How were US county boundaries drawn?

Rivers in the east, survey lines in the west, and a visible seam running roughly down the middle of the country.

There are two answers, depending on whether the surveyors arrived before the settlers or after them. Where people got there first, boundaries described ground that already had names: this river, that ridge, the corner of a named man’s field. Where the survey got there first, boundaries were arithmetic on a grid drawn before anyone lived on it. The seam shows on any national county map: east of the Ohio the lines wander, west of it they run straight for long stretches and then step sideways.

Land settled before it was surveyed

The original colonies used metes and bounds, which describes a parcel by walking its perimeter: a bearing, a distance, and a monument at each corner. The monuments were whatever stood there, a marked oak or a pile of stones or a creek bank, so the deeds are precise about direction, vague about everything else, and dependent on objects that decay or move.

Counties in that world were not designed. They were subdivided. A young state would create one large county, settlement would thicken in one corner of it, and the legislature would split off a new county described in the vocabulary the deeds used: along the river to the mouth of a named creek, up the creek to the ridge, along the ridge to a gap. Each decision was made by people who could see the ground, not by anyone holding a map of the whole state, which is why the result looks improvised. It is also why eastern counties tend to be small and irregular while their western neighbours in the same state are larger and more regular. The institution itself is covered in what a county is.

Land surveyed before it was settled

The Land Ordinance of 1785 set a different rule for the public domain, the land the federal government held and intended to sell. Before any of it could be sold it had to be surveyed into a rectangular grid, the Public Land Survey System.

The system starts from a principal meridian running north and south and a base line running east and west. From that intersection surveyors laid out townships six miles square, each divided into 36 sections of a square mile. Land was described not by walking its edge but by naming its cell: township, range, section, quarter section. A buyer could purchase a parcel he had never seen because the description depended on nothing standing on the ground, and a legislature could define a county the same way, as a block of whole townships.

The grid has one unavoidable defect: meridians converge as they run north, so a lattice of true north lines cannot hold townships of a fixed width. Surveyors reset it at intervals along standard parallels, also called correction lines, and that reset is the usual explanation offered for the sudden jogs in county and road lines across the plains. Treat it as a likely cause rather than a diagnosis: a particular jog can equally come from a river, a state line offset, or a later act of a legislature, and only the boundary description settles it.

Two regimes, side by side

Feature Settled first Surveyed first
Where it dominates The original Atlantic states, Kentucky, Tennessee, much of the South The Midwest, the plains, the Mountain West, the Pacific coast
How a parcel was described Bearings, distances, monuments on the ground Township, range and section on an abstract grid
What county lines follow Rivers, creeks, watershed divides, ridgelines Survey lines, plus a river large enough to matter
Effect on digitised data High vertex counts, borders shared with hydrography Low vertex counts, long straight segments

The columns are a summary: the frontier between the regimes is untidy, and Louisiana, Hawaii and parts of the Southwest came in with land systems of their own. Texas sits outside both. It kept its own public lands on joining the Union and was never surveyed under the federal system, yet many of its western counties are rectangles anyway. The likely reading is that the sequence, not the grid, does the work: where a boundary is drawn before anyone lives inside it, there is nothing on the ground for it to follow. The state’s 254 counties are what that ordering produces at scale. Georgia is next with 159.

The features that keep recurring

Rivers were the transport network and everyone knew where they were, so they were the first thing a boundary could be tied to. Ridgelines and watershed divides came next, a ridge being the edge of the valley a community uses, and the survey grid came last, only where it existed.

River boundaries carry a defect clerks have argued over ever since. A boundary fixed to a channel is fixed to it as the channel stood the day the statute was written, and rivers migrate, occasionally jumping their bed in a single flood. The general doctrine is that a boundary follows gradual movement but stays put after a sudden one, though the rule varies by state and by the wording of the individual statute. Where it bites, a pocket of one county is left on the far bank, attached to its neighbour physically and not legally.

A day’s ride to the courthouse

The constraint that set county size in the settled-first regime was travel. Land records, court cases, taxes and probate all required a trip to the county seat, on foot or by horse, and a county whose furthest corner lay more than a day from its courthouse produced petitions to split. The familiar version, that a county was sized so a farmer could ride in and get home by dark, summarises the pressure well enough, though the mileages popular accounts attach to it are folklore rather than statute. The constraint has since evaporated and the counties it produced have not.

What the two methods left behind

The 50 states and the District of Columbia hold 3,144 county-equivalents, and neither the vocabulary nor the size is uniform: Louisiana has 64 parishes, and Virginia’s 133 are 95 counties plus 38 independent cities. Size is the clearest fossil of the two methods. Yukon-Koyukuk Census Area in Alaska covers 145,767.49 square miles, a residual across country never subdivided for settlement; outside Alaska the largest is San Bernardino County, California at 20,068.42; Falls Church city, Virginia, cut out of a county around a single municipality, is 2.05. Boundaries outlived the governments inside them: 35 entities have no county government and 21 have no county seat.

The lines move, and the code does not always move with them

County geometry is not fixed. The Census Bureau publishes boundary changes annually, running from a few metres of annexation up to an entity vanishing: Bedford city, Virginia reverted to town status inside Bedford County on 1 July 2013 and has gone from current files.

Valdez-Cordova Census Area (02261) was replaced by Chugach (02063) and Copper River (02066) on 2 January 2019, while Petersburg Census Area became Petersburg Borough and kept GEOID 02195, so there the code held and the legal type did not. Connecticut retired eight counties for nine planning regions with the 2022 vintage, and because the regions are groups of towns rather than reassembled counties, no county to region crosswalk exists. FIPS codes are assigned alphabetically within a state, so a rename usually forces a new code, which is how Shannon County, South Dakota became Oglala Lakota County and moved from 46113 to 46102 in May 2015. The pitfalls of these identifiers are worth knowing before you build anything that joins on them.

Vintage is a property of the shape, not just the numbers

Anyone working with county data checks the vintage of the statistics; the same check belongs on the geometry. A boundary file is a snapshot of where the lines were on one date, at one scale, from one process, none of which is implied by the FIPS code in the attribute table.

This site’s maps come from the Census cartographic boundary files at 1:500,000 rather than raw TIGER/Line, because TIGER polygons extend to the legal boundary out into territorial water: a TIGER map of Virginia fills in the whole Chesapeake Bay and fuses the Eastern Shore to the mainland. The clipping and simplification are described in how this site maps counties. The rule: record which file your geometry came from and which year it is, and never assume a matching code means a matching shape.

Sources: US Census Bureau, Population Estimates Program, Vintage 2025 Population Estimates, released 26 March 2026; US Census Bureau, Gazetteer Files, 2025 Gazetteer Files, released 10 September 2025; US Census Bureau, Cartographic Boundary Files (1:500,000), 2025 Cartographic Boundary Files, released 23 April 2026. Retrieved 30 August 2026.

Figures quoted here are rebuilt from the Census Bureau files listed above each time the data pipeline runs, most recently 30 August 2026. Where a figure is computed rather than published, the text says so.