Two maps of the same continent, drawn a century apart by people with opposite interests. One records who spoke what, and where. The other records what was taken, and when. Put a coordinate through both and the gap between them is the history.
The language layer is Ives Goddard's Native Languages and Language Families of North America, the map he compiled to accompany volume 17 of the Smithsonian's Handbook of North American Indians (1996), in the revised and enlarged edition of 1999. It draws traditional ranges: where a language was spoken at first sustained contact, before the speakers of it were moved. The other layer is Charles Royce's schedule for the Bureau of American Ethnology, printed in 1899, the federal government's own accounting of the land it took from tribal nations from 1784 through 1894, together with the reservations it set aside, keyed to the treaty, statute, or executive order it claimed as authority.
Neither has been easy to query by coordinate. Goddard is a folded paper sheet, 99 by 132 centimetres. Royce is 67 map plates. Both have now been digitized by other people, openly, and this page just joins them and puts a cursor on top.
The ranges have no date
Every one of the 327 features in the Goddard dataset carries the same value in its
year field: the literal string traditional. There is no year, because the
source map assigns no year to any individual polygon.
The map is drawn to first sustained European contact, and contact did not arrive everywhere at once. It reached the Rio Grande pueblos with Coronado in 1540, the Southeast with de Soto and then the Carolina trade across the sixteenth and seventeenth centuries, the Great Lakes with the French in the 1600s, the outer Northwest coast in the 1770s but its interior only after 1805, and the Great Basin and interior Plains not until the fur trade and the emigrant roads of the 1820s to 1850s. Roughly three centuries separate the earliest of those moments from the latest. So the language layer is a composite of many different instants presented as one, and it is not a snapshot of any single year. Read it as "the traditional range", never as "the map in 1600". The regional dates in this paragraph are ours, offered as orientation. The sheet itself assigns none.
The Royce layer is the opposite: precisely dated, because a government that is taking land writes down the date.
What the readout is telling you
- Language and family come from Goddard, joined to a Glottolog glottocode so the name is unambiguous and you can follow it into the linguistic literature.
- Royce entries appear in the order they were recorded. Each names the tribes as the schedule named them, which is usually the treaty's own spelling and often a colonial rendering of an exonym rather than what people called themselves. "Dwamish" is the spelling in the 1855 treaty, and Royce copied it. Duwamish is the nation.
- Present-day nations is the National Park Service's National NAGPRA Program crosswalk from a historical entry to the federally recognized tribes today, redistributed and then revised by the Forest Service, whose revisions NAGPRA has not reviewed. It is long, and it is long for a reason: removal scattered one signatory nation across many modern reservations in many states. Where it appends "[Unrecognized]" to a nation, that is a statement about the federal government, not about whether the nation exists.
- Most entries link out to the scanned page. Royce's schedule entry on the Internet Archive covers all but one of the 717 mapped entries, the original plate at the Library of Congress covers 694, and where a treaty exists its text in Kappler on HathiTrust covers 474. Congress ended treaty-making in 1871, so later entries link to a statute (94) or an executive order (123) instead. Fifty-four link to no instrument at all.
The scrubber
Drag the year back to 1784 and forward. The language ranges do not move, because the dataset has no time dimension to move them along. The ochre does. That asymmetry is the point: one side of this history was recorded with dates and legal instruments, and the other was not.
What this is not
It is not a legal document and it will not tell you whose land you are on in any sense that a court, a nation, or a land acknowledgment should rely on. Territories were fluid, seasonal, shared, and overlapping, and a single polygon is a cartographic convenience that the underlying reality does not have. Goddard mostly draws his ranges as disjoint, but not always: Laurentian is drawn over the whole of Western Abenaki, Algonquin over Atikamekw, and two Californian pairs share identical geometry. Where ranges overlap this page names all of them, because the data gives no basis for preferring one.
Not every ochre polygon is a cession. Royce's schedule also plots reservations, and for tribes the United States never treated with at all, such as the eastern and western Apache, it plots "country claimed by them at beginning of their relations with the U.S." beside the note that no treaty of purchase was ever made. Twenty-nine entries carry only that notation, including the ones under Santa Fe and Phoenix, and several Pueblo entries are Spanish grants the United States confirmed, which is to say retained land rather than ceded land. Presence of ochre is no more a finding than its absence is.
The record is a colonial accounting. It documents dispossession thoroughly because the dispossessor kept the books. Land taken without a recorded instrument does not appear here at all, and the absence of ochre is not evidence that nothing happened.
These are not dead languages. Many of the ranges on this map belong to languages with living speakers and active revitalization programs today, and a map drawn to the moment of contact says nothing about that. If you want to know about a place, the nations whose place it is are the source. This page is a finding aid pointed at the archive, not an answer.
Sources, all open
Reproduce it
The two source datasets are a download and an HTTP query, the join is a coordinate and a point-in-polygon test, and the whole thing runs in a notebook in about thirty lines. A Colab notebook and an R notebook that do exactly this, from the live sources, accompany this page, and both assert the same worked examples. For the browser, the geometry here is additionally clipped to a lower-48 box and simplified to a 0.008 degree tolerance; run the notebooks against the sources for the full-resolution originals.
Coverage is the lower 48. Alaska is in the Goddard dataset and is not on this map, because a composite projection would have to break the continuous geometry that makes the ranges visibly cross into Canada and Mexico, which they did. Hawaiʻi is not in the Goddard source at all.