Counting the Peace: Incidence, Composition and Scale of Large-Scale Armed Conflict Across the Reign of Leto II, Set Against Pre-Reign and Post-Reign Baselines, 9719–13768 AG

1. Introduction

Across the five centuries before Leto II's reign, the surviving registers record about 0.50 armed conflicts of at least a thousand deaths in every standard year, the rate in any single century lying between 0.46 and 0.54. Across the 3,500 years of the reign they record 0.025 a year, roughly one such conflict in forty years. Whether that twenty-fold fall shows the Peace suppressing organised violence on a large scale, as its rationale says, or only the thinning of what a single authority chose to record, is the question of this paper.

We write c. 13770 AG, some fifty years after the reign closed, in the Famine Times, with the monopoly on spice broken and populations beginning to disperse outward. The later vantage gives us what the Peace's contemporaries lacked, an after to compare it with, short and poorly recorded as it is. Every source and reference below predates 13770 AG.

In the Imperial account the Golden Path, the prescient course Leto II set for humanity, rested on a tranquility enforced across the whole reign and known as Leto's Peace, a monopoly on spice, and the breeding of the line descended from Siona, invisible to prescience. The Fish Speakers, an all-female force, enforced the Peace; on the rationale the God Emperor stated, women are less given to the male aggression it was meant to tame (Fish Speaker Instructional Office, 10232–13705 AG). We test one implication of the first condition, that large-scale organised armed violence was rarer under the Peace than before or after it.

2. Sources and Method

Two registers carry the reconstruction. The Imperial list comes from the dispatch and pacification returns of the Fish Speaker Command (Fish Speaker Command, 10219–13718 AG) and, for earlier centuries, from Imperial returns catalogued in earlier work by one of us (Wrexane, 13751 AG). The Landsraad list comes from registered copies of declared kanly, the formal vendetta between Houses, and of House loss returns (Landsraad Registry of Kanly and House Losses, 9719–13768 AG), catalogued for earlier centuries in earlier work by another of us (Orrande, 13744 AG). Those catalogues coded no episodes, parties or deaths; every figure below comes from our own coding of series held at the Imperial Archives in Kaitain.

An episode is a sustained engagement between organised armed forces, or between such a force and an armed population, on one world or linked worlds, with no gap of a year or more. It is large-scale if it has at least 1,000 recorded deaths. Entries in the two lists were matched by date, place and named parties, and each episode was coded by initiating party. Category A is war between Houses or polities; B, a rising against the central authority with the response to it; C, action begun by the Imperial force with no prior rising; D, communal, sectarian or resource conflict; E, the remainder. An episode was multi-world if forces had to be carried between worlds, which required a Guild passage. A stratum of 100–999 recorded deaths was also extracted, without any claim of completeness.

Three windows are used: the 500 standard years 9719–10218 AG before the reign; the 3,500 years of the reign, 10219–13718 AG by the conventional reckoning (Kereldin, 13734 AG); and the 50 years 13719–13768 AG after it. Muad'Dib's Jihad, 10193–10205 AG, enters the first window as one Imperial-initiated episode. A diffusion reconstruction places its dead at 38–92 billion (80% interval), several thousand times the deaths in the other 251 pre-reign episodes combined, so we count it in incidence and omit it from every comparison of scale.

Decade counts are overdispersed, with a variance about 1.3 times the mean in the two long windows, so we use a negative binomial count model fitted to each window, with a time term for trends. The post-reign window and sub-periods of the reign, too short to estimate dispersion, take exact count intervals. Deaths per episode, on a log scale, come from linear regression with standard errors clustered by decade. A two-list estimator, which assumes independent lists, estimates how many episodes both lists missed (Desmarrow, 12906 AG).

3. Incidence Across the Reign

The registers hold 252 pre-reign episodes, 88 in the reign and 31 in the fifty years after (Table 1). The pre-reign rate was 0.50 per standard year (95% CI 0.44–0.58) and the reign rate 0.025 (95% CI 0.020–0.032), a twenty-fold fall (95% CI 15–26).

The fall was front-loaded. The first century of the reign held 19 episodes (0.19 per year), the next four centuries 20 (0.050) and the last thirty centuries 49 (0.016). After the fifth century the rate did not drift: the change per century was −0.5% (95% CI −4.2% to +3.5%, p = .82), against −3.2% (95% CI −6.1% to −0.2%, p = .04) across centuries 2–35. We chose the break after inspecting the counts, so the flat slope is descriptive and its p-value conditional on that choice, and the late rate describes a period picked for being quiet.

Nothing before the reign anticipates the fall. The pre-reign rate changed by −2% per century (95% CI −11% to +8%, p = .68). After the reign the rate came back: the 31 post-reign episodes give 0.62 per year (exact 95% CI 0.42–0.88), 1.2 times the pre-reign rate (95% CI 0.84–1.8, p = .28) and 38 times that of the last thirty centuries of the reign (95% CI 24–60).

