What do you do when the records were destroyed?
When records were destroyed, find out exactly what was lost: which series, which years, and which duplicates sat elsewhere. Losses are almost never total. Archives publish their own gap lists, and the copies that survived (state-level duplicates, church registers, tax lists) are your substitute set.
The order of work is always the same. Identify the loss precisely, read the holding archive's own statement about it, list what survives for that place and those years, then search the substitutes. That turns "the records burned" from a reason to stop into a shorter, more specific list of things to search.
What happened to the 1890 US census?
A fire in the Commerce Department basement on 10 January 1921 destroyed most of the 1890 population schedules. The census director reported about 25 percent destroyed outright and half the remainder damaged by water and smoke. Congress authorized disposal of the rest on 21 February 1933, and it happened in 1934 or 1935.
Fragments survive for eleven states plus the District of Columbia, carrying no fewer than 6,160 indexed names, according to Kellee Blake's account of the fire in the National Archives' own history of the 1890 census. Check them before you assume your family is not there.
- State and territorial censuses of 1885: Colorado, Florida, Iowa, Kansas, Minnesota, Nebraska, New Jersey, Rhode Island, and the territories of New Mexico and Dakota all took one, and Michigan took one in 1884.
- The 1890 special schedule of Union veterans and widows, which survives for states from Kentucky (in part) through Wyoming.
- City directories, which were annual and name heads of household with addresses and occupations.
- State and county tax lists, voter registrations, school censuses, and church membership rolls for the same years.
The phrase "the 1890 census burned" hides the useful part, which is that a decade of substitutes exists and several of them name more than the census would have.
Was the courthouse the only copy?
Courthouse fires were common and the response is always the same: find the parallel record. Deeds were often re-recorded by owners afterwards, tax and land records were duplicated at state level, and churches, newspapers, and the parent county kept their own. Virginia's losses are the best documented case.
Virginia counties sent records to Richmond for safekeeping during the Civil War and lost them in the fire of 3 April 1865 during the city's evacuation. Hanover, Gloucester, and James City are among the worst hit, and the Library of Virginia maintains a list of lost-records localities naming what each one lost and when.
- State-level tax records: Virginia's annual land and personal property tax lists survive centrally and often replace a lost census or deed index outright.
- Re-recorded deeds: after a fire, owners brought their originals back to be entered again, so a later deed book can hold a much earlier transaction.
- Chancery causes and other court files, which frequently recite the family relationships that a lost will would have stated.
- Church registers, newspapers, and abstracts published by genealogists who read the records before they burned.
What did the 1922 Dublin fire destroy?
An explosion and fire at the Four Courts on 30 June 1922 destroyed the Public Record Office of Ireland, taking the 19th-century census returns for 1821, 1831, 1841, and 1851, around a thousand Church of Ireland parish registers, and thousands of wills and title deeds.
The 1901 and 1911 returns are the only complete surviving Irish censuses, so Irish research leans on land-based name lists instead: the Tithe Applotment Books (1823 to 1837) and Griffith's Valuation (1847 to 1864), both of which list occupiers by townland. Roman Catholic parish registers were held locally rather than in the record office, so they were not in the building and largely survive.
Treat 1922 as a starting point rather than a conclusion. The Virtual Record Treasury of Ireland has reassembled a substantial amount of the lost material from copies, abstracts, and duplicates held in other archives, and it keeps growing.
Which 20th-century record losses will you run into?
Three losses account for a large share of 20th-century dead ends: the 1931 census of England and Wales, most First World War British Army service records, and the 1973 fire at the US military personnel records centre near St Louis. Each of the three has a documented substitute set.
| Loss | What happened | What to search instead |
|---|---|---|
| 1931 census of England and Wales | Destroyed by fire at a government store in Hayes, Middlesex, on 19 December 1942, unrelated to enemy action. No census was taken in 1941. The Scottish 1931 returns were stored in Edinburgh and survive | The 1939 Register, taken under the National Registration Act 1939, plus the 1921 census either side of the gap |
| First World War British Army service records | Incendiary bombing of the Arnside Street store in September 1940 destroyed roughly two thirds of about 6.5 million files; the survivors are fire-damaged and known as the burnt documents (WO 363) | WO 364 pension records, which were held with the Ministry of Pensions and escaped the fire, plus medal index cards and unit war diaries |
| US military personnel files | The fire of 12 July 1973 at the National Personnel Records Center destroyed 16 to 18 million files: about 80 percent of Army records for discharges between 1 November 1912 and 1 January 1960, and 75 percent of Air Force records for discharges between 25 September 1947 and 1 January 1964 for surnames from Hubbard, James E. onward | Auxiliary records the archives uses to reconstruct service: final pay vouchers, VA claim files, state adjutant general records, and unit morning reports |
Both national archives publish what they hold and how they reconstruct from it: the US National Archives on the 1973 fire, and researchers of the British records on the Arnside Street fire. Read the archive's own account before you accept a secondhand claim about what was lost.
How do you prove a fact when the record is gone?
A destroyed series removes one route to a fact, not the fact itself. Name the fact you need, list every record type that would have captured it, then check which of those survives for that place and year. Independent survivors that agree can prove what the lost original would have said.
This is the same method that works on any stalled line, applied to a loss you can name. The general version is in our brick walls guide, and the fact most often lost to a fire has its own ladder of substitutes in can't find a birth record.
Searching several substitute sets for one person is slow work, and it is what AncestorIQ's Deep Research does in one pass: it searches record sets across multiple archives, keeps the citation for every fact, and tells you which sources it rejected and why. Old terminology in whatever it finds is covered in our records glossary.
