Founders, Drift and the Dissolution of Managed Lineages: A Population-Genetic Reconstruction of Nine Scattering Cohorts, 13730–14310 AG

1. Introduction

Nine dispersal cohorts reconstructed from the Sisterhood's lineage registers stand at a median of about 85% of the reference neutral gene diversity at c. 15230 AG, with a range of 59% to 96%, yet a median of only three of the twelve managed lineages were present among the founders of any one cohort. Diversity was thinned. The arrangement that gave the lineages their meaning, twelve distinct lines held apart by register-controlled pairing, was carried in full by none.

The Scattering followed the death of Leto II, the Famine Times and the breaking of his spice monopoly, and noships, which prescience cannot see, made it possible. People left in numbers that no Guild or Sisterhood count recorded. We write c. 15230 AG, some fifteen centuries after Leto II's death and so at a later vantage than is usual, when scattered peoples have begun to return and Sisterhood agents to meet their descendants. For the Sisterhood the matter is practical: the managed lineages were the working stock of its breeding programme, and an earlier reconstruction of its pairing registers found them grown apart on the target index, with about 0.59 of index variance between lineages in the last recorded generations and an effective breeding size near 32.

Three questions follow. How much neutral gene diversity did dispersed cohorts keep? How many of the twelve lineages did each carry? What structure, if any, replaced the one the registers maintained? We answer for the part of the dispersal that left a trace in Sisterhood or Guild records.

2. Sources and Method

Four bodies of evidence carry the reconstruction. The Spacing Guild's heighliner manifests for declared settlement parties number 311, dated 13730–13995 AG, after which the series stops for reasons the Guild's files do not give (Spacing Guild, Office of Transit Records, 13730–13995 AG). The Sisterhood's lineage departure notices number 276, dated 13730–14310 AG, each recording a registered household that left its registered domain (Bene Gesserit Chapter House, 15188 AG). Matching by household name and date linked 118 notices to 74 manifests; the other 237 manifests carry no registered household. The remaining 158 notices match no manifest, and we treat them as departures by noship.

Those 158 notices are the only record of their departures, and no outside observer could have supplied another: in 47 known-position trials the Guild never detected a noship by instrument or by Navigator prescience, which bounds per-attempt detectability below about 7.6% (see the companion study of noship stealth, cited below). The notices thin after about 14150 AG, so late cohorts are the least recorded and every unmanifested cohort is a lower bound.

The reference collection holds 612 marker profiles, 51 for each of the twelve lineages, banked in the closing centuries of the Peace and typed on forty neutral markers, variants not expected to bear on the traits the Sisterhood selected (earlier work by one of us, Kaverin, 15204 AG). Its pooled gene diversity, the mean expected heterozygosity across markers, is 0.415, and its between-lineage share of marker variance, a fixation-index analogue, is 0.14 (95% interval 0.12–0.15, bootstrap over markers). The earlier reconstruction also found twelve lineages in the last generations of the pairing registers, though we cannot show they are the same twelve.

New typing validates the projection and does not enter it. Sisterhood agents travelling on Guild contract met four settlements whose declared destinations match those of manifested cohorts (Spacing Guild, Office of Transit Records, 15168–15216 AG), and we typed 63 residents on the same markers: 18, 17, 16 and 12 per settlement.

We grouped registered adults into nine cohorts (Table 1), M1–M5 by manifest destination and U1–U4 by departure window, 1,230 adults in all; unregistered companions are ignored. For U cohorts, founder numbers were scaled by a factor between 1.0 and 1.6 to allow for unfiled notices. Each cohort was projected to the vantage by a resampling simulation of 2,000 draws: founder genotypes sampled lineage by lineage from the reference frequencies, logistic growth to a plateau, effective size at 0.6 of census and a 30-year generation (Oressen, 15056 AG), and allele frequencies resampled each generation. Growth classes were expansive, slow and stalled, at 0.30, 0.12 and 0.05 per generation to plateaus of 20,000, 2,000 and 400 breeding adults. Class followed the manifest's declared destination or, for U cohorts, a size rule: stalled below about 150 founders (Dessarel, 14930 AG). Retention is gene diversity relative to the reference value (Talvaris, 15092 AG).

3. Founding Parties and the Lineages They Carried

The founding parties were small and drawn from few lineages. The nine cohorts range from 38 to 360 registered adults. Lineages present at founding range from one to nine, with a median of three, and seven cohorts carried fewer than half of the twelve. Unevenness cut the count further: M2 had seven lineages, but 131 of its 200 founders came from one, giving an effective number of lineages (the reciprocal of the summed squared shares) of 2.2. Across the nine cohorts the effective number has a median of 2.1 and a range of 1.0 to 4.4.

Founding itself cost little diversity. Sampling 38 founders, the fewest in any cohort, removes only about 1.3% of source diversity (1/(2N) for N founders), and median retention immediately after founding was 0.95 (range 0.79 to 0.98). The exceptions, U3 and U4 at 0.79 and 0.85, were drawn almost wholly from one lineage, which holds on average about 1 − 0.14 of the pooled diversity. U4 matches that expectation; U3 started lower because its dominant lineage was less diverse than most.

