Cadastral survey patterns

Why can you name the century a countryside was surveyed in from the shape of its fields alone?

Fly over Ohio and the ground is a chessboard: square farms, right-angled roads, everything indifferent to the hills it crosses. Fly over Kent and there is no chessboard at all — fields shaped like spilled water, lanes that curve for no visible reason. Fly along the St Lawrence and you get ribbons of land, absurdly narrow, all running back at the same angle from the river.

None of those shapes is a response to what grows there. The tempting explanation — flat land takes a grid, broken land takes irregular fields — collapses on inspection, because the American grid was driven straight over ridges and swamps, and much of the flattest ground in Europe carries fields no surveyor could have drawn. So the shapes record something else. What, exactly?

Start by asking what a boundary description has to do. It has to let a stranger, later, stand on the ground and know where the line runs. There are only a few ways to achieve that, and each carries a different bill.

The oldest is to describe the line relative to what is already there: from the great oak, north-east to the ford, then along the hedge to the stone. This is cheap — a rope, a compass and a walk. But notice its precondition. It works only if the land is already occupied, already full of agreed marks and neighbours to agree with. And notice its later cost: oaks fall, hedges are grubbed out, and adjacent descriptions written by different parties do not necessarily close, so parcels interlock imperfectly and leave slivers. Cheap now, expensive forever.

Now change the precondition. Suppose a state means to sell land that its buyers cannot visit, before anybody has settled it. What must the description do then? It must be complete and self-contained, because there is nothing on the ground to point at and nobody to ask. That forces a coordinate system: fix a base line and a principal meridian, and number everything from them. A purchaser in a distant office can then buy a parcel by name alone, and the clerk can be certain no one else has been sold it.

The American Public Land Survey is that argument made law, and its arithmetic is worth doing rather than reciting. A township is six miles square. Six by six gives thirty-six sections, each a mile square; and since an acre is one six-hundred-and-fortieth of a square mile, a section is 640 acres, a quarter-section 160, a half-quarter 80. Those are not measured quantities. They fall out of the definition — which is the point: the system was designed to be computable in an office rather than measured in a field.

Then a geometric problem the earth imposes and the rule cannot dodge. Meridians converge toward the pole, so lines run true north cannot stay six miles apart. A township measured six miles wide on its southern edge is narrower on its northern edge. The survey handles this by resetting the grid at intervals along standard parallels, and the correction shows up as a jog — a road running dead straight for many miles, then stepping sideways and resuming. That jog is the curvature of the earth made visible in a county road.

And a third arrangement, for a third problem. Where the scarce good is not area but frontage — access to a river that is the only road — dividing land into compact blocks would give some holdings water and others none. The answer is the long lot: narrow strips running back from the line of access, so every holding touches the river and every holding crosses the same succession of soils from bank to backland. It looks wasteful and it is exactly rational.

The pattern is therefore not a record of the terrain, nor of the culture in any loose sense. It records the procedure by which the land was allotted, and above all the order of two events: did the survey come before the settlers, or after them? Survey first forces a self-contained coordinate grid. Settlement first permits, and rewards, description by local reference.

Two honest corrections follow. First, terrain has not vanished — it enters as a cost rather than a cause. The grid ignores contour, so its roads climb slopes they need not climb and its farms later needed contour ploughing to stop the soil leaving; the long lot is shaped by a river's existence, though not by its shape. Constraint, not selection.

Second, the map shows the most recent allotment, not the first. Enclosure and post-war land reform have both overwritten irregular patterns with regular blocks on ground nobody re-conquered, and the earlier layer survives only in fragments — a curving hedge inherited from a strip field, a parish boundary wandering through a rectilinear countryside. So "the century it was surveyed in" means the century of the last re-allotment, which is why the dating works at all and also where it goes wrong.

That gives a clean falsification test. If the order of settlement and survey is doing the work, any territory a state surveyed before selling should carry a regular grid regardless of relief, and any territory settled before it was surveyed should carry irregular parcels regardless of how flat it is. What would refute it: a large regular grid lying over a countryside densely settled centuries earlier with no recorded re-allotment, or metes-and-bounds used as the primary instrument in a territory whose state surveyed and numbered the ground before opening it to purchase.

• Why land registration by title rather than by deed changes what a boundary dispute is actually about.

Every term the collection defines is gathered in the glossary.