The Distance Between Trees Is Part of the Forest

The gap belongs to the system

The distance between trees is part of the forest.

Count trunks and you learn how much wood stands there. You do not yet know whether a squirrel can cross without touching ground, whether a seed can move from patch to patch, or whether flame can find an unbroken road to the ridge.

Those questions belong to topology: the pattern of connection. A thousand trees in one block and a thousand trees scattered among fields contain the same count. They are not the same network.

Brain imaging faces the same trap. A region can show ordinary average activity while its relationships with other regions change. Fibromyalgia studies have reported altered links among sensory, salience, default-mode, insular, thalamic, and pain-modulation networks. Other work has found no broad change in global graph measures.

A network can change greatly while almost every node stays where it was.

When a cluster spans

Imagine a square map divided into cells. Each cell is forest with probability p and open ground otherwise. Neighboring forest cells join into clusters. At low p, the map holds islands. As p rises, clusters merge. Near a critical value, a connected group can suddenly span the map.¹

That is percolation. The change is geometric rather than explosive. Adding one cell near the threshold can join two large pieces. Total forest area changes a little; reachability changes a great deal.

Real forests are not coin flips. Trees cluster along water, slope, soil, fire history, and land use. Wind can carry embers across gaps. Fuel moisture and weather alter whether an edge is usable. A fire-spread model reported a connectivity threshold near 0.40 under arid conditions but no comparable threshold under regular weather.² Dryness changed the meaning of an edge.

Scale also changes the map. From far above, a narrow road may vanish and two patches appear joined. Zoom closer and the cut returns. Choose a maximum crossing distance for an animal and the network changes again.^1,3

Firebreaks exploit topology. Remove a narrow set of links and one spanning cluster can split into smaller ones. Yet the best break depends on wind, ignition, fuel, terrain, and spotting. A graph finds routes; it does not replace fire physics.

How to build a brain graph

First choose nodes: atlas regions, electrodes, sensors, or data-driven parcels. Then define edges: correlation, coherence, phase locking, tract estimates, or another measure. Set a threshold or keep weighted values. Every choice changes the graph.

Path length counts steps between nodes. Modularity asks whether the graph divides into communities. A hub has unusual centrality, but “central” has several definitions.

This flexibility is both power and danger. Two teams can scan similar people and build different networks because they used different atlases, frequency bands, preprocessing, edge measures, or density thresholds. They may have asked different graph questions.

A hub changes its company

Kaplan and colleagues built resting-state networks from 264 regions in 40 women with fibromyalgia and 46 controls. Global efficiency, clustering, path length, and modularity did not differ across the tested densities. Yet hub placement did: insular, sensorimotor, and other regions differed in centrality or rich-club membership.⁴

The posterior insula’s eigenvector centrality correlated with clinical pain and with glutamate-plus-glutamine. A mediation analysis linked those measures, but the design was cross-sectional. It could not determine which change came first.⁴ In a small validation cohort, several discovery-cohort hub findings did not replicate.

An EEG study used persistent homology, which follows how network components merge as the edge threshold moves. The fibromyalgia network merged more slowly and showed reduced global theta-band connectivity.⁵ This avoids betting everything on one cutoff, but it remains specific to EEG, theta frequency, and that sample.

Resting-state fMRI offers more maps. Napadow and colleagues reported stronger default-mode and executive-attention links with the insula, and selected connections tracked spontaneous pain.⁶ Fallon and colleagues found both stronger and weaker default-mode links with other regions.⁷

Flodin and colleagues reported decreased links among thalamic, insular, premotor, sensorimotor, and prefrontal regions, while other analyses in the same paper found no group difference.⁸ “Connectivity is reduced” cannot hold all directions, regions, and methods at once.

A null map is still a map

Counterevidence matters. A graph-theory study of young women found no broad difference in intrinsic global brain architecture between fibromyalgia and control groups.⁹ Age, symptom burden, medication, sample size, and methods could help explain why results differ. None can be selected after the fact as the answer without further tests.

Thresholding remains a central problem. A correlation matrix is dense: every region has some numerical relation with every other. Analysts often keep the strongest fraction of edges or test a range of densities. A hub can appear or disappear when the cutoff moves.

Forest maps face the same issue. Decide that patches connect across 50 meters and you obtain one graph. Allow 100 meters and components fuse. A threshold should answer a physical or biological question, not merely produce a pleasing picture.

Machine learning can combine many edges and classify groups, but prediction is not explanation. A model may exploit scanner quirks, motion, age, medication, or sample-specific structure. Held-out and external validation matter.

The path, not the count

Change one edge and most of the adjacency matrix stays the same. Yet if that edge bridges two modules, reach can change sharply. Location matters more than count.

This is why there is no single pain center hiding like a tree marked with paint. Pain arises through distributed sensory, affective, cognitive, autonomic, and motor systems. Connectivity findings show relationships. They cannot alone say whether those relationships caused pain, followed pain, compensated for it, or arose from another factor.

A forest teaches restraint through abundance. No tree is the forest. No gap is merely empty. A trail, creek, road, canopy bridge, and wind-blown ember each define connection differently.

At dusk, trunks become separate black bars. Above them, crowns touch. An owl crosses the valley without landing. On the map below, the forest looks broken. In the air, it is one path.