Definitions
The six definition bands — each with a named authority, an explicit criterion, and its own closure rule.
Six answers to “what is a castle?”. Each is a real definition held by real people; none is this site’s preferred one. A structure takes a verdict against each band separately, and each band publishes its own count.
The man best placed to settle the question declined to. Philip Davis built a gazetteer of roughly six thousand fortified sites and concluded:
In effect it is impossible to define ‘the castle’ since there is no one such thing. Much of the controversy in castle studies between militarists and revisionists lies in a befuddled use of examples of ‘castles’ of very different function.
These bands are not a better definition. They are the consequence of accepting that one definition was always the wrong shape.
| Definition | What it means | Closure |
|---|---|---|
| Fortified residence | The strict scholarly castle: a fortified private residence of a lord. Both halves are load-bearing — a fortress nobody lived in is not one, and an undefended great house is not one either. | |
| enumerable | ||
| Enclosure fortification | Defended enclosures with no lord and often no permanent residence — hillforts, ringforts, raths, cashels, brochs, promontory forts. The band that turns Ireland from a 4,552-castle country into a 31,431-site one. | |
| enumerable | ||
| State fortification | Works built and garrisoned by a state rather than held by a lord: artillery forts, bastion and star forts, citadels, coastal batteries, and the fortified systems of the gunpowder age onward. | |
| enumerable | ||
| Palatial seat | Buildings called “castle” in their own language whose defensive function is absent or vestigial — the German Schloss, the French château, the country seat. Where most of the world’s famous “castles” actually sit. | |
| open | ||
| Revival and folly | Castles built as architecture rather than defence: nineteenth-century romantic revivals, castellated country houses, and sham castles that are sometimes a screen wall and nothing behind it. | |
| open | ||
| Popularly a castle | What an ordinary person would call a castle and travel to see. No scholarly standing whatsoever, and by far the most useful band for anyone actually planning a visit. | |
| open |
Each band has its own page with its full criterion, its inclusions and exclusions, the authorities that hold it, the registers that bound it, and every record that satisfies it.
What “closure” means
A band is enumerable when a named register bounds it — a finite, citable list a person could in principle work through. Those bands can reach a real integer, one jurisdiction at a time.
A band is open when nothing bounds it and nothing could. “Popularly a castle” has no register and never will, because popular usage has no edges. Those bands publish a distribution or a range with its assumptions visible, never a fabricated total.
This distinction is enforced in the data, not just in the prose: a coverage cell for an open band that carries a world-population figure fails CI.
Changing a band
Bands are the dataset’s contract with its readers — moving one silently moves every number on the site. Amendments go through a decision-request issue and a PR, never a data edit.
Two are known to be unfinished. Revival / folly currently holds both what Davis calls a symbolic castle (built in castle form to assert status) and a sham castle (built as scenery); he insists these are not the same and he is probably right. And the bands are a European instrument throughout — applied to Chittorgarh they return two contested verdicts rather than knowledge, because the European lord-versus-state axis does not transfer to Rajput polity, which classified forts by terrain instead. That is a limitation of the bands, not of the fort.