+++
title = "Three Fields on the Form"
description = "A battle total can mix killed, wounded, missing, civilian, later and estimated losses under one number. The record is a category system before it is an estimate."
date = 2026-08-19
draft = false

[taxonomies]
topics = ["historical-events", "battles"]
kinds = ["attestation", "convention"]

[extra]
tier = "public"
schema_type = "Article"
canonical = ""
hero = "articles/three-fields-on-the-form/hero.webp"
hero_alt = "Nineteenth-century casualty return beside individual casualty-report slips"
diagram = "diagrams/three-fields-on-the-form.svg"
podcast = ""
mechanism = "Historical-event records track contested sequence and source status, while battle records specify forces, tactical course and outcome; the contact is that a headline total combines categories whose inclusion rules are usually more consequential than the arithmetic."
discriminating_test = "Request the inclusion table and reconcile each component to a dated source; irreconcilable category changes make the headline a convention collision."
false_friend = "A revised estimate of the same category also changes the total, but preserves the inclusion rule and identifies new evidence or a revised estimator."
symptoms = [
  "A loss figure changes across research notes with no stated definition",
  "An exposure total mixes gross, net and contingent positions",
  "A performance report compares realized losses to provisions as though they were one measure",
  "A battle casualty estimate does not state the side, category, period or source behind it",
  "Forces engaged are compared with total forces available as though they were the same field",
]
sources_verified = false

[[extra.faq]]
q = "What does a battle casualty total actually include?"
a = "It includes only the categories its source chose to count. Gettysburg's Union return records 3,155 killed, 14,531 wounded, and 5,369 captured or missing as separate fields. A total made from those fields describes Union losses over the return's period; it does not automatically describe Confederate losses, civilian deaths, later deaths, or every consequence of the battle."

[[extra.faq]]
q = "Why can two histories give different casualty figures for the same battle?"
a = "They can be using different sources, categories, or estimators. For Cannae in 216 BC, Polybius gives about 70,000 Roman killed and 10,000 captured, while Livy gives 45,500 killed. Modern military historians have proposed 10,500–16,000 Roman losses. Those figures cannot be ranked as bare totals before the counted category and the source's basis are stated."

[[extra.faq]]
q = "Can archaeology settle a disputed casualty figure?"
a = "Usually not. At Kalkriese, coins of Augustus with none minted later and Roman lead sling bolts establish dating and military presence for the Teutoburg Forest evidence; they do not count Roman dead. Archaeological finds can correct a battle's location or constrain its course, while casualty totals still depend on returns, chronicles, administrative records, or later estimation."

[[extra.faq]]
q = "When is a changed loss estimate not a category problem?"
a = "It is not a category problem when the inclusion rule remains fixed and the revision identifies new evidence or a revised estimator. A convention collision occurs when one figure includes killed and missing while another includes later deaths, civilians, or projected consequences. Both changes move a headline total, but only the second changes the thing that total is supposed to measure."

[[extra.faq]]
q = "How should an exposure or performance total be compared with another one?"
a = "Each series needs an inclusion schema, a source vintage, and a reconciliation bridge before comparison. Gross, net, and contingent exposure are different categories. Realized losses and provisions are different measures. A comparison that does not preserve those distinctions reports a difference in reporting convention as though it were a difference in risk or performance."

[[extra.sources]]
title = "\"1883 eruption of Krakatoa,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/1883_eruption_of_Krakatoa"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"1906 San Francisco earthquake,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/1906_San_Francisco_earthquake"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"1918 influenza pandemic,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/1918_influenza_pandemic"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of Agincourt,\" Anne Curry's administrative-record revision and force estimates"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_Agincourt"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of Bosworth Field,\" 2009 survey and revised site location"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_Bosworth_Field"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of Camlann,\" Annales Cambriae entry and historicity debate"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_Camlann"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of Cannae,\" Polybius and Livy casualty figures"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_Cannae"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of Gettysburg,\" recorded Union and Confederate casualty figures"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_Gettysburg"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of Hastings,\" contemporary and modern force estimates"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_Hastings"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of Stalingrad,\" dates and duration"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_Stalingrad"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of the Somme,\" dates, casualties, and phase structure"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_the_Somme"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Battle of the Teutoburg Forest,\" Kalkriese excavation finds and dating"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Battle_of_the_Teutoburg_Forest"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Black Death,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Black_Death"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Chernobyl disaster,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Chernobyl_disaster"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Great Famine (Ireland),\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Great_Famine_(Ireland"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Halifax Explosion,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Halifax_Explosion"
kind = "tertiary"
fetched = "2026-08-13"

[[extra.sources]]
title = "\"Tunguska event,\" Wikipedia"
publisher = "Wikimedia Foundation"
url = "https://en.wikipedia.org/wiki/Tunguska_event"
kind = "tertiary"
fetched = "2026-08-13"
+++

From 1 to 3 July 1863 the Union return for Gettysburg recorded 3,155 killed, 14,531 wounded, and 5,369 captured or missing.

