In 1086 the Conqueror's commissioners counted England's households, village by village. I set every place they named against the 2021 census to see whether size then still predicts size now. Mostly, it does not.
Domesday Book is the most famous count in English history. In 1086 the king's commissioners recorded, manor by manor, the villagers, smallholders and slaves who worked the land, the plough teams and mills, and what each estate was worth. It named 14,334 places. Most of them are still on the map.
That makes it a rare natural experiment. If size in 1086 still shaped England today, the big villages of Domesday would be the big towns of now, and the small ones would still be small. I wanted to test that, place by place, across 935 years.
The short answer is that Domesday is not destiny. The link between a parish's households in 1086 and in 2021 is positive, but weak. A few great places of 1086 now have only a handful of households, and some places Domesday barely noticed are now towns of tens of thousands of homes.
Each dot below is a place named in Domesday Book, sized by the households recorded there in 1086. The colour shows how much its modern parish has grown since. Warm colours grew less than the typical parish, which has about 10 times as many households now as in 1086. The darkest red ones have fewer households now than then. Cool colours grew more.
Switch to Colour by 1086 size to see the map of Domesday on its own. Search for a place, or tap any dot for its full record and a link to Open Domesday.
Many modern parishes hold more than one Domesday place, so the fair test is at parish level: add up the households of every Domesday place inside a parish, and set that against the parish's households in the 2021 census. That gives 6,376 parishes with a figure at both dates.
The rank correlation between the two is 0.22. On log scales, size in 1086 explains about 4% of the variation in size today. The link is real and it points the right way: bigger then means bigger now, on average. But it is faint.
The slope of the fitted line is 0.31. In plain terms, a parish with ten times as many households as another in 1086 has, on average, only about 2.0 times as many today. Of the 647 parishes in the biggest tenth in 1086, 139 (22%) are in the biggest tenth now. Chance alone would put about one in ten there. Meanwhile 60 of the 743 smallest tenth have climbed into the top tenth.
Almost everywhere grew. Only 214 parishes (3.4%) have fewer households now than Domesday recorded. The middle half grew between 3.9 and 33 times. And the small grew fastest.
The smallest fifth of parishes grew by a median of 36 times; the biggest fifth by 5.7 times. Some of that is what noisy counts alone would produce: a place Domesday undercounted looks small in 1086 and then seems to grow fast. So I would not lean on the exact numbers. But the overall picture does not depend on them.
I checked the result several ways, and it barely moves. Without Yorkshire, the Spearman correlation is 0.22. Using only parishes where every Domesday place has a household figure, it is 0.23. Using only parishes with at least five households in 1086, 0.23. Against 2021 residents rather than households, 0.22. And for the 3,862 Domesday places that sit alone in their modern parish, so no adding up is needed, it is 0.21.
Bigger then means bigger now, on average. But knowing a parish's size in 1086 tells you very little about its size today.
Of the 4,058 parishes with at least 20 households in 1086, 185 have fewer households today. They cluster in the east: 61 are in Lincolnshire, 23 in Suffolk and 21 in Norfolk.
Waddingworth tops the list. A caution: its 60 households come from two entries (Lincolnshire 3,12 (20); Lincolnshire 29,6 (40)) that each describe a holding of six carucates there, one of them disputed between two lords. Some of the same people may be counted twice. Even halved, the fall is steep.
At the other end are the places Domesday barely noticed. Here the modern parish carries a modern name, and the Domesday place inside it may be a hamlet now swallowed by a town. That is part of the story: much of today's population lives where Domesday recorded very few households.
Two gaps limit this list, and both hide some of the biggest risers. First, the census figures in the source exist only for civil parishes. Big towns and cities that are not civil parishes (Norwich, Canterbury, Derby, Nottingham and others, 185 such areas in all) have no 2021 figure, so they drop out of every comparison here. They hold 9% of the 1086 households. Second, 1,077 parishes contain Domesday places with no household figure at all. That is not the same as none. They include Bury St Edmunds, Dunstable and Kendal, where the entries do not give a count the data can sum.
Persistence is not the same everywhere. Within Rutland, the rank correlation between parish size in 1086 and 2021 is 0.57; within Sussex it is 0.01, which is no link at all. Of the 33 counties with at least 20 parishes, 6 reach 0.4 or more and 8 fall below 0.2.
