brain/
questionopenhistory

Was the Antonine Plague a structural shock to the Roman Empire?

Notes

Was the Antonine Plague a structural shock to the Roman Empire?

The question

The antonine-plague (165–180 CE) has been used by kyle-harper and earlier scholars (Niebuhr, Scheidel, Harris-Zelener) as a load-bearing data point for catastrophist-environmental theories of Roman decline. The argument depends on mortality being high — historically estimated at 10–25% of an empire of ~75 million people. A 2024 PLOS One epidemiological re-modeling (Karasaridis & Chalupa) produced maximum excess-mortality estimates of ~7% for measles, ~5% for bubonic-plague-with-rats, ~1% for standard smallpox. The two figures cannot both be right. Was the Antonine Plague a structural shock, or a manageable crisis that historians have mythologized via Galen's and Cassius Dio's overstatements?

Why it matters

  • For theories of Roman decline: if the plague is a manageable ~7% event, the environmental-catastrophist track (Harper 2017) loses its empirical anchor. Heather's external-pressure cascade (external-pressure-to-western-roman-collapse) doesn't depend on plague catastrophism and is unaffected.
  • For methodology: this is a case study in historical-source vs epidemiological-modeling disagreement. The historians read ancient texts as approximately accurate; the modelers find those texts inconsistent with disease-dynamics math.
  • For late-antique-little-ice-age and other environmental data points: if the Antonine Plague gets cut down, the broader catastrophist-environmental narrative loses momentum across the board.

What we currently believe

The synthesis treats the PLOS modeling as a serious challenge to Harper's catastrophist case but does not resolve the question. The traditional figures rest on Cassius Dio's "2,000 deaths a day" report from Rome in 189 and Galen's clinical descriptions — both of which are vivid but not quantitative-modeling inputs. The PLOS authors note that their models "failed to replicate" the 2,000-deaths-a-day figure under reasonable parameters, suggesting "either historical exaggeration, measurement of a separate epidemic, or unmodeled factors like seasonality."

Two things to flag:

  1. The 2,000/day figure is for the 189 secondary outbreak, not the 165–180 main wave. Conflating them inflates the original-outbreak mortality.
  2. Even a 7% excess-mortality event is non-trivial — that's 5+ million dead over an empire of ~75 million. The disagreement is over whether the plague was structural (initiating decline) or severe-but-recoverable.

Evidence we have

  • For high mortality (traditional): Cassius Dio (189: "up to 2,000 deaths a day in the city of Rome, 25% of those who were affected"); Galen's clinical descriptions; Niebuhr's 19th-c. interpretation; Scheidel (2002, ~25%); Harris-Zelener (~22-24%); Harper (2017, ~20%). All from 2026-05-13-autoresearch-causes-of-the-fall-of-the-western-roman-empire.
  • For low mortality (modeled): PLOS One 2024 — max ~7% for measles, ~5% bubonic-with-rats, ~1% standard smallpox. Authors note "the secondary literature on the Antonine Plague appears to have overestimated the impact of this pandemic." From 2026-05-13-autoresearch-causes-of-the-fall-of-the-western-roman-empire.
  • Methodological pushback: Newfield's BMCR review of Harper notes that mortality minimalists are "buried in endnotes" in Harper's narrative, suggesting the high-end consensus has been overstated in popular treatments.

Evidence we need

  • Direct paleogenetic evidence of the Antonine plague's pathogen (currently unavailable). aDNA from securely-dated 2nd-c. burials would settle the smallpox-vs-measles-vs-other question and let modelers use right parameters.
  • More fine-grained urban/rural mortality data — if cities saw 25%+ deaths while the rural majority saw <5%, both the historians and the modelers could be partially correct.
  • A direct response from Harper to the PLOS modeling — at time of writing this synthesis, Harper had not (in the sources we fetched) published a public response engaging the model on its own terms.

How to resolve

  • Wait for aDNA confirmation of the pathogen + repeat the modeling with the right parameters.
  • Engage the historian–modeler methodological gap directly: which baseline-mortality assumptions are reasonable, and how should ancient testimony be weighted against modeled disease dynamics?
  • Until one side substantially engages the other's tools, both readings stay live.

Related

Referenced by