How to Research Your Family History: A Beginner’s Guide

Work backwards, one generation at a time, and prove each link before you take the next.

The single commonest failure in family history is finding an interesting person with your surname and working forwards, hoping to meet your own line coming the other way. It feels like progress and it is how bad trees get built. Every generation you add must be joined to the one below it by a document naming both.

Before any website, write down everything you already know: names, dates, places, the stories, the half-remembered village. Mark clearly what is known and what is said — family tradition is a lead, not evidence, and mixing the two is how errors become permanent.

Then look for documents at home: certificates, the family Bible with names written in the front, photographs with anything written on the back, letters, naturalisation papers, funeral cards, military discharge papers, deeds, old passports.

Then interview the oldest people in the family, now

This is the one step that has a deadline, and the one people postpone.

Record it, with permission. Ask about places, not just names — which parish, which street, which townland. Ask who their grandparents were and what they did. Ask about the ones who left. Ask about the photographs. Names are on paper somewhere; the reason an uncle stopped being mentioned is not.

Birth, marriage and death records are the spine of a tree, because each one names the previous generation.

Marriage records are the most valuable and the most overlooked: a marriage certificate typically names both fathers, and often gives occupations and residences. One marriage record can advance two lines at once.

Death records are the least reliable of the three — the informant is a grieving relative who may not know the deceased's mother's maiden name — so treat what they say about the previous generation as a lead rather than a fact.

The census is the workhorse, because it puts a whole household on one page.

United States: 1790–1950, released after 72 years, free at the National Archives and on FamilySearch. The 1890 census was almost entirely destroyed by fire and water damage in 1921 — an enormous hole, and the reason many American lines jump from 1880 to 1900.

England and Wales: 1841–1921, released after 100 years. The 1931 census was destroyed by fire in 1942 and the 1941 census was never taken, so the 1939 Register fills the gap and is available.

Ireland: 1901 and 1911 survive complete and are free at the National Archives of Ireland. Most earlier Irish censuses were destroyed.

The technique that matters: trace a family across every census, not just one. Ages drift, spellings change, and a child who appears in 1880 and vanishes in 1890 tells you something. Then look at the neighbours — families migrated together, married each other, and witnessed each other's documents. The household two doors down is frequently a relative.

The hardest single step in most family histories, and the whole problem is this: you need the village, not the country. "From Ireland" or "from Italy" will not open a parish register. A townland or a comune will.

Where the village is usually hiding:

• Naturalisation papers — after 1906 in the United States these are excellent and frequently name the exact birthplace.
• Passenger manifests — post-1893 arrivals record last residence and, often, the relative left behind and the relative being joined. Free at the Ellis Island and Castle Garden databases and on FamilySearch.
• The marriage record in the new country, which sometimes names the parish of origin.
• Church records in the immigrant community — a priest who came from the same region often recorded it.
• The death certificate and the obituary.

And a myth to drop: names were not changed at Ellis Island. Manifests were written at the port of departure and inspectors worked from them. Families changed their own names afterwards, for their own reasons.

Before civil registration, the church kept the records — and they go far deeper than the state ever has.

• England: parish registers from 1538, one of the longest continuous series in the world.
• Ireland: Catholic parish registers, largely from the early 1800s and earlier in the towns, digitised and free at the National Library of Ireland.
• Scotland: Old Parish Registers from 1553, at ScotlandsPeople.
• Catholic Europe: required since the Council of Trent in 1563, so Italian, Spanish, Portuguese, Polish and southern German registers often run continuously from the late sixteenth century.
• Germany and Scandinavia: from the seventeenth century, and Sweden's husförhörslängder — household examination rolls recording everyone in a parish and their movements — are unmatched anywhere.

When a line stalls, these are where it usually restarts:

• Wills and probate — a will names children, in-laws, grandchildren and sometimes exactly how the testator felt about each.
• Land and deed records — the Irish Registry of Deeds from 1708, Scottish sasines, American county deed books. Land changing hands between relatives is a documented relationship.
• Military records — pension files in particular. A US Civil War widow's pension file can contain marriage evidence, family Bible pages and sworn statements from neighbours.
• Newspapers — obituaries, marriage notices, court reports, and the local paper's account of who visited whom. Chronicling America, at the Library of Congress, is free.
• Tax and valuation lists — Griffith's Valuation for Ireland (1847–64) and the Tithe Applotment Books cover the period Ireland's lost censuses do not.
• Cemetery records — and note who is buried beside whom. Plots were bought by families.

• FamilySearch — free, enormous, and the place to start. Run by the Church of Jesus Christ of Latter-day Saints and open to anyone.
• National archives — the Irish, Scottish, Belgian, French departmental, Portuguese and Spanish archives all publish substantial material free. French departmental archives in particular are outstanding.
• Ancestry, MyHeritage, Findmypast — paid, with different strengths. Findmypast for British and Irish, MyHeritage for continental Europe. Many public libraries provide free access.
• Chronicling America, Geneanet, Antenati (Italy), WieWasWie (Netherlands), JewishGen.

Useful, and usually misunderstood. The ethnicity percentages are estimates against reference panels and shift when the panels are updated. The matches are the valuable part — a shared segment is a measurement, and other people's trees are where the surnames come from.

For tracing one surname line specifically, a Y-DNA test on a man of that name is far more targeted than a general autosomal test. Our piece on what DNA and heraldry can and cannot establish goes into this properly.

1. Cite every fact. Where you found it, and when. An uncited tree cannot be checked, including by you in two years.
2. Two sources for anything that matters. One document is a claim.
3. Search spelling variants always. Spelling was not standardised. Müller, Mueller, Miller; Ó Súilleabháin, O'Sullivan, Sullivan; Mac and Mc interchangeably.
4. Record the negative searches. Writing down where you looked and found nothing saves you from looking there again next year.

This belongs at the end of the process, not the beginning, and here is why.

Arms were borne by individuals and inherited by their descendants — not by everyone sharing a surname. So arms are something you look up once you have established which family, in which place, you actually descend from. Doing it the other way round — finding arms for a surname and assuming the connection — is how most inaccurate family heraldry begins.

When you get there, our guide to verifying a coat of arms lists the armorials and registers to search, most of which are free.

The Heritage Volume sets out what the historical record holds for a surname — where it comes from, what it meant, where its bearers were concentrated, and the arms associated with it. It is an account of a name rather than an investigation of your family, and How Your Report Is Made explains exactly which parts come from public records and which our system builds.