Zoom out further and the pattern of 1086 is clearer. Add up each county's parishes and the correlation between county totals in 1086 and 2021 is 0.61 on log scales (Spearman 0.70, 33 counties). Which county a parish is in also matters about as much as its own size: in a simple model, county alone explains 5.8% of the variation in parish households today, households in 1086 alone explain 5.3%, and the two together 12.9%.
The data flags 2,581 of the 13,068 located Domesday places (20%) as possibly lost: no settlement in Wikidata and no modern place of the same name within 2 km. These are candidates, not findings. A grid error, a renamed place, a surviving farm or a quarter of a town can look lost.
The possibly lost places were small. Their median was 7.8 households in 1086, against 19.0 for the rest, and together they held 8% of all the households Domesday recorded. They are most common in Cornwall (52%) and Devon (38%), and rarest in Huntingdonshire (3.5%). A name that now survives only on a farm, or that has changed, is hard to match, so read the county pattern with care.
Domesday recorded much more than people. Perhaps the richest places of 1086, or the ones with mills and plenty of plough teams, were the ones set up to grow. I tested that with a regression of 2021 households (log scale) on 1086 households, plus the value of the estate per household, plough teams per household, and whether a mill was recorded, across 5,799 parishes with all of these figures.
They add almost nothing. Households alone explain 5.3% of the variation; adding wealth, ploughs and mills takes it to 6.3%. Each has a small positive coefficient. The 1086 value of a parish ranks with its 2021 size about as well as its households do (Spearman 0.23, against 0.24). Parishes with a mill grew by a median of 8.4 times, those without by 9.4.
One modern fact beats all of them: the area of today's parish. Adding it lifts the explained share to 19%, and the coefficient on 1086 households falls from 0.38 to 0.13. One reading is that part of what looks like persistence is land: a large parish had room for many households then and has room for many now.
Domesday does not count people. It lists heads of households by status: villagers, smallholders, cottagers, slaves, freemen, burgesses and others. Their wives, children and dependants are not there. The "households" figure here is the sum of those heads, as the source data builds it. That is why I compare it with households in 2021, not with population. Turning Domesday's heads into a head count needs a multiplier, and historians disagree about the right one.
The survey is also uneven. London and Winchester are not in it. The far north was largely not surveyed, and Yorkshire's record is thin: only 648 of its 2,148 places have a household figure. The figures for Essex, Norfolk and Suffolk come from Little Domesday, and here they are parsed from the translated text rather than from a coded database. An absent figure means the entry does not state it, which is not the same as zero.
Data. The realm-domesday repository joins the Hull Domesday Project's coded Domesday to the modern map: 14,334 places (13,068 located) and their entries, each place linked to the 2021 civil parish containing the centre of its 1 km grid square. The repository reports that this parish agrees with Wikidata's 90% of the time; a place near a boundary can land in its neighbour.
Counting households. I sum the households of every entry that describes a place. When one entry describes several places, I share its households equally among them, as the repository's own then-and-now views do. 2,522 places take a share of a multi-place entry. Places an entry only names as dependencies get nothing from it.
Parishes. For each modern parish I add up the households of its Domesday places and compare them with Census 2021 households (ONS). The main test uses the 6,376 parishes with figures at both dates. Each parish is given the Domesday county of most of its places.
Statistics. Pearson correlation and least-squares slope on log10 households, and Spearman rank correlation on the raw figures. The wealth model is ordinary least squares with robust standard errors. Growth ratios use at least 10 households in 1086 for risers and 20 for fallers, to keep tiny numbers from dominating; the "from almost nothing" list takes parishes with 5 households or fewer.
Known problems. Big unparished towns and cities have no 2021 figure in the source and are left out. Some entries carry totals for a group of places while describing only one; 289 places draw households from entries with an "altogether" or "in total" clause, and dropping the parishes involved leaves the correlation almost unchanged (Spearman 0.22). Domesday figures were parsed from Hull's coding and agree with Open Domesday's hand-coded figures for 98.7% of manors on villagers and 98.9% on smallholders, in the repository's check of 450 manors. "Possibly lost" is a candidate list, never a finding.