Wiring explained 5% (worm), 11% (zebrafish) and 24% (fly) of reliable pair-specific FC; much fly and mouse wiring signal was a smooth spatial field; of three behaviours only swimming raised the share

Question: how much of the pair-to-pair pattern of functional connectivity (FC) is specific to the wiring of each pair, and not to node effects, geometry or brain state? Data: worm (40 freely moving C. elegans, 119 NeuroPAL neurons, Cook connectome), fly (50 flies, 75 neuropils, FlyWire), larval zebrafish (22 fish, 65 regions, Legare et al. 2025 projectome) and mouse group fMRI (304 parcels, Allen tracer model). Method: we removed node effects and a distance function from group-mean FC, measured the split-half reliability of the pair-specific residual (the ceiling), and reported the fraction of the ceiling that wiring features explain (share). Nulls: strength- and distance-matched surrogate wiring, and a new neighbour-swap null that gives each node the wiring of a near node. States were compared inside individuals with permutation tests and jackknife CIs. Results: pair-specific FC was highly reliable (ceiling 0.92 worm, 0.85 fly, 0.89 zebrafish), but wiring explained only 0.047 [0.041-0.053], 0.236 [0.175-0.301] and 0.106 [0.080-0.132] of it. Surrogates gave 0.003, 0.076 and 0.023. Wiring of the nearest node kept 10% (worm), 20% (zebrafish), 78% (fly) and 70% (mouse) of the real effect, so most of the fly and mouse effect is a smooth spatial field; node-specific shares were about 0.04, 0.09 and 0.05. Forward vs reversal (worm) and rest vs walk (fly) did not change the share (p = 0.11, 0.15). Swimming nearly doubled it in zebrafish (0.086 to 0.165, p = 0.005; 0.077 to 0.117 after global signal regression, p = 0.014). Reliable state changes of the pattern in worm and during dark flashes were not explained by wiring (0.3% and 2%). In mouse, tracer-model R2 of pair-specific group FC was 0.41 under medetomidine+isoflurane, 0.17 under isoflurane and 0.12 awake, and 0.12, 0.04 and 0.04 beyond nearest-neighbour-swapped wiring. Limitations: wiring and activity come from different individuals; the ceiling counts individual differences as noise; region scale except worm; mouse has group data only and no ceiling; linear wiring model; several controls were added after the first results.