Field guide
How to Read a War: A Newcomer's Guide to Conflict Literacy
How to Read a War: A Newcomer's Guide to Conflict Literacy
Edition: 23 September 2026. Built from the 120-dossier Wars of the World research corpus — every figure below is attributed to the dossier that carries it, and ranges stay ranges.
Start here
Five things to know before you trust any number about any war.
1. Almost every death toll is a range, not a number. The American Civil War's toll is given as 650,000–850,000 (a census-based band); the Thirty Years' War as 4.5–8 million (a modern estimate band); World War I as 15–22 million depending on what you include. When you see a single clean number, someone rounded the argument away.
2. "War" and "armed conflict" are not interchangeable. Trackers like UCDP (the Uppsala Conflict Data Program) classify armed conflicts by intensity thresholds — typically 25 battle-related deaths a year to count at all, 1,000 for "war." So a "conflict" can be real, documented, and killing people without being called a war.
3. Every number has a claimant. Gaza's post-October-2025 toll (1,381 killed, 4,757 injured per the OCHA flash update of 18 September 2026) is relayed from Gaza's Ministry of Health, which is administered by Hamas and disputed by Israel. Ukraine's civilian toll (16,874 killed through July 2026) is verified case-by-case by the UN's human-rights mission, which says the real figure is higher. Ask "who counted, and how" before "how many."
4. Military and civilian deaths come from different machinery. Military deaths are routinely contested on both sides (Ukraine's are explicitly UNVERIFIED — no adopted figure). Civilian deaths are sometimes verified case-by-case (Ukraine), sometimes compiled from ministries (Gaza), sometimes estimated from censuses decades later (the Taiping Rebellion's 20–30 million is a speculative census-based band).
5. "Ceasefire announced" and "war over" are different events. A ceasefire was announced in Gaza on 10 October 2025; fighting continued and OCHA kept recording fatalities for eleven months after. Announced endings are claims, not facts.
The jargon decoder
- UCDP — Uppsala Conflict Data Program, the academic gold standard for coding which conflicts are active and how many battle deaths they produce. You'll hear this when someone says a war is "UCDP-listed."
- ACLED — Armed Conflict Location & Event Data project. Event-level data: battles, strikes, civilian targeting, protest events. When a dispatch says "per ACLED," it means someone scraped and coded news/event reports, not that a body was counted twice.
- IISS Armed Conflict Database — the International Institute for Strategic Studies' tracker of current conflicts. A war "covered by IISS" means it's on the recognized active list.
- OCHA / OHCHR — UN humanitarian coordination / UN human-rights office. OCHA publishes the Gaza fatality flash updates; OHCHR's Ukraine mission (HRMMU) verifies civilian casualties one case at a time.
- Battle deaths — people killed in combat operations, distinct from total war deaths (which can include disease, famine, and displacement mortality).
- Tier 1 / 2 / 3 — the validation grade used in this corpus: Tier 1 means multiple independent sources corroborate; Tier 2 means strong institutional corroboration with contested points; Tier 3 means thin or single-source evidence. A Tier 3 dossier is not wrong — it is honestly flagged as unverifiable in parts.
- UNVERIFIED — a claim that surfaced but could not be corroborated in the research pass. Not false; not adopted. The most honest word in the whole corpus.
Five doors in: where to start reading
These are entry points, not rankings — five conflicts chosen because the evidence is strong, the stakes are legible, and each teaches a different lesson.
1. The Russo-Ukrainian War (2014–present) — to follow a war live. Tier 1. This is the best-documented current war in the corpus: UN-verified civilian data (16,874 killed, 51,273 injured through July 2026), ACLED event data, UCDP coding. The lesson it teaches is the split between verified civilian harm and unverifiable military losses — and how drones went from battlefield novelty to, in HRMMU's description, "hunting" frontline civilians. Start here if you want to understand how a war is tracked in real time.
2. World War I (1914–1918) — to see how historians argue. Tier 1 on the framework, Tier 2–3 on the totals. The causes alone are a century-long seminar: Fritz Fischer's 1961 German-expansionism thesis versus the multi-causal structural accounts (Christopher Clark's The Sleepwalkers, 2012, argues Europe "sleepwalked" in). The 15–22 million death range teaches inclusion criteria — whether influenza, post-war conflicts, and the Armenian genocide count changes the total by millions.