Those are not three ways of saying the same thing. They are three fields on a form. They cover one side, over a bounded period, and the last field is already two categories sharing a box.

A reader who turns them into a Gettysburg death toll has altered the record before doing any arithmetic.

The precision belongs to the return system rather than to the battle. By 1863 armies filed regimental casualty returns. Ancient and early medieval armies did not, and the tempting conclusion is that a modern battle has a number while an old one has an estimate.

The more useful conclusion is that both start with a decision about which kind of loss was going to be counted at all.

## One name, several measures

Cannae was fought in 216 BC. Polybius gives about 70,000 Roman dead and 10,000 captured. Livy, writing generations later, gives 45,500 dead. Neither was there, and modern military historians have argued Roman losses down to something nearer 15,000. The received version of that spread is that one of them found the right number and the others did not. The evidence does not permit the arrangement. Polybius supplies a count of killed and a count of captured. Livy supplies a different killed count. The modern figures are reconstructions from a third kind of reasoning entirely.

Two words, Cannae casualties, collapse a named historian, a side, a category, a period and an estimator into one quantity.

That is not an ancient peculiarity. It is what a good modern record makes visible. Gettysburg's three Union fields state on paper the distinctions that Cannae's narratives leave implicit. The later record is better because its components can be inspected apart, not because its total is more impressive.

## Two death tolls, two questions

Chernobyl shows the same structure off the battlefield.

The immediate count is 2 people killed by debris and 28 who died of acute radiation syndrome within three months. The long-term estimates run to a UN figure of up to 4,000 eventual cancer deaths in Ukraine, Belarus and Russia, and a wider European figure of 16,000.

The 28 were diagnosed individually. The thousands are excess cancer across an exposed population, and they exist only inside a dose-response model. The two are not repeated attempts at one hidden total.

They differ on consequence, on time horizon, on population, and on method. What they share is the phrase death toll, which is doing all the damage. The label makes the count look like one quantity whose uncertainty could be settled by preferring a better source. Under the label, the inclusion rule has changed. Observed deaths and projected excess cancer cannot be reconciled until their categories are put back on the page.

## What each instrument can settle

Battle evidence comes from several instruments, and none of them speaks to every field. A chronicle can describe a force or a loss. A muster roll or purchase order is an administrative record kept for logistics rather than for persuasion. A casualty return is a post-battle accounting. An excavation can establish place, date, terrain and presence. Kalkriese is the boundary worth keeping intact. Tony Clunn's finds there in 1987 were Augustan coins, none minted later, and Roman lead sling bolts; excavation under Wolfgang Schlüter and then Susanne Wilbers-Rost produced further coin, weapon and skeletal material. The site supports Kalkriese as the final stand of the Teutoburg Forest disaster in AD 9.

It produces no casualty total, and no amount of further digging will produce one.

That is a different failure from disagreement. The instrument is sound, the finding is real, and the question was addressed to the wrong instrument.

## The table comes before the total

The working record is a table with a row per component: side, period, category, source, and whether the figure was observed or estimated. The total is the last row, and it travels with its inclusion rule attached. That separates a revision from a collision. A figure can move because new evidence changed an estimate of the same category, and then the source or the estimator that changed should be nameable. Or it can move because one account counts the missing and another does not, or because one switched from bodies to projections.

The second is not disagreement inside a shared measure. It is two measures wearing one name.

Financial loss records have an advantage over Cannae, in that the primary rolls may still exist. They have an obstacle Cannae does not, in that access restrictions can leave an existing ledger merely attested.

Gross, net and contingent exposure. Realised losses and provisions. A reported figure and the components under it. None of those are synonyms waiting to be summed. They are inclusion decisions waiting to be inspected.

Polybius and Livy disagree about Cannae, and the disagreement is the most the surviving record can support.

Gettysburg's clerk disagreed with nobody. He filled in three fields because three fields were on the form.

The form is the finding. Somebody decided in advance what would be countable, and every total downstream is a report on that decision rather than on the battle.