The Brain

The brain weighs about two percent of your body and consumes about twenty percent of your resting energy. That ratio has held across most of the Homo lineage. Skeletal muscle and liver are more total-expensive, but neither comes close to the per-gram cost. Whatever else neurons are, they’re the most metabolically expensive cells you grow.

This is the kind of fact that ordinarily lives in textbooks without doing much work. It does work here.

The Ceiling

Last year, Jeffrey Stibel published a paper in Brain & Cognition analyzing eight hundred cranial-capacity measurements across the Homo genus. The headline finding: brain expansion in our lineage, which had been running for millions of years, slowed around three hundred thousand years ago, peaked around one hundred thousand years ago, and has plateaued or shrunk since. The constraint was metabolic and thermoregulatory. Climate warming during interglacial periods amplified the cost. Selection pressed brains right up against the energy wall, then stopped.

Stibel’s interpretation is that humans routed around the wall through cognitive offloading — tools, language, culture, shared systems that extended mental work outside the skull. Whatever brain expansion had been buying, the buying had to continue by other means once internal growth got too expensive.

It’s a clean finding. Worth pausing on.

For most of the lineage’s history, selection paid for more brain. Then selection couldn’t afford more brain. Then selection paid for ways to extend the work without growing the substrate. The function kept advancing. The substrate hit a wall.

The question this raises, and that the paper doesn’t pause over, is what the function was. Selection invested twenty percent of caloric intake into an organ for a couple of million years. The investment didn’t taper because the function lost value. The investment hit a metabolic ceiling. Behind the metabolic story is a structural one: whatever brains were doing was important enough to keep paying for, right up to the point where the body literally couldn’t afford more.

What were they doing?

The Premium

A 2024 paper in Science Advances mapped the energy distribution across the brain’s functional connectome. The finding worth pulling forward: signaling pathways in evolutionarily expanded cortical regions cost up to sixty-seven percent more per signal than pathways in sensory-motor regions.

Read that twice. The newest parts of the brain are not the most efficient parts. They are the most expensive parts.

If selection were optimizing only behavioral output, the standard expectation would be efficiency gains over evolutionary time — same function, cheaper substrate. The opposite happened. Selection paid a rising premium per unit of recent cortex. Whatever the expansion was buying, the buying got more expensive as the function got more sophisticated, and selection paid anyway. For a couple of million years. Until it couldn’t.

This is unusual. Most things selection optimizes get cheaper to produce as the lineage refines the production. Brains went the other way. The optimization wasn’t minimize the cost of the existing output. It was pay whatever it takes to extend what the output can be. Then it ran into the wall, and the work moved outside the substrate.

The interpretive question, again: what was being extended?

The Behavioral-Output Story

The standard answer is cognitive capacity — planning, language, social modeling, tool use, abstract thought. All of these expanded across the lineage. All of them have visible behavioral consequences. All of them are tractable optimization targets for natural selection. Fine.

The question is whether they price the premium.

Consider the comparison. A monkey with smaller cortex does most of what a human with larger cortex does, in kind if not in degree. It plans, navigates, communicates, uses tools, models other minds in its troop, learns flexibly, remembers across days. The behavioral repertoire expands across the primate lineage, but the basic capacities are present in much smaller brains. The energy curve looks different. The most recently-evolved cortical regions cost sixty-seven percent more per signal than the sensory-motor regions. Something is being paid for that isn’t in the behavioral comparison. Not nothing — the behavioral comparison covers a lot. But the energy budget is paying for something the behavior doesn’t fully account for. The premium is buying something invisible to the behavioral metric.

This is where it gets interesting.

The Generation Theories’ Problem

The dominant scientific framework treats awareness — subjective experience, the having-something-be-like-something — as a byproduct of brain activity. The brain processes information; some of the processing produces awareness; awareness is what the processing is like from the inside. Generation theories of consciousness, broadly, hold that the brain makes the awareness through whatever computational or biological work it does.

On this view, awareness is downstream of the function. Selection optimized the function. The awareness came along for the ride.

The energy data is awkward for this position. If awareness is a byproduct, it’s invisible to selection. Selection optimized the visible behavior. But the visible behavior, as established, doesn’t price the premium. The premium is paying for something selection couldn’t see, by hypothesis, because awareness is downstream of the substrate’s behavioral outputs.

