Nomen Errans

Historical geography often begins with a proper noun and quietly treats it as a stable key. A place record called Dacia seems to promise that every occurrence of the word can be joined to the same object. That is convenient for a database and dangerous for an argument.

The corpus separates three things that prose too easily collapses. The lexical form is the written word. The referent is what a particular use makes the word indicate. The relationship is an evidenced claim about how two uses are connected. Repeating the first does not establish either of the other two.

This distinction matters even inside Roman administration. The Trajanic province, the later Dacia south of the Danube and the dioecesis Daciae are not three dates for one stable polygon. They are different administrative objects. The coin legend is another kind of use again: a mint makes the name circulate without supplying a boundary. Ptolemy’s coordinate catalogue, Eutropius’s narrative, the Notitia Dignitatum and the imperial coin series answer different questions. The ledger keeps their source families because their silences differ as much as their words.

The release page does not copy their chronology into prose. The panels below are compiled from the same tables as the data release. Correcting a date or locator in the corpus changes the essay on the next deterministic build; demoting a row below review removes it from the public sequence. The text cannot preserve an older, more confident version by accident.

Each step is one reviewed use. The fate class names the operation performed on the word: applicatio applies it to an administrative or representational referent; translatio carries it to different ground; restitutio deliberately restores an earlier name. These labels classify what the source does. They do not claim a genealogy between sources.

The medieval step is the fracture that makes the method necessary. On the Hereford Mappa Mundi, around 1300, Dacia belongs to the Scandinavian north. The source is one membrane and therefore has no page or sheet number to pretend to precision; the whole work is the locator. Its reviewed row says what the map calls the region. It does not say that Denmark descends from the Roman province, that the name moved there by one recoverable route, or that a medieval reader understood every earlier career gathered on this page.

The early-modern return is different again. Ortelius’s Daciarum Moesiarumque vetus descriptio belongs to the Parergon, the historical supplement to the Theatrum. Its Dacia is an antiquarian reconstruction: an ancient province made present on a printed sheet through a Renaissance reading of old authorities. The fate class is therefore restitutio, not continuity. The map’s date tells us when Ortelius restored the province to view, not that the province had persisted until 1595.

The source apparatus beneath the panels is as important as the sequence above them. A repository, citation and locator let a reader ask whether a row says only what its witness can support. Confidence is spelled out rather than encoded by colour. Careers without an honest Atlas layer say so instead of opening the nearest available geometry and hoping proximity will be mistaken for evidence.

The diagram below makes the problem deliberately uncomfortable. It has reviewed nodes, a chronological direction, and relationship lines that had to be earned one at a time. Every line is a derivation, a revival, or - drawn dashed - an explicit homonym only: two careers that share a string and nothing else. Not one asserts continuity, because the schema will not let a continuity edge exist without an attestation to carry it, and none has been cited. Other relationships have passed review and are still not drawn, because they land on a use that has not; a reviewed edge is never permission to draw a line to a referent nobody has cleared. Chronological order alone draws nothing at all.

Some careers can still open useful map states. The Trajanic province opens the Roman sites and network at AD 150. The coin legend opens the province it names, not a guessed findspot for an unspecified coin. Ortelius opens the research-tier attestation that records the reconstruction. Every link carries the essay slug, so the Atlas can return the reader to this passage rather than dropping them at a generic catalogue.

Other careers correctly open nothing. The southern province and the civil diocese have no compiled geometry. Pointing either at the Trajanic layer would turn the same word into the same territory. Scandinavia is outside the extent of every Dacia layer; that absence is the essay’s finding, not a missing feature request.

The result is a map that distinguishes three answers: mapped, not compiled and outside coverage. Most digital maps offer only present or absent. Here the reason for absence carries the argument, because an unbuilt layer and an irrelevant extent are not the same epistemic state.

Historical-map rights do not follow automatically from the age of the work. The ledger assessed six candidate witnesses. The Hereford Mappa Mundi photograph needs permission; the Tabula Peutingeriana digitisation, an Ortelius impression and a coin specimen need object-level rights review; the exact Notitia page remains unlocated. They stay research-only, so this essay describes them and reproduces none of their images.

One later document clears the production bar. Decree no. 194 of 16 October 1974 survives on page 1 of the Romanian official bulletin. The scan is sufficient, its object identity is fixed, and the source page records the official text and its document photograph as exempt from Romanian copyright protection. The operative article adds Napoca to Cluj. That makes it a useful boundary object for this essay: rights permit the page to be shown, but the two name-use readings attached to it remain normalized rather than human-reviewed, so neither is promoted into the career sequence above.

That asymmetry is deliberate. Reuse permission answers whether the page may be shown. Scholarly review answers what claims the publication is prepared to make from it. A permissive licence cannot promote an interpretation, and an excellent interpretation cannot supply image rights. The two gates meet only at release.

The wandering in Nomen Errans is not a biography of a word with every missing year filled. It is a record of distinct, inspectable employments. Between them lie gaps in evidence, changes of institution, shifts of medium and, in the Scandinavian case, different ground altogether.

This changes what an atlas can responsibly do. It can compare referents without merging them. It can direct a reader from a reviewed use to a relevant layer and back. It can expose the reason a layer is absent. It can record a candidate relationship without drawing it. What it cannot do is make sameness of spelling stand in for sameness of territory, people or authority.

The most consequential mark in the diagram is therefore the one line it does not draw: the unbroken path from Trajan’s province through every later Dacia. No edge in the ledger is a continuity, because continuity has to cite an attestation and not one does. A continuous path would be visually satisfying and historically premature. Its absence makes a smaller claim than the lines around it, but a more useful one: every connection in an atlas should be able to say who reviewed it, what source supports it and what kind of relationship it asserts.

Continue through The Border, inspect naming as an archival problem in The Archive, or return to

Terra Sigillata for the longer history of how Dacia was measured, depicted and withheld.