Romanian words, debated origins
Which words might come from Dacian?
Dacian has left no texts that can be read without dispute. Barză, brad and mazăre are among the Romanian words traditionally assigned to a pre-Roman layer, often called “Dacian” or “Daco-Thracian”. The strength of the evidence varies by word. Compare twelve Romanian–Albanian pairs with the dictionaries’ explanations and reservations.
- 12
- words with an Albanian mirror and a dictionary verdict
- about 90
- words shared with Albanian in Brâncuș’s inventory; about 160 proposed as substratum by Russu (1959)
- 25 of 42
- plant names Decev accepts as Dacian
The mirror is alive: old pairs, on either side
- barză bardhë “white”
- brad bredh “fir”
- mazăre modhull “vetch”
The word atlas
Twelve words with debated origins
Each card sets a Romanian word beside its Albanian mirror and the dictionary's verdict — the exact wording, quoted, so you can check it. The badge says how well the verdict is supported, the small note says what weakens it, and each card links to its dictionary page.
How to read the badges
- WELL SUPPORTED: several dictionaries and studies call the word “autohton” (or “Thraco-Dacian”), and nobody overturns the Romanian–Albanian correspondence. The badge measures the strength of the argument — not absolute certainty.
- PROPOSED AS SUBSTRATUM: an important traditional attribution, but a major dictionary abstains, marks it “uncertain” or compares elsewhere — or the direction of borrowing is disputed.
- DISPUTED: the competing etymologies are serious, and no dictionary signs “autohton”. Likeness alone makes no origin.
barză n.
the long-legged white bird
bardhë WELL SUPPORTED nature
bardhë means “white”: the mirror is the adjective, not the bird. DER prefers Latin *gardea.
“cuv. autoh.; cf. alb. bardhë” (DEXI)
brad n.
the Carpathian conifer; at Christmas, the trimmed tree
bredh WELL SUPPORTED nature
“cuv. autoh.; cf. alb. *breth, bredh” (DEXI)
mazăre n.
the sweet seed in the pod
modhull WELL SUPPORTED garden
DEX cites it as modhullë; today's Albanian standard says modhull — a wild vetch, not the garden pea.
“Cf. alb. modhullë” (DEX); “(cuv. autohton)” (DE)
țap n.
the male goat — and the beer mug
cjap PROPOSED AS SUBSTRATUM sheepfold
DER thinks the road runs the other way: the word “passed from Romanian” to its neighbours, Albanian included.
“cuvânt de origine traco-dacă” (MDA2); “Cf. alb. cap, cjap” (DEX)
balaur n.
the giant serpent of fairy tales
bollë WELL SUPPORTED nature
Widened sense: Albanian bollë is an ordinary snake; the Romanian balaur is a monster.
“cuv. autoh.; cf. alb. bollë “șarpe”” (DEXI)
strungă n.
the gate the sheep pass through at milking
shtrungë PROPOSED AS SUBSTRATUM sheepfold
DER holds that “from Romanian come undoubtedly” the Albanian forms — the direction is disputed.
“Cf. alb. shtrungë” (DEX); “Cuv. autohton (Philippide, …)” (DRAM)
grumaz n.
the neck, the nape
gurmaz PROPOSED AS SUBSTRATUM body
DER derives it from Latin gruma; “from Albanian” is “insufficient”. Today's standard Albanian lacks the word.
“Cuv. autoht.” (NODEX); “Cf. alb. gurmaz, gërmaz” (DEX)
năpârcă n.
the legless lizard, like a small snake
nepërkë PROPOSED AS SUBSTRATUM nature
DEX prints it undiacritized (“neperke”); the Albanian standard says nepërkë. DER: “Albanian must come from Romanian.” The direction is disputed.
“Cuvânt autohton (Russu, …)” (DRAM); “Cf. alb. neperke” (DEX)
rânză n.
rennet; the ruminant stomach
rrëndës PROPOSED AS SUBSTRATUM sheepfold
MDA2 marks it “ns” (uncertain); DER starts from Slavic ręsa, “catkin”.
“cuv.autoh.; cf. alb. rrëndës “cheag”” (DEXI)
țarc n.
the fenced pen, for sheep or toddlers
thark WELL SUPPORTED sheepfold
DEX cites the cark variant (and Greek tsárkos); the Albanian standard says thark.
“cuvânt de origine traco-dacă, cf. alb. cark, thark” (MDA2)
măgură n.
the mound, the rounded hill
magulë PROPOSED AS SUBSTRATUM nature
Scriban: “from Romanian comes Albanian mágulĭă”; DER sees a bare correspondence — the direction is disputed.
