The Headless Escalation: How Commitment Accumulates Without a Chooser

Escalation is usually explained by a chooser or by its absence. Strategy places the ranking of the end at the centre, explicitly or as the standing priorities of an official mind; the literature on the man on the spot places it at the frontier; accounts of inadvertent escalation place it nowhere, because the outcome outran everyone's foresight. This thesis proposes a fourth position, which it calls accumulation: a sequence of acts, each within the authority of the office that takes it, none of which ranks the end the sequence adds up to, and each of which raises the cost of reversing it, until the apex accepts in full view what no one chose. Accumulation is neither strategy nor inadvertence. It is acceptance without choice.The thesis tests the claim by process tracing on 1818–1914, in two arenas: the imperial frontiers of Britain, France, Russia and the United States (Part A), where the arrival of the telegraph tests whether distance drives the pattern, and great-power crises from Crimea to July 1914 (Part B), where crises that ended in war are compared with crises that stopped short of it. A codebook fixed before coding, Bayesian inference over five families (accumulation and four rivals), hoops stated in advance, and strategic controls coded at the same depth decide each sequence. A sequence counts as accumulation only if the documents establish a causal sequence that excludes a ranking at the node as well as at the apex; the official mind is tested on its own prediction, that the centre's responses track fit with its standing priorities rather than the cost of reversal. The comparative tests (distance, the official mind, regime) rest on sixteen to eighteen tier-1 episodes and decide direction, not magnitude.The process-traced core also serves as the authentication protocol of an escalation chronotope (Filsecker 2026f): a structural representation of escalation episodes from 499 BC to 1931, built from degraded records and frozen before the primary record is read. Its primary coding is made by an AI coder under procedural blinding, independent of the author's prior reading of the cases; what the coder brings from its training is measured on an ablation sample, not assumed away. A degradation test measures what the degraded record recovers, and carries that error, as a lower bound, to the ancient episodes that cannot be authenticated. The ancient episodes extend the representation at hypothesis level: the thesis makes no claim that the pattern recurs across the period, and the degradation test is a test of this application of the method in one documentary domain, not a validation of the method in general.