Reconstructing the Unreconstructable: Casualty Scale and Heighliner-Network Diffusion in Muad'Dib's Jihad, c.10193–10205 AG

1. Introduction

Muad'Dib's Jihad followed Paul Atreides's accession to the Golden Lion Throne and ran for roughly twelve years, from about 10193 to about 10205 AG. Its scale is the most contested figure in Imperial history. The Emperor himself, in a reckoning made during the war's final phase and preserved by his court, put the dead at some 61 billion, with 90 worlds sterilized and 500 others demoralized. Other surviving tallies place the total nearly seven times lower or more than twice as high.

Writing in 10248 AG, we do not adjudicate between those tallies by preferring one witness. We ask instead what range of totals the war's logistics can physically support. Every Fedaykin company and every Qizarate mission that left Arrakis did so aboard a Spacing Guild heighliner, because the Guild's monopoly on foldspace transport left no other route between worlds. The Guild's transit records therefore fix, within limits, where the Jihad could go and how quickly.

A second question follows from the first. Theological accounts explain why mobilization began and why it was so rapid, locating the cause in the fusion of Paul's charismatic authority with a legitimating legend seeded in advance. Those accounts say less about why some worlds were devastated and others passed through the war almost untouched. We test whether a world's position in the transport network predicts the severity of its toll better than its population or its prior exposure to Bene Gesserit prophecy-seeding.

2. Sources and Method

The transport layer rests on heighliner passage and cargo manifests in the Spacing Guild Operational Archives (series TR-4), consulted under restricted access, and cross-checked against the CHOAM Directorate's freight assessment ledgers. From these we reconstructed a directed graph of 1,148 worlds (nodes) joined by 6,930 regular transit links (edges), each link weighted by passage frequency in 10191–10193 AG. A second, political layer of Landsraad fief and alliance ties was coded from the Landsraad Public Archives. Both graphs are our reconstruction, not archival documents.

Casualty testimony comes from four families of record: the Imperial court histories, the Qizarate's enumeration of the fallen, the Landsraad House loss returns and Sardaukar after-action records held on Salusa Secundus. Pre-Jihad populations are taken from CHOAM assessment ledgers, uncertain by perhaps a factor of two for peripheral worlds.

Estimation has two separate stages. Stage one calibrates the spread of the Jihad on timing alone, using the first documented arrival of Fedaykin or Qizarate personnel on 203 worlds. Stage two converts simulated reach into deaths through explicit per-world mortality assumptions. The casualty figures we judge most trustworthy, available for 37 worlds where two sources with opposed interests agree within a factor of two, are held out of both stages and used only for validation.

3. The Source Estimates and Their Interests

Each surviving tally was produced by an institution with a stake in the answer. The Qizarate, Muad'Dib's priesthood and the administrative arm of the Jihad, recorded some 140 billion dead and 132 worlds purified, its term for sterilized. Its enumeration, deposited with the Commission of Ecumenical Translators, served a doctrine in which the scale of the Jihad testified to its divine sanction; inflation served that purpose directly. Oderan's critique of the Jihad tallies treated this figure as arithmetic in the service of doctrine and set it aside; we extend his reading to all four families of record, since the minimizing sources were shaped by interest no less than the magnifying one.

At the opposite pole stand the Landsraad House loss returns, filed between 10206 and 10211 AG, which sum to about 9 billion dead and 31 sterilized worlds. The Houses had reason to minimize: acknowledged losses implied either defeat or collaboration, and several returns were filed by Houses that had supplied the Jihad with ships' stores. The Sardaukar after-action records are the most difficult to situate. The main body of the Sardaukar stayed with the exiled Shaddam IV on Salusa Secundus; we read the series, on our own interpretation of the archive, as observer reports from that exile combined with returns from detachments pressed into Imperial garrison service under the Atreides. The records document about 9.1 billion deaths across roughly 38% of the engagements attested in any source; scaled by that coverage share, they imply roughly 24 billion, a figure limited as much by coverage as by interest.

The Emperor's reckoning reaches us through the court histories compiled under Princess Irulan, whose office had every reason to present the reign as both mighty and burdened. Paul's own motive, as the court histories portray it, was self-indictment, which could push a count upward. Yet an Imperial count drawn from Qizarate reports and Guild returns would also inherit their gaps, and a reckoning made before the last campaigns closed could not include them. We therefore treat 61 billion as one interested witness among four.

4. A Transit-Network Diffusion Model

We model the Jihad as a contagion over the transit graph, following the contact-graph approach developed for heighliner routes in the Mentat literature. A world becomes eligible for arrival once a linked world hosts Jihad forces, with a per-year transmission probability proportional to the weight of the link. Only the transmission scale and a departure delay for Arrakis were fitted, by maximum likelihood against the 203 documented arrival dates. The calibrated model reproduces the observed median arrival lag of 2.9 years after the Jihad's opening (simulated median 3.1 years; 80% interval 2.6–3.7).