“Cf. alb. magulë” (DEX); “Cuv. autohton (…)” (DRAM)
leurdă n.
the wild garlic with white flowers
hudhër PROPOSED AS SUBSTRATUM nature
DEX compares Bulgarian levurda, not Albanian — the weakest link in the atlas.
“Cuv. autoht.” (NODEX); “cf. alb. hudhër “usturoi”” (DRAM)
The trap word: gard
gard n.
the fence
gardh DISPUTED
DEX ’09 compares and stops: “Cf. Alb. gardh, Slavic gradŭ”. MDA2 pulls it into Slavic: “E: OSlav. градъ, cf. Alb. gardh”. No “autochthonous” label anywhere — this is what a mirror that fails looks like. The gard lesson: likeness alone proves nothing about origin.
Three of them, drawn
Twelve checked words, each with its mirror and its verdict. The method behind them is an investigation — and it does not believe in coincidences.
Why Albanian?
Romanian is a Romance language — yet a clutch of words and a few old features have special parallels in Albanian. These correspondences are one of the main windows onto the linguistic prehistory of the Balkans. What kind of kinship links the ancestors of the two languages — shared inheritance, very old neighbourhood, or something else — stays debated. That is why Albanian appears at every step of this page: it is the mirror, not the verdict.
How to read a language with no dictionary
No secure inscription says “brad means fir”. The method is an elimination carried through — attestations, dialects, sound correspondences, old senses — with one hard question at the end: loan or inheritance? Follow brad through the five sieves.
Latin?
No accepted Latin etymon explains the form; the sister Romance languages lack it.
Other layers?
The languages Romanian touched — Slavic, Hungarian, Turkish — are checked, with the possible Indo-European parallels.
Albanian?
Bredh is the proposed correspondent: the same tree, related forms — not identical.
Loan or inheritance?
Sound patterns and chronology must show whether it is an old loan or a deeper shared layer.
The cautious verdict
A word many linguists count as substratum — not a “proven Dacian word”.
Sieve 1 of 5
Why the pattern matters
A single word says little. But dozens of Romanian–Albanian correspondences point to a significant prehistoric link — shared substratum, very old contact, or something else: the nature of the link stays partly debated.
One point of method: historical linguistics never compares two words by their looks. It asks for regular sound correspondences, attested old senses and plausible chronologies — without them, any likeness stays an elegant coincidence.
One precision: this page says “Dacian” for short. Some linguists, Georgiev among them, place the substratum in a “Daco-Moesian” language, distinct from Thracian proper. More: substratum status is little disputed, but the words' origin — Thraco-Dacian, Albanian, or a third language — stays uncertain. A 2025 Cambridge synthesis urges caution: the Albanian–Romanian link may be shared inheritance or old neighbourhood. The label is debated; the correspondences stand.
How many are there? It depends how strictly you count
The Brâncuș inventory counts about ninety words shared with Albanian, judged by its author to be part of the autochthonous stock; a 2025 Cambridge synthesis counts, on its own account, 70–90 Albanian–Romanian cognates. The wider inventory associated with Russu, from Limba traco-dacilor (1959), proposes about a hundred and sixty items for the substratum. Choose the criterion and watch the atlas: on strict criteria the “proposed as substratum” badges go out — an illustration of rigour, not a checked subset of Brâncuș's ninety.
Loose criteria (Russu): about 160 words
Dropped: 7 “proposed as substratum” plus gard — 5 “well supported” remain.
On this page, five of the twelve cards carry the “well supported” badge, seven the “proposed as substratum” one. Gard drops first — it was never in the race. The badges are the page's own, after the dictionaries — not subsets of the inventories above.
Traces in manuscripts
A few dozen plant names, in copies of a Greek book
The Greek physician Dioscorides writes a book about plants in the 1st century, De materia medica. The lists of ethnic synonyms — what the Dacians call them, the copyists say — belong to the late manuscript tradition called Pseudo-Dioscorides: we know neither their author nor their exact date of collection. The classic hypothesis (Wellmann; Russu) places them in the 3rd century, from an older glossary; all we know for sure is that they appear in the late copies, the oldest being the Vienna one, from around 512. Between forty-two and some sixty names carry the “Dacian” mark.
dielleina
dielina, diellena
henbane · Hyoscyamus niger
The manuscripts give several forms. A 2023 study ties it to Albanian diell, “sun”.
skiare
wild teasel · Dipsacus fullonum
The gloss attributed to the Dacians for wild teasel, kept in a 6th-century Vienna manuscript.
stirsozila
centaury · Centaurium erythraea
Mind the source: this name comes from Pseudo-Apuleius’ Herbarius, not from Dioscorides.
