RUGet in touch
WebsitesSystemsWorkNotesAbout
All work

A family tree to 1715: seven generations from confession lists and tax censuses

A family tree for another family, from one rural parish. In two days of paid archive access AI read confession lists, tax censuses and registers, and the line went up seven generations — to a man born in 1715. The tree is kept apart and never mixed with others.

Context

The client was another family. The starting point was relatives’ notes with the names of a few generations and the name of a village. The task: check the notes against documents and go as far up as the archives allow.

Before the mid-19th century there are hardly any registers, and those that exist often leave out rural people’s surnames. The main sources became confession lists — yearly lists of parishioners by household, with everyone’s age — and tax censuses. Access to scans was paid by the day, so the work was planned in advance: first a list of files from the free catalogue, then the access.

What was built

A line built from ages. A household is found in one year’s confession list, then the same household in lists and censuses of other years: 1792, 1795, 1805, 1808, 1811, 1840, 1850, 1858, 1877, 1883. The age in each document gives the year of birth within ±1. A match across several documents confirms these are the same people. The 1850 census gave a date of death outright: “died in 1837”. In the tree such years are marked “about”: of 38 known birth years, 22 are calculated from ages.

Registers where they exist. From the mid-19th century the line is confirmed by marriage, birth and death records, down to five births in 1893–1906. The maiden name of one wife comes from a marriage record. Another is inferred from several documents taken together and marked as probable.

Searching before 1792. The agent went through the parish title pages of about 15 volumes of 18th-century confession lists — around 1,600 scans. None of them has a section for the village. That is a result too: it is now known where to look next, and that the next step is a written request to the archive, not browsing online.

A separate tree. This is another family, and it is never mixed with the tree from the first case. In the same Gramps Web database its people carry a tag, and the tree is exported as its own GEDCOM. The client received the tree, an HTML page with cards, an archive of scans and 21 composites. A composite is one image per find: the file cover, the index, the leaf with the entry, an arrow showing where the entry is, and a source caption.

Scale: 39 people, 12 families, 54 events, 7 generations, earliest year of birth 1715. The main work fitted into two days of archive access.

What it looks like

A household from a 1792 confession list: each person’s age becomes a year of birth. Scan blurred, names hidden
A household from a 1792 confession list: each person’s age becomes a year of birth. Scan blurred, names hidden, mobile version
A household from a 1792 confession list: each person’s age becomes a year of birth. Scan blurred, names hidden
The second tree in numbers: seven generations, earliest year of birth 1715. Years before the mid-19th century are calculated from ages
The second tree in numbers: seven generations, earliest year of birth 1715. Years before the mid-19th century are calculated from ages, mobile version
The second tree in numbers: seven generations, earliest year of birth 1715. Years before the mid-19th century are calculated from ages

Why this way and not another

Age as a date, not a guess. We could have stopped at registers and at the mid-19th century. Confession lists give an age, not a date, and every birth year from them is ±1. But they come almost every year, and one person shows up in a dozen documents over half a century. That is enough to identify someone without a surname and go three generations higher. The cost: those dates stand in the tree as “about”, not as exact dates.

Catalogue first, payment second. Access is paid by the day, so the files were chosen in advance from the free catalogue, and the paid days went to reading only. The cost: years that are not digitised (marriages for some years, deaths for others) are left to order. Registers bound out of order or surviving only as fragments were read by the agent in full, year by year.

How to search handwritten registers and where AI goes wrong is in the note “Finding ancestors in handwritten registers”. The first tree and the infrastructure behind both are in the case “A family history from handwritten archives”.

Other case studies

Case studies and personal projects

An international parcel shipping website: a manager assembles any page without a developerThirty-four ready blocks and more than thirty country pages. A manager assembles any page from them and no developer is needed.A package forwarding website: a catalogue of a thousand products and a store directoryMore than a thousand products, categories, a store directory and a button that copies the American address — on the website for forwarding purchases from American stores.A corporate website: forty-four blocks, ten form scenarios and news from LinkedInForty-four ready blocks, ten form scenarios, and company posts from LinkedIn arrive in the news on their own. A manager assembles any page without a developer.