Cognitive control networks in human and macaque

Using fMRI during a multi-step saccadic maze task, this study demonstrates that while macaques possess a distributed brain network corresponding to the human multiple-demand system for cognitive control, there are notable species differences in the specific organization and extent of lateral frontal cortex activation.

The Gist

Imagine your brain has a super-organized project manager. In humans, scientists have known for a long time that this manager lives in a specific set of neighborhoods across the brain. They call this the "Multiple-Demand" (MD) system.

Think of the MD system like a Swiss Army Knife. No matter what complex problem you face—whether it's solving a math equation, planning a route, or switching tasks quickly—this team of brain regions lights up to help you figure it out. It's the part of your brain that says, "Okay, what's the goal? What are the rules? Let's make a plan and execute it."

For years, scientists have wondered: Do monkeys have this same Swiss Army Knife? Or is this high-level "project management" a uniquely human superpower?

This paper sets out to answer that question by putting both humans and macaque monkeys in a digital maze.

The Experiment: The Maze Race

The researchers designed a video game-like task for both species.

• The Human Version: You see a grid on a screen. You start in the middle. You have to use your eyes (saccades) to jump from square to square to reach a hidden goal. Along the way, some paths are blocked (red) and some are open (yellow). You have to remember the goal, look at the options, and choose the shortest path. It's like navigating a maze while blindfolded, but you can only move one step at a time.
• The Monkey Version: The monkeys played a simplified, two-step version of the same game. They had to remember where the goal was, wait for a signal, and then jump to the correct open spot to get a fruit smoothie reward.

They also had a "Control" version of the game where there was no goal and no choices. The path was just a straight line of forced moves. This was the "autopilot" mode, used to see what the brain does when it's just following orders without thinking.

The Findings: A Family Resemblance

When the researchers looked at the brain scans (fMRI) of both humans and monkeys while they were solving the maze, they found some fascinating similarities and differences.

1. The "Yes, They Have It" Moments (The Similarities)
Just like humans, the monkeys' brains lit up in the same general neighborhoods when they had to think hard.

• The Frontal Lobe (The CEO): Both species activated the front part of the brain, which is responsible for planning and decision-making.
• The Parietal Lobe (The Map Reader): Both species used the back-top part of the brain to understand space and location.
• The Insula (The Alarm Clock): Both species activated the deep center of the brain that helps with awareness and switching attention.

This suggests that monkeys do have a "Multiple-Demand" network. They aren't just reacting to the world; they are building a mental model of the task, holding the goal in mind, and making decisions, just like we do.

2. The "But..." Moments (The Differences)
While the locations were similar, the shape of the activation was different.

• Human Precision vs. Monkey Bulk: In humans, the "project manager" team is made up of nine distinct, tiny neighborhoods that work together but stay separate. It's like a boardroom with nine specific experts sitting at a round table.
• The Monkey "Blob": In the monkeys, these nine neighborhoods seemed to merge into one big, single patch of activity. It's less like a boardroom and more like a single, large open-plan office where everyone is shouting at once.
• The Missing Piece: Humans showed a lot of activity in the "Dorsal Attention Network" (the part of the brain that helps you focus your eyes on specific things). The monkeys didn't show this as clearly. The researchers suspect this might just be because the monkeys' data was a bit "noisier" or because the task was too simple for them to need that extra level of focus.

The Big Picture: What Does This Mean?

Think of the human brain as a high-tech, modular smartphone with many distinct apps running in separate windows. The monkey brain is like a rugged, older model phone that does the same core functions but with fewer distinct windows and more overlapping features.

The Takeaway:
This study confirms that the hardware for complex thinking (the ability to hold a goal in mind and make flexible decisions) is shared between humans and monkeys. We didn't invent the "project manager" from scratch; we inherited it.

However, humans seem to have refined the system. We took that basic monkey network and evolved it into a more specialized, finely-tuned team of nine distinct regions. This evolution likely allowed humans to handle much more complex, abstract, and multi-layered problems than our primate cousins.

In short: Monkeys have the same "brain tool" as humans, but humans have upgraded the tool to have more specialized, distinct parts, allowing us to tackle even harder puzzles.

Technical Summary

1. Problem and Objective

The study addresses a fundamental question in comparative neuroscience: Does the macaque brain possess a functional analogue to the human "Multiple-Demand" (MD) system?

• Background: In humans, the MD system is a well-established, distributed network of brain regions (lateral/dorsomedial frontal, anterior insula, parietal, and posterior temporal cortex) that shows increased activation across a wide variety of cognitive demands. It is hypothesized to encode a "task model" necessary for cognitive control.
• The Gap: While electrophysiological studies in macaques suggest similar neural coding for task events, and limited imaging studies hint at overlapping regions, there has been no direct, whole-brain fMRI comparison between human and macaque activation patterns using a comparable cognitive task. Previous estimates of a macaque MD system relied on functional or structural connectivity in anesthetized animals, which may not reflect the active cognitive control network.
• Objective: To directly compare whole-brain fMRI activation patterns in humans and macaques performing a multi-step spatial navigation task requiring goal-based decision-making versus a control task with similar motor demands but no goal-based decisions.