Decev's reckoning: of 42 names, 25 Dacian, 10 Latin, 7 Greek; in Pseudo-Apuleius, 16 of 31. That is why this page gives no fixed number. Beyond plants, the language also left proper names and place names — real traces, but too short to make sentences. And beyond the manuscripts, the only possibly Dacian written trace — the DECEBALUS / PER SCORILO sealmarks — sits on separate seals and is disputed (Mócsy).
Four dates, four thresholds
The words need a little history around them. That much: four lines, while the story in stone stays with the siblings.
106
Dacia becomes a Roman province, after two wars against Decebalus.
271
Aurelian decides to abandon the province.
1887
Hasdeu prints the first volume of Etymologicum Magnum Romaniae.
1970
Russu gathers the inheritance in Elemente autohtone în limba română.
The story in stone: the Dacian fortresses and the Dacian frontier.
What other languages took from Romanian, Albanian among them: Romanian words in other languages.
Can the words say where Romanian formed?
In short: they feed the debate but do not settle it alone. The Romanian–Albanian correspondences constrain theories of contact and formation — an exact map they do not give.
The south-of-the-Danube hypothesis
- Roesler, Romänische Studien (1871): the Romanians formed south of the Danube.
- Weigand (1895; 1927) restates the case on linguistic tracks; Schramm, Ein Damm bricht (1997), restates Roesler and Weigand.
The continuity hypothesis
- Rosetti, Russu, Brâncuș: the stock shared with Albanian is Dacian inheritance kept north of the Danube.
- Madgearu answers Schramm: the shared words are Dacian survivals in Romanian.
“Daco–Roman continuity cannot be proved from the place names north of the Danube — and the names south of the river point to no southern cradle either.” That is why the words matter.
Match the mirrors
Twelve Romanian–Albanian pairs, shuffled. Pick a word, then its correspondent.
Romanian
Albanian
Score: 0 of 12
Twelve of twelve — in the game. The verdict stays with the dictionary.
The pairs, face up
| barză | bardhë |
| brad | bredh |
| mazăre | modhull |
| țap | cjap |
| balaur | bollë |
| strungă | shtrungë |
| grumaz | gurmaz |
| năpârcă | nepërkë |
| rânză | rrëndës |
| țarc | thark |
| măgură | magulë |
| leurdă | hudhër |
What we do not know (and say so)
- When the substratum language died — no extinction date is known.
- How many there are, exactly: about ninety and a hundred and sixty are inventories using different criteria, not bounds on a proven number of Dacian words.
- Who borrowed from whom, word by word — the mirror shows kinship, not direction.
- Whether “Dacian” is the right label for the whole inheritance.
We know an old layer exists. The Dacians left no dictionary — but you speak its possible words every day.
Sources
Dictionaries first, then the bedrock books, the manuscripts and the scholarship. Each word card links to its dictionary page.
Dictionaries
- barză, brad, mazăre, țap: dexonline
- balaur, strungă, grumaz, năpârcă: dexonline
- rânză, țarc, măgură, leurdă: dexonline
- gard (the disputed case): dexonline
Bedrock books
- Grigore Brâncuș, Vocabularul autohton al limbii române, Bucharest, 1983
- I. I. Russu, Elemente autohtone în limba română, Bucharest, 1970
- I. I. Russu, Limba traco-dacilor, Bucharest, 1959
- D. Decev, Die thrakischen Sprachreste, Vienna, 1957
The manuscript and the Dacian plant names
- Vienna Dioscurides (EN-wiki)
- List of Dacian plant names (EN-wiki)
- The Dacian language: Decev, Wellmann, Asher & Simpson (EN-wiki)
- Crăciun 2023, on diellina (DOI)
The count and the quarrel
- Romanian words of possible pre-Roman origin, with the Russu 1970 bibliography (EN-wiki)
- Friedman & Joseph 2025, ch. 1: 70–90 cognates (Cambridge)
- Romanian school manual (grade XI) on Russu 1959: 160 (PDF)
- Substrate in Romanian: state of the dispute (EN-wiki)
- Curta 2022, Aging Levee (DOI)
- Schramm 1997, Ein Damm bricht (ECMI record)
- Roman Dacia (EN-wiki)
Type: Fraunces, Newsreader and Atkinson Hyperlegible, under the SIL Open Font Licence.