Family Patterns · Corpus scans
Testing the shape at scale.
Three machine-scale evidence streams, each targeting a different failure mode of the single-witness argument. Findings recorded honestly — the positive results, the negative replication, and the canonical counter-tradition. Every claim carries a verbatim quote or a reproducible pipeline.
Jump to: Gutenberg pilot · Cinema scan · Novel survey
Gutenberg pilot — the 3-generation motif surfaces in period prose
Four-stage pipeline (rsync mirror → keyword scan → Haiku sieve → Sonnet judge → quote-verification) run end-to-end on 500 randomly sampled English Project Gutenberg books. Calibration, not evidence. The one legitimately load-bearing result is methodological: the pipeline finds the full three-generation motif in period prose, tags it correctly, and the anti-fabrication quote check holds.
The crosstab that would matter at scale (n=12 — do not cite)
Of the confirmed hits, twelve had both the same-sex in-law axis (harmony vs strife) and a child outcome (fortunate vs blighted) stated in the source:
| In-law axis | Fortunate | Mixed | Blighted |
|---|---|---|---|
| Harmony | 3 | 3 | 0 |
| Strife | 0 | 2 | 4 |
Zero off-diagonal corners, in the schema's predicted direction. Twelve machine-labelled data points prove nothing; a diagonal that holds at the full run's projected scale (hundreds of dually-rated hits, same-run counter-cases included) would be worth arguing about. The full-corpus run projects ~145,000 extracts, ~21,000 confirmed hits, ~750 three-generation criss-cross rows for ~$800.
Named exhibits (all quote-verified against source)
- Lycophron (Herodotus, via Lodge's Best of the World's Classics) — the complete strife motif in a single classical source: Gen 1 Procles (mother's father) at war with Gen 2 Periander (father, who killed the Gen 2 carrier Melissa). The grandfather-grandson affinity is itself the channel that ignites the feud ("who killed your mother?"). Descent bond strained; Lycophron banished, beggared, dead estranged. mgf→grandson, strife, blighted.
- King Baldwin IV (Besant, Jerusalem) — moralised cross-line heredity, stated in the source: "the evil life of Jocelyn bore its fruits in the leprosy of his grandson." Gen 1 Jocelyn of Edessa → unnamed daughter → Baldwin.
- Elizabeth Montagu's son (her letters) — the mother-claiming-the-boy phenomenon in the primary source: her infant, in her own eye, has her father's eyes. Claimant class = parent.
- Henry II (Bright's History of England) — never-met modelling of the maternal grandfather (Henry I). The gold class of Phase-2-via-legend cases.
- Louis Agassiz (his Life and Correspondence) — harmonious counterpart: maternal grandfather the "good physician," mother the carrier of the ambition, outcome fortunate.
- J. Pierpont Morgan (Foote) — flagged namesake case and an explicit resemblance-contrast claim: named for the maternal grandfather, "hard to find a greater contrast." The namesake-inflation control earning its keep.
Bias attached, per Gutenberg spec §9: pre-1929 Anglo-American literate corpus, overwhelmingly male subjects; biography selects for prominence; Phase 1 cannot state base rates (only the same-run counter-cases, which are recorded, can — in Phase 2); namesake convention inflates the mgf→grandson signal (flagged per hit). The paternal-grandmother / granddaughter crossing is near-empty in the pilot — the predicted Victorian-publishing artifact, not schema evidence either way.
Cinema register scan — cross-sex parent-child dialogue is denser in emotional vocabulary
Vocative-address dyad detection on two film-dialogue corpora, followed by an emotional-word density scan on the dialogue exchanged within each identified parent-child pair. The Cornell corpus shows a clear cross-sex lift; the modern IMSDb corpus does not cleanly replicate — documented honestly.
Cornell Movie-Dialogs — 617 films, 114 identified dyads
| Combo | Dyads | Words | Emo / 100w | |
|---|---|---|---|---|
| Father–Daughter | 16 | 13,454 | 3.37 | cross-sex |
| Mother–Son | 30 | 19,167 | 3.03 | cross-sex |
| Mother–Daughter | 16 | 10,623 | 2.72 | same-sex |
| Father–Son | 52 | 40,258 | 2.65 | same-sex |
| Cross-sex (MS + FD) | 46 | 32,621 | 3.17 | |
| Same-sex (MD + FS) | 68 | 50,881 | 2.66 |
Cross ÷ Same = 1.19 ×. Both cross-sex cells beat both same-sex cells; the ordering FD > MS > MD > FS is monotone. Controlled for the child's gender: given a son, mother–son beats father–son 3.03 vs 2.65 (1.14 × lift for the cross-sex parent); given a daughter, father–daughter beats mother–daughter 3.37 vs 2.72 (1.24 ×). Both "mommy's boy" and "daddy's girl" halves of the schema's 2-generation claim show up separately, with the daughter case the stronger of the two.