4. Who Fought, and How Large

The Peace did not abolish organised violence; it changed who practised it (Table 2). War between Houses and polities, category A, fell from 121 episodes to 6, a 141-fold fall in rate (95% CI 62–320), and five of the six fell in the first two centuries. Risings fell about tenfold and Imperial-initiated action about ninefold. These two categories always involve the Imperial force and rose from 37% of episodes to 78% (odds ratio 6.2, 95% CI 3.5–11.0; χ²(1) = 45.0, p < .001).

The intervals in Table 2 ignore clustering in time and are somewhat narrow. The reading is nonetheless clear: under the Peace the Emperor's standing force became the principal party to large-scale violence. That is a finding about who fought, not about whether the fighting was warranted.

Scale fell too, though less steeply than incidence. Excluding the Jihad, the median pre-reign episode killed about 3,200 people and the median episode of the reign about 2,100; the ratio of geometric means was 0.62 (95% CI 0.49–0.80, p < .001). In the post-reign window the geometric mean matched the pre-reign value (ratio 0.98, 95% CI 0.56–1.7).

5. Scale and Transport as Dimensions of the Fall

The fall steepened with scale. In the lower stratum of 100–999 recorded deaths the lists hold 508 pre-reign and 366 reign episodes, a 9.7-fold fall (95% CI 8.5–11.1) against 20-fold at large scale; the ratio of the falls is 2.1 (95% CI 1.6–2.7). Small episodes per large one rose from 2.0 to 4.2, so organised violence did not so much vanish as shrink, though this stratum is less completely recorded.

The fall also steepened with dependence on transport. Before the reign 74 of 252 episodes (29%) were multi-world; within it 3 of 88 (3.4%) were, all in the first two centuries. Multi-world episodes fell 173-fold (95% CI 54–550) and single-world episodes 15-fold (95% CI 11–19); the odds of an episode being multi-world were 11.8 times lower in the reign (95% CI 3.6–38; χ²(1) = 25.1, p < .001).

We offer a hypothesis here and do not claim to have shown it. A force that must cross between worlds needs a heighliner passage, the Guild's navigators depend on spice, and the Emperor held the spice. If the monopoly let the Peace refuse passage to armed forces, the episodes that needed passage would vanish first. The Fish Speakers on the ground would produce the same pattern, and the registers cannot separate the two.

6. Discussion

Four alternatives would also produce a twenty-fold fall. The first is a decline that began before the reign; the pre-reign rate shows no such drift. The second is that a single authority recorded less. Two-list estimates put completeness at about 86% before the reign (95% CI 80–92%) and about 80% within it (95% CI 68–97%), which corrects the fall to about 19-fold. Because the lists are unlikely to be independent, both figures are probably high, but the reign's records would have to be less than half as complete as the earlier ones for the fall to drop below ten-fold, and under 15% as complete for it to drop below three-fold.

The third alternative is that the Jihad exhausted the capacity for war, but the decade before the reign, 10209–10218 AG, held 7 episodes, or 0.70 per year (exact 95% CI 0.28–1.44), and shows no lull. The fourth is that violence declines with time whatever the regime, which predicts a low rate after the reign too; the rate was back at the pre-reign level within fifty years. That rise coincides with the Famine Times and the breaking of the monopoly, and our data cannot apportion it between the end of enforcement and famine and dispersal. The low rate belonged to the reign and not to the age.

The stated rationale holds that tranquility was enforced, and the results fit that description in three particulars. Suppression was strongest where violence was largest and where it needed transport. The violence that remained was mostly the force's own, 78% of episodes. And after the first five centuries, in which most of the fall took place, the rate held without detectable trend for three thousand years.

7. Limitations of the Record

Every series was kept by a party to the Peace or by the Houses it constrained. The Fish Speaker Command wrote the returns that supply 90% of the reign's episodes and had an interest in recording order; the Landsraad list, which captured only 35% of reign episodes, is a weak check on any omission. The two-list estimate assumes independence and so tends to overstate completeness, and the post-reign completeness of about 55% (95% CI 37–100%) makes the post-reign rate a floor.

Recorded deaths are rounded and often reported by the winning side, so the 1,000-death threshold is blurred. Initiating party was coded by consensus among the three of us, with no blind second coder, so the category and transport results are secondary to the incidence result. The reckonings of the reign differ by a decade or two at either end, which could move a handful of episodes but would not change the result. We count organised armed conflict only; famine, displacement and smaller violence lie outside the study.

Within these limits the evidence supports a plain statement. Large-scale organised armed conflict was about twenty times rarer across the reign than in the five centuries before it, and about thirty times rarer in its final thirty centuries. The fall was steepest at the largest scales and for conflicts that needed transport between worlds, it left the Imperial force as the principal party to the violence that remained, and it was reversed within fifty years of the reign's end.