Cognitive capacity and control in the evolution of intelligence

This paper presents an evolutionary model of working memory that bridges cognitive theory and evolutionary pressures, revealing a synergistic yet prioritized relationship between storage capacity and control that predicts distinct cognitive regimes and diverse intelligence patterns across animal lineages.

The Gist

The Big Idea: Building a Better Brain

Imagine evolution as a massive construction project. Animals are constantly trying to build better "minds" to solve problems in their world, like finding food, avoiding predators, or navigating social groups.

But building a smart brain is expensive. It costs a lot of energy (calories). So, evolution has to be a very careful accountant. It has to decide: Where should we spend our energy?

This paper asks a specific question: Should we spend our energy on Capacity (how much information we can hold) or Control (how well we can manage and protect that information)?

The Two Main Ingredients

To understand the answer, let's use two metaphors:

1. Capacity = The Size of the Backpack.
   Imagine your working memory is a backpack.

   • Low Capacity: A tiny pencil case. You can only hold a few things. If you try to put too much in, things fall out.
   • High Capacity: A giant hiking backpack. You can carry a tent, a stove, and a week's worth of food.

2. Control = The Zipper and the Organizer.
   Imagine the backpack is full of loose items tumbling around in the wind (noise).

   • Low Control: The zipper is broken, and you have no pockets. The wind blows your important map away, and your snacks get mixed up with your socks.
   • High Control: You have a perfect zipper, Velcro straps, and labeled pockets. You can instantly grab the map you need and keep the wind from messing up your other items.

The Discovery: They Work Together (Synergy)

The researchers built a computer model to see how these two things evolve together. They found something surprising: Capacity and Control are best friends. They don't fight; they help each other.

• The Analogy: Imagine you are trying to find a specific needle in a haystack.
  • If you have a tiny haystack (low capacity), it doesn't matter how good your search skills are (control); there's barely anything to search.
  • If you have a huge haystack (high capacity) but bad search skills (low control), you will never find the needle because the wind blows the hay everywhere.
  • The Sweet Spot: If you have a huge haystack AND great search skills, you can find needles in massive piles of hay that others can't even touch.

The paper calls this Synergy. Having a bigger backpack makes your organizing skills more useful, and having better organizing skills makes a bigger backpack worth carrying.

The Evolutionary Strategy: "Big Backpack First"

So, if they are so good together, why don't animals just get both at the same time? The answer is in the cost.

The researchers found that evolution follows a specific order of operations:

1. Phase 1: The "Passive Storage" Phase.
   When an animal is just starting to get smarter, it's too expensive to build complex "organizing skills" (Control). Instead, evolution prioritizes Capacity. It's like buying a bigger backpack first. You need somewhere to put things before you can organize them.

   • Who fits here? Simple animals like jellyfish or sea anemones. They can hold a little bit of info, but they can't really "think" about it or protect it from distractions.

2. Phase 2: The "Control-Enhanced" Phase.
   Once the backpack is big enough, it becomes worth the energy to buy a good zipper and organizer. Now, the animal starts investing in Control. This is where animals like monkeys, dogs, and maybe even humans start to shine. They can hold a lot of info and focus on the important parts.

   • The Finding: The study tested humans and Rhesus monkeys. Both were in this "Control-Enhanced" zone. They both had big backpacks and good organizers.

3. Phase 3: The "Capacity-Heavy" Phase.
   Eventually, if an animal gets really smart, the organizers are working so well that adding more organizers doesn't help much anymore. At this point, the only way to get smarter is to make the backpack even bigger.

   • The Prediction: The researchers think that eventually, the most super-intelligent lineages might stop focusing on "organizing" and just focus on "carrying more stuff."

What Did They Actually Test?

To prove their theory, the researchers didn't just guess; they ran an experiment.

• The Game: They used a "Retro-Cue" task. Imagine you are shown four pictures of fruit. Then, the pictures disappear. A moment later, a clue (a flash of light) tells you, "Hey, remember the apple!"
• The Test: They played this game with 346 humans and 10 monkeys.
• The Result:
  • Humans had both a bigger "backpack" (Capacity) and better "organizers" (Control) than the monkeys.
  • Crucially: Even though humans are super smart, they still prioritized the size of the backpack over the quality of the organizers. This matches the model's prediction: Capacity is the foundation. You can't have effective control without enough space to store information first.

The Takeaway

This paper solves a puzzle about why animal intelligence looks so different.

• It's not just about being "smart." It's about how you spend your energy budget.
• The Rule: Evolution says, "First, build a big storage room (Capacity). Once that's done, hire a good manager (Control). If you get really rich, build a bigger storage room again."
• The Diversity: Different animals are at different stages of this construction project. Some are just building the walls; others are installing the security systems. This explains why a monkey is smarter than a fish, but a human might eventually evolve to be smarter than a monkey in a completely different way.

In short: You need a big brain to hold the information, but you need a smart brain to use it. Evolution says: Get the big brain first.

Technical Summary

1. Problem Statement

The paper addresses a fundamental gap in evolutionary biology and cognitive science: the lack of a unified framework explaining how environmental constraints shape the specific mechanisms of intelligence.

• The Gap: While it is established that environmental complexity drives the evolution of larger brains and higher intelligence, existing theories often treat "intelligence" as a monolithic trait or focus on abstract phenotypic outcomes without modeling the underlying cognitive mechanisms.
• The Core Question: How do two specific cognitive components—Capacity (the total amount of information that can be stored/processed) and Control (the ability to regulate and allocate resources to achieve goals)—co-evolve? Do they evolve via trade-offs or synergies, and how does metabolic cost influence this evolutionary trajectory?
• The Challenge: Characterizing this requires linking the restrictions on cognitive mechanisms (e.g., noise, interference) with the evolutionary pressures (e.g., metabolic energy constraints) that shape them.