IMSDb — 1,095 modern films, 186 dyads: replication does not hold cleanly
The Cornell finding deserved a replication on a second, more modern corpus. IMSDb is where the honest half of the story lives: the emo-density effect flips or flattens.
| Combo | Dyads | Emo / 100w | |
|---|---|---|---|
| Mother–Daughter | 34 | 3.50 | same-sex |
| Father–Son | 71 | 2.93 | same-sex |
| Father–Daughter | 39 | 2.86 | cross-sex |
| Mother–Son | 42 | 2.74 | cross-sex |
| Cross-sex | 81 | 2.80 | |
| Same-sex | 105 | 3.09 |
Cross ÷ Same = 0.91 × — the direction reverses. Two candidate explanations, both plausible, neither cleanly separable in v1:
- Pipeline noise. IMSDb has no gender labels (inference from SSA name data covers ~35 % of characters); no explicit reply-chain edges (sequential-line heuristic); and screenplay parsing is messier than Cornell's clean JSON. Top-20 audit shows ~4 residual false positives. These systematically push toward null.
- Real corpus / era shift. IMSDb includes Black Swan, Carrie, The Fault in Our Stars, A Good Person, Persian Version, Barbie — a genuinely rich seam of 2000s–2020s mother-daughter drama. The MD cell being highest is not obviously fake. It might reflect a real shift in what emotional dialogue writers give to mother–daughter pairs, an under-written combination in earlier eras.
The volume direction (cross-sex exchange count vs. random pairing under per-film cast gender ratio) is weakly preserved on both corpora at ~1.05 ×. The two-corpus signal for the schema's intimate-track claim is therefore mixed — directional on volume, absent-or-reversed on emo density in modern data. Treat the Cornell 19 % lift as provisional, not settled, until the noise and the corpus-era shift can be separated.
Canonical-novel survey — the schema is a counter-tradition claim
Narrator-voice resemblance-and-inheritance claim extractor run on five public-domain multi-generation novels. The finding is honest and worth carrying: only one of the five stages the schema's cross-sex maternal-grandfather chain as its central inheritance mechanism. Three stage same-sex paternal-line skip instead. That is important context — the schema is asserting a pattern the Western literary tradition mostly does not reach for.
| Novel | Explicit 3-gen chain | Schema? |
|---|---|---|
| Buddenbrooks (Mann, 1901) | Arnoldsen → Gerda → Hanno (maternal grandfather → mother → son) | ✓ |
| The Forsyte Saga (Galsworthy, 1906–21) | Old Jolyon → young Jolyon → Jolly/Jon (paternal-line same-sex) | ✗ |
| The Way of All Flesh (Butler, 1903) | John → George → Theobald → Ernest (paternal-line 4-gen skip) | ✗ |
| Seven Gables (Hawthorne, 1851) | Colonel Pyncheon → Judge Pyncheon (paternal 5–7 gen skip) | ✗ |
| Wuthering Heights (Brontë, 1847) | 2-gen only: Linton ← Isabella (cross-sex); Cathy Jr ← Edgar (cross-sex, partial) | ✓ (2-gen) |
Four of five novels either confirm the schema (Buddenbrooks, Wuthering Heights at 2-gen) or stage its opposite (Forsyte, Way of All Flesh, Seven Gables). This does not refute the schema — canonical novelists were working with their own theories of heredity, and same-sex paternal-line skip is exactly the pattern the schema's §9.5 two-channel model predicts for craft transmission, which is what family-saga novels naturally foreground (business, trade, family style, "how we do it").
Hawthorne's own meta-thesis is worth quoting; his Seven Gables narrator explicitly names the same-sex paternal pattern as the mechanism of the Pyncheon curse:
"it is too fruitful a subject, this of hereditary resemblances,—the frequent recurrence of which, in a direct line, is truly unaccountable, when we consider how large an accumulation of ancestry lies behind every man at the distance of one or two centuries." Hawthorne, The House of the Seven Gables, 1851
The schema's claim is about essence, which sits at a level that literary tradition addresses less often and less directly than family style, business succession, or physical resemblance. That the schema's specific chain is the pattern only Mann reaches for is neither refutation nor confirmation — it is the counter-tradition the schema is arguing against, named openly.