Meso-structural domains, not aggregated networks, reveal the action of structural selection

Ecological networks are often analyzed as aggregated structures, an approach that has yielded important insights but implicitly assumes that selection acts uniformly across communities. We refine this perspective by showing that structural selection becomes detectable only when analyses focus on meso-structural scales. Using a detailed trophic network, we quantified local structural environments through order-2 egonets and evaluated how structural traits shape interaction geometry. Aggregated representations captured broad patterns but showed no structural gradients, whereas egonets revealed strong axes of differentiation. Hierarchical asymmetry emerged as the dominant meso-structural trait, with local connectivity contributing secondary structure. Principal component analyses and Manhattan distances showed that meso-structural domains retain the heterogeneity through which selection acts. Structural selection was sparse but concentrated in hierarchical domains, identifying the meso-structural scale as the level at which evolutionary signals become detectable in ecological networks. Together, these results position structural selection as an evolutionary process acting on ecological structure, expanding how selection can be conceptualized in complex systems. One-Sentence SummaryStructural selection becomes detectable only at meso-structural scales, revealing evolutionary gradients that vanish in aggregated networks and expanding how selection is conceptualized in structured ecological systems. Teaser textEcological networks are usually analyzed as large, aggregated structures, but this perspective hides the evolutionary signals that operate at finer scales. By zooming into meso-structural domains, local neighborhoods captured through order-2 egonets, we uncover strong structural gradients that disappear in whole-network representations. These meso domains reveal where structural selection operates, exposing hierarchical asymmetry as the dominant axis of differentiation. Our results show that aggregated networks smooth out the heterogeneity through which selection acts, whereas meso-structural environments retain the variation that shapes interaction geometry. This work reframes how we detect evolutionary processes in ecological networks and shows that structural selection is sparse, localized, and fundamentally meso-structural, converging with classical natural selection when selection is mediated by interaction geometry--a perspective consistent with recent theoretical work on enemy-tracking constraints in tri-trophic systems by Miranda-Perez et al. in prep.