Place names
A Socratic walk-through of place names — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #What can the names on a map tell you about who lived there before?
Look at a map of eastern England and the endings cluster oddly. Grimsby, Whitby, Selby, Scunthorpe — thick on the ground in Lincolnshire and Yorkshire, and almost absent in Devon. Nobody planned that distribution, and no single authority ever imposed it.
Which raises a question worth taking seriously: if names were coined by ordinary people naming their own surroundings in their own language, then the map is a record of who was doing the naming, and where. How much weight can that record actually bear?
Reasoning it through
REASONING #Start with what makes a name survive. A name is useful only if it is shared, so it passes from speaker to speaker rather than being reissued. That gives names a peculiar property: they outlive the language that coined them. Newcomers who cannot parse a name will still use it, because everyone else does, and it becomes an opaque label — a fossil of a vocabulary nobody around it speaks any more.
Which features should fossilise hardest? Ask what a newcomer would need to name and what they would not. A farmstead you built yourself needs a name; you will coin one in your own language. A large river you found already there does not — everyone locally already calls it something, and you adopt it. So the prediction is that big natural features carry the oldest layer and settlements carry the newest.
England fits. The major river names — Thames, Severn, Trent, Avon, Ouse, Derwent, Exe — are overwhelmingly not English; most are Celtic or older, and several are simply the word for water or river in a language the English did not speak. That is the deepest stratum, and it is deep precisely because it was inherited rather than coined.
Now the layers above it. The Romans built forts and towns; the English who came after found the ruins and did what people do with an unfamiliar structure — described it. Old English ceaster, borrowed from Latin castrum, gives the -chester, -caster and -cester in Chester, Doncaster and Cirencester, and these track Roman sites closely. Note what such a name really records: not Romans naming their own town, but the English recognising a Roman one.
Then the English themselves, naming what they made: -ton from tun, a farmstead; -ham, a homestead; -ley, a woodland clearing; -ing, the people of somebody. These blanket the country, which is what you would expect from the language that became the ordinary speech of it.
And then the Norse. Old Norse by means a farm or village, and thorp a secondary or outlying settlement; -thwaite marks a clearing, -toft a plot. Where do they occur? Concentrated in Yorkshire, Lincolnshire, the East Midlands and the north-west — and the concentration corresponds closely with the Danelaw, the region conceded to Scandinavian rule in the late ninth century. A distribution of farm names lines up with a political boundary drawn by treaty, without anybody intending it to.
Set beside that the Normans, who conquered the whole country a century and a half later and left remarkably few new settlement names — mostly they attached family names to existing ones, as in Ashby de la Zouch or Melton Mowbray. Why the difference? Consider who arrived. A thin layer of landholders takes over estates and records its ownership; a mass of farmers clearing and settling land coins names for the places it makes. The toponymic record is not measuring conquest. It is measuring who was doing everyday naming.
The analogy
THE ANALOGY #Think of the map as a road surface that has been resurfaced several times, where each layer of tarmac was laid only over the potholes rather than the whole road. Dig anywhere and you can read which resurfacings reached that spot, and in what order — but you learn about the patching crews, not about who drove on the road.
Road layers stay put in strict order, whereas names get reshaped by later speakers, deliberately replaced, or carried wholesale to another continent, so the sequence on the map can be disturbed in ways a genuine stratigraphy cannot.
Clarifying the model
THE MODEL #That last clause is where the honest limits live, and there are three worth naming.
The first is folk etymology. A name whose meaning has gone opaque gets reanalysed into words the current speakers do recognise, and the reanalysis then sticks in the spelling. The most elegant example is one already mentioned: incoming English speakers heard the Brythonic word abona, meaning simply "river", and took it for the river's proper name — so "River Avon" says river twice. Multiply that by centuries of scribes rationalising unfamiliar syllables and some names now testify to a misunderstanding rather than to a settlement.
The second is renaming. Names are also assets, and authorities replace them: New Amsterdam becomes New York, colonial administrations write over indigenous names, states rename towns after leaders and then rename them again. Every such act deletes a layer rather than adding one, and it is not random — the layers likeliest to be erased belong to the people with least power, which biases the record in a consistent direction.
The third is subtler. Names record language, not ancestry. A dense field of -by names proves that Norse was being spoken there by people naming farms; it does not by itself settle how many people came, or whether locals adopted the naming fashion of a new elite. That distinction has been argued at length over the Anglo-Saxon settlement, and place names alone were never able to resolve it — it took evidence of a different kind, genetic and archaeological, to constrain the numbers. Dating is coarse for the same reason: a name-element can stay productive for generations after the group that introduced it arrived, so a -by is evidence of a period, not a year.
A picture of it
THE PICTURE #How to readRead down the years and treat each row as a layer added to the same map rather than a place replaced. Note that two rows share the same span: the English were coining their own farm names and describing Roman ruins in the same centuries, so the strata overlap rather than stacking cleanly. The Norman row is deliberately thin — a conquest that changed the rulers changed the names very little. The last row is not a period at all but a process running throughout, which is why the record is a disturbed stratigraphy rather than a clean one.
What became clearer
WHAT CLEARED #Place names are a record of everyday naming, and everyday naming is done by whoever is living somewhere and making things there. That is why the oldest names cling to rivers nobody had to name, why the -by belt reproduces a ninth-century political frontier, and why the Norman conquest is nearly invisible on a map it entirely governed. The map reads as a stratigraphy of settlement — but a stratigraphy that later speakers reshaped by misunderstanding and later rulers edited on purpose, so it is best evidence about language and worst evidence about numbers.
Where to go next
ONWARD #- Why transferred names, like Boston in Massachusetts, break the stratigraphic logic entirely.
- How field names and minor features preserve older layers than settlement names do.
Key terms
TERMS #| Term | What it means |
|---|---|
| Toponym | a place name, considered as evidence. |
| Danelaw | the region of England under Scandinavian law and settlement from the late ninth century. |
| Generic element | the part of a name giving the kind of place: by, ton, ceaster, thorp. |
| Folk etymology | reanalysis of an opaque name into familiar words, which can then freeze into the spelling. |
| Brythonic | the Celtic language spoken in Britain before English, source of many river names. |
Every term the collection defines is gathered in the glossary.