England is scattered with the ghosts of villages: house platforms under pasture, a church alone in a field, a name on an old tax roll. I mapped all 2,222 in Beresford's Lost Villages and set them against Domesday Book to ask where they were, how big they were, and when and why they emptied.
Walk across a sheep pasture in Northamptonshire or the Yorkshire Wolds on a low winter afternoon and the ground starts to talk. A sunken lane runs between grassy banks. Rectangles of raised turf mark where houses stood. Ridge and furrow, the ploughland of a vanished open field, runs up to the edge of it all. These are deserted medieval villages, and England has thousands of them.
In 1954 the historian Maurice Beresford brought them to national attention, and with the archaeologist John Hurst he compiled the first county-by-county gazetteer. The University of Hull has turned that 1968 list into an open database, Beresford's Lost Villages, with a location for every site and, for some counties, its full documentary record: Domesday Book, the medieval tax rolls, the census, and a written account of what is known.
The popular story is simple: the Black Death came in 1348, the villagers died, the village was gone. The textbook answer is that lords later turned the fields over to sheep. I wanted to see what the record itself says, site by site, and to ask a question that needs a second dataset: were the doomed villages already different in 1086?
Each dot is a site. Colour shows the presumed date of depopulation where the record gives one; grey dots have none, because the full written record has so far been completed for only 14 counties. Filter by date or type, search for a place, or tap any dot for its figures and a snippet of the record's own description.
The gazetteer is densest in Yorkshire (374 sites), Lincolnshire (220) and Northumberland (165). But raw counts mix two things: how many villages a county had, and how hard people have looked. A fairer measure asks what share of each county's Domesday places later became lost villages.
On that measure the pattern is stark. In Lincolnshire, 20% of the places named in Domesday Book are now in the gazetteer as deserted; in Hampshire and Hertfordshire, nearly as many. In Cornwall, Devon and Shropshire, almost none are. Of 31 Domesday counties, 13 lost one place in ten or more, and 8 lost fewer than one in thirty.
This is the old map of English farming. The losses run through the champion country of nucleated villages and great open fields, from Lincolnshire and the Midlands down to Hampshire and Berkshire. In the west and south-west, where people lived in scattered hamlets and farms, there were fewer villages to lose and fewer that the 1968 searchers recognised. I cannot fully separate those two causes, and the Hull team itself expects the list to grow.
Domesday Book lets us look at these villages long before they emptied. I matched each lost site to the Domesday place of the same name within 5 km, using the Hull Domesday Project data as prepared for my Is Domesday Destiny? study. 1,271 sites found a match. Leaving aside shrunken, shifted and doubtful sites, and matches where a "Little" or "North" site met a plain Domesday name, that gives 1,110 Domesday places later deserted, against 11,819 that were not.
The doomed were smaller, though not tiny. The median deserted place had 13 households in 1086 (heads of household, not people), against 17 for the places that survived. Because the deserted places cluster in particular counties, the fairer test is within each county: ranked against its own county's Domesday places, the typical doomed village sat at the 33rd percentile, and 72% were below the county's middle place.
The risk peaked in the middle. 11% of places with 6 to 20 households in 1086 later appear as deserted, against 2.5% of places with more than 40. Big villages almost never vanished. The very smallest places, with 5 households or fewer, were not the most at risk either; many of them were outlying farms or manors that were never villages, and their loss would rarely leave earthworks that someone in 1968 would list.
The doomed were poorer in the other measures too. The median deserted place had 4.0 plough teams against 5.5, and was valued at £2 3s against £3. But per household they ploughed about the same (0.29 teams against 0.30). They were not worse farms; they were smaller communities, with less to hold on with when hard times came.
The villages that vanished were not marginal places. They were ordinary, smallish villages in good farming country.
A classic explanation for desertion is that villages pushed onto poor, high or wet land in the land-hungry thirteenth century were the first to go. I measured the ground under every Domesday place: height, local relief within a kilometre, distance to the nearest watercourse and to the nearest Roman road.
The raw medians differ a little: deserted places sat on gentler ground and slightly further from water. But within each county almost all of that disappears; on every measure the typical deserted place ranks between the 50th and the 55th percentile. The differences reflect where the lost villages are, in the lowland champion belt, not where in their own landscape they were set. The median site in the whole gazetteer lies at 70 m above sea level, and only 82 lie above 200 m.
