Do ≥5σ claims hold up better than 4–<5σ claims? Preregistered Stage-0 feasibility pilot for a cohort study of Physical Review Letters, 2000–2015 (W4, v0.4.3)

How often do published claims of new physical effects later prove wrong, and do claims reported at 5σ or above hold up better than those at 4 to below 5σ? This project will trace the subsequent outcomes of positive empirical claims in Physical Review Letters (2000–2015), comparing their originally reported test significance. Signal claims and their physical interpretations will be assessed separately. This deposit preregisters Stage 0 only, a feasibility pilot. The population comprises PRL articles listed in NASA ADS with an arXiv classification in hep-ex, nucl-ex or astro-ph, including subcategories. Comments, replies, errata, publisher’s notes and retractions are excluded. A documented count before registration returned 4,224 records. Initial sampling targets 100 articles across four disjoint strata based on numerical sigma expressions of at least 4σ, significance language, claim language and the remainder. Smaller strata are included in full. If coding agreement falls below 80%, one externally versioned rule revision followed by up to 20 additional articles is permitted. The 20-hour working-time cap may be extended only through an externally versioned budget amendment. The pilot assesses eligible-paper counts per significance band, source accessibility, coding feasibility and workload. Selection rules, decision thresholds, conservative treatment of unclear information and normative files bound by SHA-256 checksums are specified in advance. Harald Zierhut is the sole author and human coder. A separate, documented AI extraction serves as a consistency check. AI assistants also supported planning, drafting, programming and review; all plan decisions were approved by the author. Prior case knowledge and exposure are disclosed; complete blindness is not claimed. No pilot sample has been drawn and no outcomes have been coded. Later stages require separate registrations.