The Connectivity Constraint
Why bigger brains cannot simply keep the same wiring density
The scientific idea
A larger brain does not merely require more neurons. It must also connect those neurons, and axons and synapses take up space, consume energy, and impose delays. This creates a fundamental scaling problem.
Suppose a network contains (N) neurons. If every neuron could connect with every other neuron, the number of possible undirected pairwise connections would be
\[ C = \frac{N(N-1)}{2} \]
Neuron number therefore grows as (N), but the number of possible pairwise connections grows approximately as (N^2). Doubling the number of neurons would roughly quadruple the number of possible connections in a fully connected network.
Brains do not solve this problem by wiring every neuron to every other neuron. Instead, biological nervous systems are sparsely connected and highly structured. Most neurons communicate with only a tiny fraction of all other neurons. Many connections remain local, while a much smaller number of longer-range projections link regions, modules, and processing systems.
This organization allows brains to increase neuron number without requiring connectivity, wiring volume, and metabolic cost to increase as though the nervous system were an all-to-all network.
Adding neurons is relatively cheap. Connecting every new neuron to everything else is not. Large brains can scale only by being selective about their wiring.
Launch The Connectivity Constraint
Keep the HTML file and this Quarto file in the same folder.
What is new in this version of the demo
The earlier sparse view made the point mathematically, but it looked too much like a set of identical hub-and-spoke rosettes. That regularity weakened the teaching effect.
The revised sparse view still preserves the same conceptual lesson, but it now reorganizes the network into a more appealing and more plausible neural landscape:
- irregular local neighborhoods rather than identical circular modules,
- variable cluster sizes and shapes rather than repeated rosettes,
- mostly short local connections within neighborhoods,
- a small number of selective long-range bridges between neighborhoods, and
- a much clearer visual contrast between dense clutter and economical organization.
This matters pedagogically. The student should not merely be told that modularity is a solution. The solution should look like a solution.
How to run the demonstration
The demonstration begins with a small network in Dense toy network mode. In this deliberately extreme condition, every displayed neuron is connected to every other displayed neuron.
1. Begin with the dense network
Click + Add 5 neurons several times.
Watch four things at once:
- the number of neurons,
- the number of actual edges,
- the total wire length, and
- the Metabolic + mass cost meter.
At first, adding a few neurons seems harmless. The network is small, the number of lines is manageable, and the cost meter remains in the green range.
Continue adding neurons.
The important observation is that the number of neurons and the number of connections are no longer increasing at the same rate. If the network contains (N) neurons, a fully connected undirected network contains
\[ \frac{N(N-1)}{2} \]
connections.
For example:
| Neurons | Pairwise connections |
|---|---|
| 5 | 10 |
| 10 | 45 |
| 20 | 190 |
| 40 | 780 |
| 80 | 3,160 |
The number of neurons has increased by a factor of 16 from 5 to 80, but the number of pairwise connections has increased by a factor of 316.
That divergence is the connectivity constraint.
2. Use Auto-grow
Press Auto-grow to let the network expand automatically.
The visualization will rapidly become crowded with connections. More importantly, the simulated wiring burden will begin to dominate the cost meter. The meter progresses from green to yellow and eventually to red.
When the cost passes the demonstration’s limit, growth stops and the network freezes.
This is not meant to represent a literal biological threshold. The stopping point is a teaching device that makes the scaling problem visible: a system that attempts to preserve extremely dense connectivity eventually spends an increasing fraction of its resources simply on wiring.
The demonstration contains two different costs.
Neuron tissue grows approximately linearly because each additional neuron adds another unit of cellular machinery.
Wiring burden depends on both the number and length of connections. In dense mode, the number of edges grows approximately as (N^2), so wiring quickly becomes the dominant term.
The sparse/modular solution
Once the dense network has hit its limit, do not reset the demonstration.
Instead, click Sparse / modular.
The same neurons reorganize into irregular local neighborhoods spread across the canvas. Most connections now remain short and local, while only a small number of longer-range bridges connect one neighborhood to another.
Several things happen immediately:
- the number of actual connections falls sharply,
- total wire length decreases,
- the wiring component of the cost meter drops,
- the system returns to the green range, and
- additional neurons can now be added without immediately overwhelming the wiring budget.
This is the critical comparison. The demonstration has not solved the problem by removing neurons. It has changed the architecture of connectivity.
What the student should see in the new sparse layout
The visual transformation is part of the lesson.
In Dense mode, the screen becomes a dark and increasingly unreadable mesh. That clutter is the visible counterpart of a bad scaling rule.
In Sparse / modular mode, the same network becomes legible again. The student now sees:
- neighborhoods of neurons that are locally interlinked,
- lots of short wiring,
- only a few long-distance pathways, and
- empty space between neighborhoods rather than exhaustive connection everywhere.
The point is not that the sparse display is anatomically literal. It is that it now better conveys the principle that a larger nervous system must rely on organized selectivity rather than indiscriminate wiring.
Local connections and long-range connections
The sparse network contains two simplified kinds of organization.
Local connectivity links neurons within neighborhoods. In the revised visualization, these are mostly short-range connections among nearby neurons, producing a more organic local mesh.
