When Does Hierarchy Arise? A Threshold Argument for the Dominance of the Pyramid over Flat and Network Structures

We ask: from what number of entities N does a flat or network organisation begin to spontaneously restructure into a hierarchy, and why does the pyramid win? We present a parsimonious argument based on three independent mechanisms — cognitive capacity, communication cost, and task asymmetry — and show that the emergence threshold is not a single number but a triple (Nc, Nₖ, Nᵣ) sensitive to task class: a cognitive threshold set by Dunbar's limit on mutual acquaintance (~150), a communication threshold set by the O (N²) scaling of pairwise ties against bounded individual capacity (~10), and a coordination threshold set by the deadline for collective decisions (5–10 under combat pressure, 10³–10⁴ under asynchronous conditions). We derive a closed-form expression for the critical size at which the binding threshold is crossed, N* (Cₘax, Tcrit) = min (Nc, Cₘax + 1, ⌊b^ (Tcrit/τ₀) ⌋), and argue that a network structure — while resolving the communication threshold — still loses to hierarchy on task asymmetry (Garicano 2000) and on addressability of responsibility. The resulting necessity theorem states that any stable coordination architecture over a population exceeding Nc under time-critical, high-stakes, heterogeneous tasks must contain at least one hierarchical layer. The predicted logarithmic growth of hierarchical depth in population size is tested against an empirical sample of eight real organisations spanning nearly five orders of magnitude (start-up, hospital, Roman legion, university, Alphabet, Catholic clergy, US Army, Walmart): ordinary-least-squares regression yields L ≈ 1. 21 log₁₀ N + 0. 06 with R² ≈ 0. 86, corresponding to an effective branching factor b ≈ 6. 7, well within the Graicunas range. The paper adds no new mathematical result to the existing literature but unifies the lines of Simon (1962), Graicunas (1937), Williamson (1975), Dunbar (1992), Radner (1993), Garicano (2000), and Powers & Lehmann (2014) into a single three-threshold framework, separates three independent reasons why flat and network forms fail, and checks the depth prediction against organisational data. Keywords: hierarchy, organisational structure, bounded rationality, Dunbar's number, span of control, Graicunas, task asymmetry, logarithmic depth, emergence threshold, formal social theory