{
  "session": 72,
  "question": "Do the audited one-rune groups preferentially occupy observable structural positions?",
  "scope": "LP2 pages 0 through 55, with current apostrophe-joined lexical grouping; no glyph deletion, cipher search or translation.",
  "features": ["red_span", "section_edge", "paragraph_edge", "page_edge", "line_edge", "literal_stop_adjacent", "nonrunic_data_adjacent"],
  "feature_policy": "Binary per word. Edge means word begins or ends at that edge; a wrap inside a word is not an edge. Red uses the thirteen retained manually aligned spans. Literal stop means .,; in the immediately flanking rune gaps, without assigning a semantic role. Nonrunic data means an ASCII letter or digit in those gaps.",
  "primary": "Randomly reassign singleton labels among all lexical words within each section, retaining each section's exact singleton count. Statistic is the largest positive standardized excess across the seven features. Analytic means and variances use stratified hypergeometric sampling. Compare the same maximum on every trial.",
  "secondary": ["Same comparison limited to words of one through three runes", "Same comparison excluding all red-span words", "Short-word comparison excluding all red-span words"],
  "draws": 9999,
  "seed": 72001,
  "controls": ["Planted singleton labels preferentially at printed line edges, retaining section counts", "Five already decoded LP1 groups: report ordinary singleton identities and structural feature counts without treating the small sample as a population model", "Check gap mapping on within-word wraps, apostrophe joins and page joins; compare observed counts to direct feature sums"],
  "limits": "This is a conditional placement comparison, not an English generative null and not a test that proves words or markers. Short-word sensitivity partially addresses length and typesetting differences. Most line and punctuation positions come from the saved master, whose physical coverage is incomplete. No discovery claim follows from an enrichment by itself.",
  "decision": "A reproducible enrichment can nominate a source-inspection question, not justify relabeling every singleton as an instruction. An unremarkable placement result supplies no positive support for a general structural-marker explanation.",
  "timing": "Written before computing Session 72 outcomes; earlier singleton and layout findings were already known. This is not an independent preregistration."
}