either the predicted reward (ventromedial prefrontal cortex) or the degree to which that attribute is being attended (dorsal anterior cingulate cortex, dACC). Within the dACC, adjacent regions track correlates of uncertainty at different levels of the decision, regarding what to attend versus how to respond. These findings bridge research on perceptual and value-based decision-making, demonstrating that people dynamically integrate information in parallel across different levels of decision-making.
Introduction
In classical laboratory settings, decision-making is often treated as a linear, sequential process: first, an individual is directed to a specific target of attention, and then they decide on a motor response. However, naturalistic decisions are rarely this tidy. For instance, while writing a report, an incoming email notification forces a parallel evaluation: how much attention should be diverted to the email versus the current task, and what specific actions (keystrokes) should follow?
This study addresses a significant gap in cognitive neuroscience: how the brain integrates these hierarchical levels of decision-making in parallel. Specifically, the researchers investigated whether distinct or overlapping neural circuits track the evidence used to select a target of attention versus the evidence used to select a physical action.
Experimental Design and Methods
The researchers utilized fMRI to scan 34 healthy participants performing a "bivalent" perceptual task. Participants viewed "random dot kinematograms"—clouds of moving, colored dots—and were required to judge both the dominant color (red vs. blue) and the dominant motion (up vs. down).
Crucially, participants provided only a single button press that could be correct for one, both, or neither attribute. To manipulate attentional priority, the rewards for correct color versus motion judgments changed over time in distinct epochs. In some epochs, rewards were equal; in others, one attribute was worth twice as much as the other. Behavioral data were analyzed using mixed-effects logistic regression and Drift Diffusion Modeling (DDM) to simulate normative attentional allocation.
Key Insights and Neural Mechanisms
The study revealed several critical aspects of hierarchical decision-making:
- Parallel Integration: The brain does not wait to select a target of attention before planning an action. Instead, it processes "what to attend to" and "how to act" simultaneously, integrating evidence for both levels within the same decision window.
- vmPFC as a Value Tracker: The ventromedial prefrontal cortex (vmPFC) tracks evidence in proportion to its reward value. If motion evidence is worth twice as much as color, the vmPFC's response to motion is approximately twice as strong, effectively calculating the utility of the incoming information.
- dACC as an Uncertainty Manager: The dorsal anterior cingulate cortex (dACC) tracks the absence of evidence (uncertainty) across a rostrocaudal gradient. Caudal (posterior) regions are sensitive to "Response Conflict" (motor uncertainty), while more rostral (anterior) regions are sensitive to "Attentional Conflict" (uncertainty about where to focus).
Results
Behavioral results indicated that participants did not merely follow the 2:1 reward ratio; they "over-attended" to the high-reward attribute. DDM simulations confirmed this strategy is mathematically normative when accounting for the "effort cost" of switching attention.
Neural analyses demonstrated a clear dissociation: while vmPFC activity increased with the strength of evidence for the chosen option, dACC activity increased when evidence was sparse or conflicting. Furthermore, the whole-brain analysis highlighted a striking distinction along the dACC: more caudal regions tracked the lack of evidence for the high-reward attribute, while more rostral regions tracked the presence of distracting evidence from the low-reward attribute.
Discussion and Limitations
These findings demonstrate that the prefrontal cortex dynamically integrates information across different levels of the decision hierarchy. However, the authors note that attentional allocation was inferred through behavioral modeling rather than direct measurement. Future research incorporating eye-tracking could provide more direct evidence of visual attention deployment. Additionally, while this study focused on reward-perceptual strength conflicts, further work is needed to determine if this dACC hierarchy persists in more complex scenarios involving entirely different tasks.
Summary updated