To move from reach to mortality we assigned each reached world to a severity band by simulated occupation length and applied band-specific mortality priors: 40–95% of the pre-Jihad population for worlds that suffered sterilization, and 0.5–6% for worlds that were occupied without sterilization. The low end of the first prior reflects usage in the tallies themselves, where a sterilized world is sometimes one whose agricultural and water capacity was destroyed, with deaths following over years through famine and dispersal; raising that lower limit to 70% shifts the 80% interval for total dead to 44–95 billion. These priors are assumptions, stated so they can be contested; they are the widest source of uncertainty in the model. Across 5,000 simulations the model places the number of affected worlds between 560 and 720 and total dead between 38 and 92 billion (80% intervals), with a median of 58 billion and a 95% interval of 29–118 billion. Only 11 of the 5,000 simulations (0.2%) produced 9 billion dead or fewer, and 38 (0.8%) produced 140 billion or more.

Validation used the 37 held-out worlds. For 29 of them (78%) the corroborated toll fell inside the model's per-world 80% interval, close to the nominal rate. The eight misses were split, with five worlds where the model overpredicted deaths and three where it underpredicted; all five overpredictions were agricultural worlds on which surrender preceded the arrival of Fedaykin units.

The source range of 9–140 billion is about 2.4 times as wide as the model's 80% interval of 38–92 billion (widths of 131 and 54 billion). Across the reached worlds, log betweenness centrality correlated with earlier simulated arrival (Spearman ρ = −0.52) and with longer simulated occupation (ρ = 0.47). The Emperor's figures sit within the model's intervals on both deaths and affected worlds (90 sterilized plus 500 demoralized gives 590). The Qizarate's total lies above even the 95% interval, and the Landsraad returns fall well below it.

5. Centrality and the Distribution of the Toll

A world's documented severity is an ordered category, so we fitted a proportional-odds ordinal regression. The outcome, coded for 486 worlds whose status can be established from at least one source, is unaffected (211 worlds), demoralized (214) or sterilized (61). Predictors, standardized to one standard deviation, were log betweenness centrality in the pre-Jihad transit graph, log pre-Jihad population and a documented seeding index. The last measures the density of Missionaria Protectiva activity, meaning the Bene Gesserit programme that planted the Panoplia Propheticus, a body of protective prophecies and superstitions, on selected worlds; we coded it on a four-point scale from Bene Gesserit and Commission records and treated it as quasi-continuous. Because neighbouring worlds shared exposure through the transit graph, standard errors were clustered by transit sector (41 sectors).

In the joint model, each standard-deviation increase in centrality multiplied the odds of falling in a more severe band by 2.41 (95% CI 1.96–2.97, p < .001). Seeding density also raised the odds of a worse outcome (OR 1.31, 95% CI 1.07–1.61, p = .01). Population, once centrality was included, showed no clear association (OR 1.22, 95% CI 0.99–1.50, p = .06); centrality and population were moderately correlated (r = 0.38).

Single-predictor models make the comparison directly. The centrality-only model had an AIC of 842, against 908 for seeding and 921 for population. Under ten-fold cross-validation it assigned 64% of worlds to the correct band, compared with 53% and 51% respectively, and 44% for a model that always predicts the modal band. Adding the Landsraad alliance layer as a further centrality measure improved AIC by only 3 points, which suggests that political ties added little once transport position was known.

6. Discussion

Two results stand out. The first is that the physical logistics of the war comfortably support a total in the tens of billions and give little support to either extreme in the record: fewer than 1% of simulations reach the Landsraad or the Qizarate figure, although the mortality priors leave that judgement open to revision. This does not vindicate the Emperor's figure; it shows only that the Guild's own manifests can accommodate it, and a count near the centre of an independent interval may still be right for the wrong reasons.

The second is the dominance of transport position. Worlds that stood on many shortest routes between others fell more often into the severer bands, whatever their size. The simulations reached such worlds earlier and held them longer, which offers a plausible mechanism for the association, though the records themselves cannot confirm it world by world. Seeded prophecy mattered too, and its effect runs in the expected direction, perhaps because seeded populations were drawn more readily into the Jihad's quarrels, or because the Qizarate dealt more severely with worlds whose prophecies it judged rival or corrupted. Its predictive weight, however, was well below that of centrality. Comparative work on implantation success across worlds indicates why the seeding index is a noisy measure: implantation took root unevenly, depending on local receptivity and the length of operative presence.

These findings complement the theological account without competing with it. Charismatic authority fused with seeded legend explains the speed and fervour of the first mobilization; the Guild's routes explain where that mobilization could be carried and in what order. The transit records indicate that the Jihad was also, perhaps chiefly in its distribution of suffering, a movement of freight.

7. Limitations of the Record

The Guild manifests are incomplete for worlds served by irregular or unrecorded passages, and the Guild's restricted archive may have withheld entries touching its own conduct during the war. Worlds destroyed early leave few records of any kind, so the 486 coded worlds overrepresent those that survived to report. Population figures carry wide error, and the mortality priors drive most of the width of the final interval. Clustering by sector addresses shared exposure only in part, and residual network dependence may make the regression intervals somewhat too narrow.

Only 37 worlds could be treated as corroborated, and agreement between opposed sources is itself only a proxy for accuracy. Our intervals should be read as a statement of what the surviving logistics permit. Further recovery of House and Guild records, especially from sterilized worlds, is the only route to narrowing them.