{
  "session": 57,
  "status": "fixed before computing local frequency results",
  "question": "Do extra red-heading boundaries inside the existing substantial sections coincide with an unusually large local change in rune composition?",
  "heading_definition": "Enlarged red initial with adjacent red heading runes. Exclude normal-sized red labels, numeric item labels and red callout passages.",
  "unsolved_scope": "LP2 pages0-55, excluding solved pages56-57",
  "heading_page_offsets": [[0,0],[3,0],[3,119],[6,71],[8,0],[15,9],[23,0],[27,0],[33,0],[33,91],[39,119],[40,0],[54,0]],
  "extra_heading_page_offsets": [[3,119],[6,71],[33,0],[39,119],[54,0]],
  "window_runes": 64,
  "heading_handling": "The red heading ends at its visually aligned delimiter. Omit the heading and compare the preceding64 runes with the following64 body runes.",
  "statistic": "For each boundary, use the two-sample multinomial likelihood-ratio G statistic for its2x29 rune-count table. Primary statistic is the maximum G across the five extra boundaries.",
  "reference": "Random word-start cuts within each boundary's same old section, omitting a gap equal to that heading's rune length. Both64-rune flanks must lie inside the section and avoid every annotated red heading and every gap containing nonrunic ASCII letters or digits. Sample independently and reject trials whose compared windows overlap within a section.",
  "null_meaning": "A descriptive random-placement reference in the observed text, conditional on the eligibility and nonoverlap rules. Not a model of random English, an identified cipher, or a universal no-reset hypothesis.",
  "simulations": 4999,
  "seed": 3301057,
  "tail": "One-sided plus-one Monte Carlo comparison of the maximum G. This accounts for choosing the largest of the five displayed boundaries under the specified placement reference.",
  "control": "A synthetic uniform background with30-percent concentration on runeF in each left window and30-percent on runeI in each right window. Use the same window locations, candidate rules and maximum statistic. This is a deliberately strong sensitivity check, not a power guarantee for real key changes.",
  "reporting": "Report every boundary and its G, the primary result, reference quantiles, all candidate counts and the control. Do not change window size or select another statistic after seeing outcomes.",
  "stop_rule": "No key search. If the maximum is not unusual, do not promote a secondary local fluctuation. A positive result would need a separate validation and would not identify plaintext or an encryption reset.",
  "limitations": "Five retrospective boundaries, one window size. A key or cipher-state change can preserve rune frequencies and remain invisible to this test. Plaintext-topic differences could also produce a composition change. Word-start matching does not reproduce paragraph syntax or semantic roles."
}