2. Methodology

The authors developed a mechanistic evolutionary model grounded in the cognitive science of working memory, specifically using the retro-cue task as a paradigm.

A. Cognitive Model (The Mechanism)

The model simulates an animal tracking multiple items (e.g., fruit locations) in a noisy environment.

• State Variables:
  • Capacity ($\eta$): The total pool of cognitive resources available.
  • Control ($\kappa$): The rate at which resources are actively directed toward an optimal allocation strategy (e.g., prioritizing a cued item) to counteract noise.
  • Noise: Represented as Gaussian white noise ($\epsilon$) that corrupts representations over time.
• Dynamics: The allocation of resources to items ($x_i$) is modeled using a stochastic differential equation (Ornstein-Uhlenbeck process):
  $$dx_i/dt = \kappa(\hat{x}_i - x_i) + \epsilon_i + \lambda_i$$
  Where $\hat{x}_i$ is the optimal allocation setpoint (determined by cue reliability $q$), and $\lambda_i$ ensures total capacity is conserved.
• Recall Probability: The probability of recalling an item is a function of allocated resources, following a diminishing returns curve: $F(x_i) = 1 - e^{-x_i}$.

B. Evolutionary Model (The Selection Pressure)

The model treats the evolution of $\eta$ and $\kappa$ as an optimization problem balancing fitness benefits against metabolic costs.

• Fitness Function ($W$):
  $$W(\eta, \kappa) = \bar{R}(\eta, \kappa) - c(\eta + \kappa)$$
  Where $\bar{R}$ is the expected recall performance (benefit) and $c$ is the metabolic cost per unit of investment.
• Optimization: The authors used Monte Carlo simulations and an evolutionary algorithm to find the optimal values of $\eta$ and $\kappa$ that maximize fitness across varying levels of metabolic benefit and task complexity.

C. Empirical Validation

To test the model's predictions, the authors:

1. Conducted a new experiment with 346 human participants on a retro-cue task (tracking up to 4 items with partially reliable cues).
2. Re-analyzed existing data from 10 rhesus macaques ($Macaca mulatta$) performing a similar task.
3. Fitted the model to both datasets using a hierarchical Bayesian approach (Stan) with a neural network surrogate to approximate the likelihood of the stochastic simulation. This allowed for direct estimation of species-specific $\eta$ and $\kappa$.

3. Key Contributions

1. Mechanistic Bridge: The paper successfully bridges the gap between information-processing theories (cognitive science) and evolutionary theory by creating a mathematically explicit model where cognitive constraints directly influence evolutionary fitness.
2. Synergy Discovery: It demonstrates that capacity and control are synergistic, not competitive. Increasing capacity enhances the efficacy of control, and vice versa.
3. Evolutionary Regimes: The model predicts three distinct evolutionary regimes based on the level of metabolic investment:
   • Passive Storage: Low investment; capacity exists without control.
   • Control-Enhanced: Intermediate investment; control is heavily favored due to high synergy with existing capacity.
   • Capacity-Heavy: High investment; capacity is prioritized over control because the marginal returns of adding more control diminish when capacity is already high (noise is sufficiently suppressed).
4. Quantitative Prediction of Diversity: The model provides a framework to classify different animal lineages into these cognitive regimes based on their metabolic investment and environmental complexity.

4. Results

• Synergy: The analysis confirmed that capacity and control amplify each other. Control is most effective at intermediate levels of capacity; with very low capacity, noise dominates and control is useless; with very high capacity, control is already effective, so additional control yields diminishing returns.
• Prioritization of Capacity: Across all levels of metabolic benefit, evolution prioritizes investment in capacity over control. This is because control cannot function without a space (capacity) to hold representations. Mathematically, capacity contributes linearly to performance, while control contributes sublinearly ($\sqrt{\kappa}$).
• Empirical Findings (Humans vs. Monkeys):
  • Humans exhibited significantly higher absolute values for both capacity ($\eta$) and control ($\kappa$) compared to rhesus macaques.
  • Relative Investment: Both species showed a higher investment in capacity than control.
  • Regime Placement: Contrary to the hypothesis that high-investment species (humans) would be in the "capacity-heavy" regime, the data suggested both humans and monkeys currently reside in the "control-enhanced" regime. This implies that for these species, increasing control still yields significant marginal gains, likely due to the complexity of their environments and the specific nature of the tasks tested.

5. Significance

• Explaining Cognitive Diversity: The paper offers a principled explanation for why different animals possess different "kinds" of intelligence. It suggests that diversity arises not just from different environmental pressures, but from the specific interaction between metabolic constraints and the functional synergy of cognitive components.
• Testable Framework: By defining capacity and control mathematically, the authors provide a method to empirically measure these traits across species, moving beyond qualitative descriptions of "smart" or "dumb" animals.
• Implications for Brain Evolution: The findings suggest that the evolution of complex brains follows a trajectory where capacity is the prerequisite for effective control. Lineages may evolve through a sequence: first expanding storage (capacity), then refining allocation (control), and potentially returning to massive capacity expansion once control efficiency plateaus.
• Future Research: The model highlights the need for tailored experiments to determine if highly intelligent species (like cetaceans or corvids) have indeed transitioned into the "capacity-heavy" regime, and how task complexity and cue reliability modulate these evolutionary paths.

In summary, this work provides a rigorous, quantitative framework demonstrating that the evolution of intelligence is driven by a complex interplay between the storage limits of the brain and the regulatory mechanisms of attention, constrained by the metabolic cost of maintaining these systems.