4. Drift after Founding

Most of the loss came afterwards. By c. 15230 AG the median cohort retained 0.85 of reference gene diversity, a fall of about 15%, and cohort values ran from 0.96 for the largest expansive cohort to 0.59 for the smallest stalled one (Table 2). Expansive cohorts lost about 0.1% of diversity per generation; the three smallest stalled cohorts, M5, U4 and U3, lost 1.0% to 1.2%. In those three a median of 17.5% to 27.5% of markers had become fixed or lost, against none in M1, M2 and U1.

Drift after founding accounted for more than half of the total loss in five cohorts and for more than four-fifths in three (M4, M5 and U2); founding accounted for more than half in the other four (M1, M2, U1 and U3). In most cases what a party met after landing mattered more than who it was when it left.

Setting effective size at 0.4 or 0.8 of census moves the median retention to 0.79 or 0.88, a 27- or 33-year generation to 0.84 or 0.85, and halving or doubling the plateaus to 0.83 or 0.86.

Typed settlements give a partial check. Predictive intervals include typing noise from samples of 12 to 18, so an upper limit can exceed 1. Three typed values lay inside their 95% predictive intervals: 0.93 against 0.91–1.02 for M1, 0.83 against 0.77–0.95 for M3 and 0.81 against 0.74–0.94 for M4. The settlement matched to M5 returned 0.34 (bootstrap interval over markers 0.19–0.49), below its interval of 0.44–0.73, which points to a smaller founding party or an unrecorded later bottleneck.

5. What Replaced the Lineage Structure

Within a cohort, lineage structure could not outlast the register. Under random mating the variance among individuals in lineage ancestry halves each generation, so under 1% of the founding variance remains after seven generations, and every cohort had 34 or more. Only strict lineage endogamy would have preserved the old divisions, at the cost of splitting a party of 38 to 360 founders into smaller sub-populations with added drift. The typed samples are too small to test for substructure inside a settlement.

Structure did reappear along another axis. Differentiation among the nine cohorts, the between-cohort share of marker variance, was 0.07 at founding and 0.18 at the vantage (95% interval 0.16–0.20) in the simulations, against 0.14 between lineages in the reference collection. The descendants differ from one another more than the lineages did, by place of settlement and not by line of descent.

Pooling restores diversity faster than lineage coverage. One cohort chosen at random holds a median 0.85 of reference gene diversity, three pooled hold 0.95, five 0.97 and all nine 0.99. A single cohort carries a median of three lineages at founding and three pooled carry nine. Five carry all twelve in 67 of 126 combinations (53%), seven in 34 of 36 (94%), and eight in every case.

6. Discussion

The managed structure was lost in two stages with different causes. At founding, the composition of small parties removed most lineages from each cohort, and no later process could restore an absent lineage. Afterwards, drift in stalled cohorts removed diversity at a rate set by their size. The end of register-controlled pairing joined the stages, because without it no cohort could hold its lineages apart.

Neutral markers understate what was at stake. The earlier reconstruction found 0.59 of target-index variance between lineages, far above our 0.14 for neutral variation, which indicates that selection and assortative pairing, not isolation, kept the lineages distinct on the traits that mattered. We measured no index in any descendant, but if matching on the index lapsed with the register, its structure would have eroded at least as fast as the neutral structure.

For the Sisterhood, no single returning population can serve as a window onto the old lineages. Reconstituting them would require pooling at least five cohorts, accepting that rare variants carried by few founders are gone, and rebuilding the pairing control that gave the lineages their identity.

7. Limitations of the Record

The sources are partial and interested. The Guild's manifests and the Sisterhood's notices were compiled by bodies with their own reasons to record or omit departures, and the notices cover only registered households. Unmanifested cohorts rest on notices alone, scaled by an assumed 1.0 to 1.6, and late cohorts are the least recorded. Admixture with unregistered companions would raise diversity and blur lineage identity further.

The projection fixes quantities the record leaves open: effective size, growth class, plateau and a uniform 30-year generation. Sensitivity runs move the median retention by at most 0.06 but leave the growth-class assignment untouched. Validation rests on four settlements, matched to cohorts by destination sector alone, and all four typed values fell below their predicted medians, which hints that the projection is slightly optimistic. We attribute no settlement to any named returning order; the Sisterhood's hypothesis that one such order descends from Leto II's Fish Speakers merged with Reverend Mothers is contested, and nothing here bears on it.

Within these limits the record supports three conclusions. No cohort carried the managed structure intact, and five of nine carried a quarter of the lineages or fewer, so the lineages as the Sisterhood defined them did not survive dispersal in any settlement. Neutral gene diversity largely survived in aggregate, about 85% in the median cohort and about 99% across all nine, but its arrangement changed from lineages to settlements. What remains of the managed structure is fragments spread across settlements, without the register that held them together.