2. Methodology

Experimental Design

• Task: A multi-step saccadic maze task.
  • Maze Condition (High Demand): Subjects navigate a grid from a central start point to a specific goal location. At each step, they must choose between two "open" (yellow) locations and one or more "blocked" (red) locations. The correct choice requires maintaining the goal in working memory and selecting the path that minimizes distance to the goal.
  • Control Condition (Low Demand): Subjects follow a pre-determined path where only one location is open at each step. No goal maintenance or complex decision-making is required.
• Subjects:
  • Humans: 34 participants (fMRI).
  • Macaques: 2 male rhesus monkeys (Macaca mulata).
• Task Variations:
  • Humans: Performed a complex maze with variable path lengths (2, 4, or 6 steps).
  • Macaques: Performed a simplified version restricted to 2-step paths to ensure task engagement and feasibility within the scanner environment.

Imaging and Analysis

• Human fMRI:
  • Scanner: 3T Siemens Prisma.
  • Preprocessing: HCP minimal preprocessing pipelines (undistortion, motion correction, surface-based registration to CIFTI grayordinate space).
  • Analysis: General Linear Model (GLM) contrasting Maze vs. Control. Results mapped to the HCP multimodal parcellation (7) and network definitions (8) to identify activation in core MD parcels, Dorsal Attention (DA), and Cingulo-Opercular (CO) networks.
• Macaque fMRI:
  • Scanner: 3T whole-body scanner with custom coils.
  • Preprocessing: FSL and MrCat tools. Data included rigorous artifact rejection for motion and global intensity fluctuations.
  • Analysis: GLM contrasting Maze vs. Control. Results mapped to the F99 macaque surface and parcellated according to the 27-region atlas.
• Cross-Species Comparison:
  • Macaque activation maps were warped to the human cortical surface using a resting-state fMRI-based transformation (RheMAP-Surf) to facilitate direct spatial comparison.

3. Key Results

Human Findings

• Replication of MD System: The human data robustly replicated the canonical 9-patch MD system.
• Network Expansion: Activation extended beyond the core MD parcels to include significant portions of the Dorsal Attention (DA) network (particularly dorsal parietal cortex) and adjacent regions of the frontoparietal network.
• Specificity: Strong activation was observed in lateral frontal, dorsomedial frontal, anterior insula, parietal, and posterior temporal cortices.

Macaque Findings

• Broad Correspondence: The macaque brain showed widespread activation related to cognitive demand, with significant overlap in:
  • Dorsomedial Frontal: Pre-SMA and anterior cingulate (areas F6, 24c, 24c').
  • Lateral & Medial Parietal: Areas 7a, 7op, 7m, 23b/c, and 31.
  • Insula/Orbitofrontal: Activation in Iam extending to orbitofrontal cortex.
  • Posterior Temporal: A patch in anterior V4v corresponding to the human posterior temporal MD patch.
• Key Divergence (Lateral Frontal Cortex):
  • Humans: Showed multiple distinct activation patches across lateral frontal cortex.
  • Macaques: Showed a single, large, largely dorsal activation patch extending from the dorsal premotor cortex (F7) through the frontal eye fields (8ad) to the dorsolateral frontal cortex (9d, 46d).
  • Ventral Absence: Unlike humans, the macaque data showed minimal to no activation in the ventrolateral prefrontal cortex (vlPFC) in the combined analysis, though one individual monkey showed some ventral extension.
• Key Divergence (Parietal):
  • Humans showed strong activation in the Dorsal Attention (DA) network (dorsal parietal).
  • Macaques showed primarily ventral parietal activation; dorsal parietal activation was absent in the combined group data (though present in one individual).
• Unique Macaque Activation: Significant activation was observed in sensorimotor cortex (areas 3a/b), likely reflecting the physical movement constraints and lack of head fixation compared to humans.

4. Key Contributions

1. Direct Cross-Species Validation: This is the first study to directly compare whole-brain fMRI activation patterns of cognitive control in humans and macaques using a matched behavioral paradigm.
2. Refinement of the Macaque MD Network: The study identifies a specific set of macaque regions that functionally correspond to the human MD system, validating the existence of a distributed cognitive control network in non-human primates.
3. Identification of Species Differences: The study highlights that while the network exists in both species, the topography differs. Specifically, the macaque lateral frontal activation is more consolidated into a single dorsal patch, whereas the human system is fragmented into multiple distinct patches, with a notable difference in the recruitment of ventrolateral frontal and dorsal parietal regions.
4. Methodological Advancement: Demonstrates the feasibility of using simplified, goal-directed tasks in awake macaques to map cognitive networks, providing a bridge between electrophysiological single-unit data and human fMRI.

5. Significance and Implications

• Evolution of Cognitive Control: The findings suggest that the distributed MD system is an evolutionarily conserved mechanism for cognitive control. However, the human system appears to have undergone expansion and fragmentation, particularly in the lateral frontal cortex, potentially supporting more complex or specialized forms of cognitive control.
• Interpretation of Connectivity Studies: The results suggest that networks defined by structural connectivity or resting-state functional connectivity in anesthetized animals may not perfectly map onto the active cognitive control network observed during task performance.
• Future Research: The absence of vlPFC activation in the macaque group (despite its known role in control) suggests that future studies may need larger sample sizes or different task parameters to fully capture the ventral components of the macaque MD system. The study reinforces the utility of the macaque as a model for human cognitive control while cautioning against assuming exact anatomical homology.

In conclusion, the paper confirms that macaques possess a distributed brain network recruited by cognitive challenge that closely mirrors the human MD system, but with distinct topological differences, particularly in the lateral frontal and parietal cortices.