Long-range connectivity is concentrated in a much smaller number of selective bridges between neighborhoods. These links allow information to travel across the network without requiring all possible distant pairs to be directly connected.
The exact arrangement in the simulation is schematic. Real brains contain many overlapping networks, recurrent circuits, topographic projections, hierarchical pathways, and specialized long-range fiber systems. The important principle is that long-distance connectivity is selective rather than exhaustive.
Why this matters for brain evolution
Increasing neuron number can improve computational capacity only if the nervous system can still fit inside the head, supply its tissue metabolically, and transmit information efficiently.
Axons create several costs.
They require volume. A long axon must physically occupy space, and large numbers of long-range fibers contribute substantially to white matter.
They require energy. Action potentials, synaptic transmission, maintenance of ion gradients, and cellular transport all consume metabolic resources.
They also create time delays. Signals traveling between distant parts of a larger brain cannot be treated as instantaneous.
A large brain therefore cannot simply be a geometrically enlarged version of a smaller brain with every neuron preserving the same relationship to every other neuron. Scaling requires changes in organization.
Sparse connectivity, local processing, modular organization, and selective long-range pathways are all ways of controlling the cost of a larger nervous system.
The “77-pound brain” thought experiment
The demonstration refers to the provocative idea of a human brain with a mass on the order of 77 pounds if increasing neuron number were accompanied by much denser, rodent-like scaling of connectivity and wiring volume.
The number is useful because it makes the constraint memorable. It should not be interpreted as the output of the network simulation or as a claim that a rodent brain is literally an all-to-all network.
The dense mode is deliberately much more extreme than any real mammalian nervous system. Its purpose is to expose the mathematics cleanly:
\[ N \longrightarrow N^2 \]
when connectivity approaches an all-pairs rule.
The biological lesson is broader. As brains become larger, they must become increasingly economical about which neurons communicate directly and how far those connections travel.
The Dense toy network is not intended to model an actual rodent brain.
Rodent brains are already sparse, structured networks. The comparison is meant to illustrate the general scaling pressure that arises if connectivity and wiring density are preserved too aggressively as neuron number increases.
The biologically important contrast is therefore not “rodents are fully connected, humans are sparse.” It is that increasing brain size places progressively stronger constraints on the fraction, length, and organization of neural connections.
What to look for
Run the demonstration several times and focus on a different quantity each time.
Neurons versus connections
In dense mode, compare the Neurons and Actual edges counters. Neuron number increases steadily, while connection number accelerates.
After switching to sparse/modular mode, notice that adding neurons no longer produces the same explosion in edge count.
Wiring rather than neuron number
Watch the two components of the cost meter.
Early in dense growth, neuron tissue accounts for much of the cost. As the network expands, wiring becomes increasingly important.
This illustrates a general principle in brain scaling: the difficulty is not simply making more neural tissue. A major constraint is connecting that tissue without allowing wiring volume and energetic cost to dominate the system.
Visual organization
Dense mode quickly becomes visually unreadable. That visual clutter is not itself the scientific argument, but it provides an intuitive counterpart to the mathematics.
The revised modular network remains much more legible because most interactions occur within local neighborhoods and only a few longer-range paths span the larger network.
This makes it easier for students to appreciate that the evolutionary solution is not merely fewer connections, but better-organized connections.
Controls
+ Add 5 neurons adds neurons manually.
Auto-grow repeatedly adds neurons until you stop the simulation or dense wiring exceeds the toy cost limit.
Dense toy network connects every displayed neuron to every other displayed neuron.
Sparse / modular reorganizes the same neurons into irregular local neighborhoods with limited long-range bridges.
Reset returns the simulation to its initial dense network.
The following keyboard shortcuts are also available:
- Space — start or stop Auto-grow
- D — switch to dense mode
- S — switch to sparse/modular mode
- R — reset
- Up arrow — add five neurons
A useful classroom sequence
A particularly effective way to use the demonstration is to avoid explaining the solution in advance.
First, begin in dense mode and add neurons slowly. Ask what the students expect to happen if the neuron count doubles.
Then use Auto-grow and allow the network to hit the cost limit.
At that point, ask a different question:
If removing neurons is not an acceptable solution, what property of the network could evolution change?
Only then switch to Sparse / modular.
In the revised visualization, the transition is especially useful because the same crowded network reorganizes into a visibly economical structure. Students can see that the answer is not to eliminate neurons, but to change the wiring architecture.
The broader principle
Brain evolution is constrained by physics.
More neural tissue can provide more computational machinery, but that tissue requires energy, space, communication, and time. The nervous system therefore cannot increase neuron number independently of the architecture that connects those neurons.
The solution is not one special anatomical trick. It is a family of related principles found throughout nervous systems:
- dense local circuitry,
- sparse long-range connectivity,
- modular and hierarchical organization,
- selective projection systems,
- hubs and bottlenecks, and
- extensive reuse of intermediate processing stages.
These organizational principles allow a nervous system to become larger without paying the impossible cost of connecting everything directly to everything else.
Brain size is constrained not only by how many neurons can be built, but by how economically those neurons can be connected.