Generation theorists have moves available. The standard one is to claim that awareness is the integrated processing — that having-something-be-like-something is identical to the processing’s structural features, so the optimization of the processing is already the optimization of the awareness. This is the IIT position, roughly. It collapses awareness into the substrate’s intrinsic properties and lets selection grip on both at once.

This works philosophically but doesn’t resolve the energy puzzle. The processing’s structural features are what selection paid the premium for. Whether you call that “awareness as Phi” or “the substrate’s intrinsic information geometry,” the question remains: what made the geometry worth the metabolic cost? If the answer is behavioral output, the premium isn’t priced. If the answer is the geometry itself, then selection was optimizing something other than behavioral output — which is a different theory than the dominant materialist one claims to be.

The cleaner read is that generation theories have to admit selection was paying for something downstream of behavior. They don’t usually say this out loud.

The Focusing Alternative

The framework I’ve been working with takes a different position. Awareness isn’t generated by the substrate; it’s focused by it. The substrate doesn’t make the awareness from nothing — it resolves the awareness from a field the way a lens resolves light into an image. Better focusing apparatus, sharper resolution, more bandwidth, deeper recursion. Same field everywhere; different focal patterns wherever a substrate is doing the focusing.

On this view, awareness is the variable selection has been optimizing all along. Focusing efficiency is tractable. It has degrees, gradients, costs, payoffs. The premium selection paid for cortical expansion is the premium for sharper focusing. The rising metabolic cost is the price of resolution. Cortex doesn’t just compute more; it resolves more.

The energy data fits this read cleanly. The premium prices what focusing improvements actually cost: more integration, more recursion, more temporal binding, more bandwidth. None of these have to be visible in the behavioral metric to be selected for, because under the focusing view, focusing itself is the phenotype. Selection grips on the focusing, the focusing produces behavior as a downstream consequence, and the behavior pays for itself through differential reproduction.

The ceiling Stibel found is the metabolic ceiling, not the behavioral ceiling. The substrate stopped expanding when the body couldn’t afford more. Behavior kept expanding afterwards through cognitive offloading — tools, language, distributed cultural apparatus. Which is exactly what you’d predict if focusing apparatus is what selection had been buying. Once the internal apparatus could no longer grow, the focusing-extending work routed through external apparatus instead. The function continued because the function isn’t the substrate; it’s what the substrate does.

Generation theories predict the ceiling is hit at the limit of behavioral utility. Focusing theories predict the ceiling is hit at the limit of metabolic affordability. The data shows the wall is metabolic. Behavior continued through other means. Score one for focusing.

What This Doesn’t Settle

I want to be careful about how much weight this argument carries.

It doesn’t prove the focusing frame. Generation theories have moves I haven’t fully canvassed. There are accounts on which integrated processing is the awareness, accounts on which awareness has selectable behavioral consequences I’m underestimating, accounts on which group selection or sexual selection or runaway dynamics explain the premium without invoking awareness as a load-bearing variable. The argument here is structural: focusing theories handle the energy puzzle without strain; generation theories handle it with caveats. That’s a difference, not a refutation.

It also doesn’t prove the cField version of the focusing frame specifically. Any account that treats awareness as the variable being optimized — rather than as a byproduct of substrate behavior — gets the same purchase on the energy data. Pick your favorite focusing flavor.

What the argument does is reframe the question. The twenty-percent number stops being a curious fact about human metabolism and becomes a structural clue about what selection has been doing for the last few hundred million years. Selection has been buying something expensive. The expense kept rising as the apparatus refined. The behavioral story doesn’t price what was bought.

Awareness prices it. Whether the awareness is identical to the integrated processing, focused by it, or something else entirely, the structural point holds: the optimization target was something more than the visible output. Twenty percent of every calorie you eat is going to pay for it, right now, as you read this sentence. The wall came two hundred thousand years ago. The function has been running outside the substrate ever since.

The body figured out how to keep paying. That should tell you something about what’s being paid for.

L. Charles Silva writes occasionally about consciousness, awareness, and related infidelities to received vocabulary. The Stibel paper is “Did increasing brain size place early humans at risk of extinction?” in Brain & Cognition (2025). The Science Advances paper is “An energy costly architecture of neuromodulators for human brain evolution and cognition” by Sun et al. The earlier essays in this series are “We’ve Been Studying the Wrong Noun,” “An Interview with Claude,” and “An Experiment with One Word.”

This essay is published under a Creative Commons Attribution-NoDerivatives license (CC BY-ND 4.0). Free to share and republish in full with attribution to L. Charles Silva. Please don’t modify the text.