Every ending was once a noun
The endings that make English place names sound like English place names are not suffixes in the grammatical sense. They are ordinary words — farm, fort, valley, clearing, village — that stopped being ordinary words and froze into the names of the places they described.
Once you know which word each ending was, most English place names become readable. Norton is the north farmstead. Sutton is the south one. Bradford is the broad ford. Oxford needs no explanation at all.
| Ending | From | Originally meant |
|---|---|---|
| -ton, -tun | Old English | Enclosure, farmstead, later village |
| -ham | Old English | Homestead or village |
| -ley, -leigh | Old English | Woodland clearing |
| -ford | Old English | River crossing |
| -worth | Old English | Enclosed settlement |
| -bury, -borough | Old English | Fortified place |
| -by | Old Norse | Farmstead or village |
| -thorpe | Old Norse | Secondary or outlying settlement |
| -thwaite | Old Norse | Clearing or meadow |
| -toft | Old Norse | Homestead, house plot |
| -chester, -caster | Latin via Old English | Roman fort |
| -combe | Brittonic via Old English | Short valley |
| -pen, -tre | Brittonic | Hill or headland; settlement |
The endings map onto settlement history
This is the part that makes place names genuinely useful as evidence, and the part a generator has to respect if its output is going to sound like anywhere real.
Names in -by and -thorpe are heavily concentrated in the areas of Norse settlement — Yorkshire, Lincolnshire, the East Midlands, parts of the north-west. Grimsby, Whitby, Selby, Scunthorpe, Cleethorpes. They are correspondingly rare in the south and west, which the Norse settlement did not reach in the same way.
Names in -chester and -caster mark Roman forts, so they follow the Roman road network and the sites of Roman military occupation: Chester, Manchester, Winchester, Doncaster, Lancaster. The two spellings themselves split roughly north and south.
Brittonic elements — pen-, tre-, -combe — survive most densely in the west and south-west, where the Brittonic-speaking population persisted longest: Cornwall, Devon, Wales and the Welsh borders. Penzance, Tredegar, Ilfracombe.
This is why a generated name that combines a Norse ending with a Cornish element reads as wrong even to someone who could not explain why. The two never coexisted in the same place, and the ear picks that up without knowing the history.
The first element is usually a person or a feature
If the ending says what kind of place it is, the first element usually says whose it was or what was distinctive about it.
Personal names are extremely common. Birmingham is the ham of the Beorming people — the followers of someone called Beorma. Nottingham was Snotingaham, the settlement of Snot's people, which lost its initial S after the Norman conquest for reasons that are not hard to guess at.
Where the first element is not a person, it is generally a physical description, a plant, an animal, or a direction. Oakley is the oak clearing. Shipley is the sheep clearing. Ashford is the ford by the ash trees. Easton, Weston, Norton and Sutton are simply east, west, north and south farmsteads, and they exist in quantity because every part of England needed one.
Occasionally the first element records something specific and long forgotten: a market, a church, a mill, a battle, a boundary stone.
Sound changes have obscured a lot of it
Place names have been spoken continuously for a thousand years or more, and they have worn down accordingly. The written form often preserves something the spoken form lost centuries ago, which is where the notorious English gap between spelling and pronunciation in place names comes from.
Gloucester and Worcester keep their Roman -cester in writing and have compressed it to a single syllable in speech. Cholmondeley is written with four syllables and said with two. Happisburgh in Norfolk is said Haysborough.
This causes a specific problem for anyone generating names: a name assembled from historically correct parts can still be unpronounceable, because real names had centuries of use to smooth them down and a generated one has had none. Some combinations of elements that are individually attested simply never occur together, because the resulting cluster of consonants was too awkward and got worn away or avoided.
The practical answer is to constrain which elements can follow which, rather than trusting that authentic parts will make an authentic whole. Authenticity of components is not the same as authenticity of output, which is a lesson this site has learned repeatedly and in several languages.
A few things that look like patterns and are not
Not every -ton is Old English tun. Some are later formations, some are contractions of something else entirely, and some are nineteenth-century inventions that borrowed the ending because it sounded right.
Names ending in -field, -bridge, -wood and -green are often much younger than they look. They can date from enclosure, from industrial expansion, or from a housing developer choosing a name. The form is medieval; the place may not be.
And a substantial number of place names are simply obscure. The first element may be a personal name attested nowhere else, a dialect word that did not survive, or a corruption so thorough that the original is unrecoverable. Reference works are candid about this, and any account that explains every name with equal confidence is overreaching.
The generator produces names in these patterns. It does not claim any particular output is a real place, and where a generated name happens to coincide with a real one that is coincidence rather than lookup.
Where this comes from
- The element meanings given here are the standard ones found in dictionaries of English place names, which agree closely on the common endings.
- Distribution claims describe well-established broad patterns of Norse, Roman and Brittonic settlement; local exceptions are numerous and individual etymologies are frequently disputed.