For many sites the gazetteer copies each appearance in the national tax records. They let us watch villages thin out. The figures are not like for like: the 1377 poll tax counted nearly every adult over fourteen, while the 1524 and 1525 subsidies missed the poorest. Still, the drop is telling.
Of the 147 villages with a count in both years, the median fell from 46 poll-tax payers to 7 subsidy payers. Below are the biggest of them in 1377, with their Domesday figure (the minimum number of people the gazetteer counts there) and the 1801 census.
The name usually outlived the village. The National Archives' E179 database of tax records gives a last document year for 1,622 of these places, and for 93% of them the last document is after 1600. The median last appearance is 1678. A township went on paying tax as a unit long after its houses had gone, which is one reason desertion is so hard to date from the documents alone.
For 308 sites, all in the 14 fully recorded counties, the record gives a presumed date of depopulation. Most are ranges ("1350-1450", "15th-16th century"), with a median width of 100 years, so I spread each site's weight evenly across its range and added them up.
The Black Death does not dominate. Only 20% of the dated weight falls in the century from 1348 to 1449, and 6% before it. In all, 74% falls after 1450. More falls in the century from 1450 to 1549, the age of the enclosure complaints (26%), and the busiest single decade is the 1500s. A good deal comes later still: 28% between 1550 and 1699, and 20% from 1700 onwards, when villages were cleared for landscape parks and enclosed by Act of Parliament.
That fits what historians have argued since Beresford: the plague and the long fall in population afterwards weakened many villages, but the final blow usually came generations later, when a lord or a grazier turned the open fields into sheep pasture. Wharram Percy, the most famous lost village of all, tells the story in miniature.
For 542 sites the gazetteer has a written account. I searched those accounts for the main stories of desertion. A mention is not a verdict: an account may raise enclosure only to doubt it. But the counts show what the evidence keeps returning to.
Enclosure and conversion to pasture (68 sites) and emparking (63) come first, with monastic landholders (53) close behind; some monastic houses, the Cistercians above all, cleared villages to make granges. The Black Death or plague is mentioned for 38, and the sea or flooding for 11, including the lost port of Ravenserod at the mouth of the Humber.
The gazetteer is the 1968 list, digitised. It is not every lost village: the Hull team expects the sites identified since 1968 to at least double the total, and plans to add them. Coverage reflects where people searched as well as where villages died. The full documentary record exists only for 14 counties (540 sites), so the dates, descriptions and scheduled status describe that sample, which leans to the south and east and to the East Riding of Yorkshire. Presumed dates are the record's judgement, often broad. Domesday counts heads of household on manors, not people in villages, and a lost village's households may sit inside a neighbour's entry. Name matching can err.
The gazetteer. I read all 2,222 records from the Village Explorer feature service of Beresford's Lost Villages (University of Hull). The service gives each site three positions (a grid reference, grid coordinates and latitude and longitude). For 25 sites they disagree; I kept the position that a second field confirms to within 1 km, preferring the grid reference. Site types: 2,000 deserted villages, 74 deserted hamlets, 52 shrunken, 13 shifted and 13 migrated villages, and 70 doubtful sites. The database's own ID number is not unique (some numbers are shared by sites in two counties), so I key sites by the service's row id.
Cleaning the figures. Each documentary field is free text. A bare number is used as a value. A number with a note in brackets ("896 (Holme with)") usually means a figure shared with other places, so it counts as recorded but is not used as a size. "Not recorded" means the source survives but the place is not in it; "No surviving record" and blanks are left out of every denominator. Money is in pence, as in the source.
Dates. "1350-1450" is read as written; "15th century" as 1400 to 1499; "early", "mid" and "late" as the first 40, middle 40 and last 40 years of a century; "post 1761" as 1761 to 1810; text naming several phases gives several ranges. Each site's weight of one is spread evenly over its years.
Domesday. Each site is matched to a Domesday place within 5 km whose name agrees (after setting aside brackets, Great/Little and similar, and allowing small spelling differences), or that shares a Domesday entry with the site's own Domesday reference. Where both figures exist, my matched household count and the gazetteer's own minimum of individuals agree within one for 78% of 898 sites (rank correlation 0.78). 73 matches where the site's name has a qualifier the Domesday name lacks are left out of the comparison.
Setting. Height and relief from the AWS Terrain Tiles (about 100 m cells here); watercourses from OS Open Rivers, without canals; Roman roads from Itiner-e. Comparisons use medians, Mann-Whitney tests and within-county ranks.