3. The Gothic Wars and the fall of Western Rome (376–476) — to meet the debate at its oldest. Tier 1 on events. No aggregate death toll exists for the entire century — a useful shock for anyone who expects every war to come with a body count. The causes are the most debated question in late-antique historiography: Peter Heather's external-shock thesis (the Hunnic push, the loss of Africa's tax base in 439) versus internal fragmentation (Guy Halsall) versus the "transformation, not fall" school (Walter Goffart and the Vienna school), answered by the catastrophists (Bryan Ward-Perkins). Start here to learn that history's biggest arguments are about causes, not events.
4. The Peloponnesian War (431–404 BCE) — to read the first war reporter. Tier 1. Thucydides is the reason we have the concept of a war narrative at all. The corpus flags even here: no aggregate death toll, the plague's death toll is a 75,000–100,000 scholarly spread, and the plague's pathogen is unknown. Start here if you want the origin story of military history — and a reminder that the oldest sources are also the least countable.
5. The Gaza War (2023–present) — to watch attribution discipline in action. Tier 2 (downgraded from Tier 1 because no independent tracker publishes a verified consolidated fatality total). Every figure in this dossier is attributed and none is settled: Gaza MoH numbers cited by the UN and disputed by Israel, ~1,200 killed on 7 October 2023 (Israeli government), the October 2025 ceasefire that didn't stop the killing. Start here to practice the one skill this whole guide exists to teach: reading the attribution before the number.
Mistakes everyone makes once
Treating a point estimate as a count. "The Thirty Years' War killed 8 million" sounds precise. The research says 4.5–8 million is a modern estimate band, and regional population-loss percentages conflate death, flight, and fertility collapse. Quote the band or don't quote it.
Reading ancient sources as spreadsheets. Herodotus gives 1.7–5.2 million Persians at Thermopylae and Salamis; modern scholars unanimously reject those figures. Sima Qian's 450,000 Zhao dead at Changping rests on his word alone. Ancient authors wrote literature, propaganda, and state ritual — not casualty ledgers. (The "buried alive" detail at Changping is later embroidery; Sima Qian says executed.)
Confusing displacement with death. Sudan has 14+ million displaced (UN, 2025–2026) — the world's largest displacement crisis. That is not 14 million dead. Displacement and death are counted by different agencies with different methods; collapsing them inflates one and erases the other.
Trusting one belligerent's math. Caesar's own boast — 1,000,000 Gauls killed, 1,000,000 enslaved, via Plutarch — is the ancient equivalent of a press release. Ashoka's Rock Edict XIII (100,000 killed, 150,000 deported at Kalinga) is the victor's own published account; unusually frank, still one claimant. Modern versions of this mistake: adopting either side's military death claims in Ukraine, which the dossier leaves UNVERIFIED.
Assuming announced endings ended things. Beyond Gaza: the Iran–Israel twelve-day war's June 2025 ceasefire left "uneasy deterrence"; the PKK's 2025 ceasefire declaration (1984–2025 armed phase) is a claim whose implementation sits outside this corpus. A ceasefire is an event in the war, not necessarily the end of it.
Where to go next
The trackers named across this corpus's current-conflict dossiers, each good for something different:
- UCDP (Uppsala) — which conflicts are active, how many battle deaths. Start here to see the shape of the world's wars.
- ACLED — events: battles, strikes, civilian targeting. Best for "what happened this month" in places like the Sahel or Myanmar.
- IISS Armed Conflict Database — the strategist's overview of major current conflicts.
- UN OCHA — humanitarian snapshots (its Gaza flash updates are the primary casualty source for that war).
- UN OHCHR / HRMMU — verified civilian casualties for Ukraine, case by case; the closest thing to a count anyone has.
One rule of etiquette: cite the range, name the counter, keep UNVERIFIED visible. The researchers who built this corpus graded 71 dossiers Tier 1, 42 Tier 2, and 7 Tier 3 — and labeled every unverifiable point instead of smoothing it. Do the same when you talk about war, and you'll already be ahead of most of the internet.