ILLE · -UL · -A · -I
Why does the article stand at the tail?
Le loup, el lobo, il lupo — the major western Romance languages keep their article in front. Romanian tacks its definite article onto the end: lupul. In the Balkans, though, it is not alone. The tail comes from Latin ille — but why it landed at the tail is still disputed among linguists.
- ille
- the Latin demonstrative the tail comes from
- 1521
- the article fully formed in Neacșu's letter
- 3 branches
- Romance · Albanian · Slavic — Greek sits out
Where the tail comes from
Ille, the demonstrative turned definite article
The material is Latin and nobody disputes it: the Romanian tail comes from the demonstrative ille. In Late Latin, ille was already shifting from demonstrative towards article — the sixth century dates that Romance grammaticalization, not the moment Romanian fixed its article at the tail.
“Starting with the 6th century, the Latin demonstrative/adjective pronouns ille and ipse underwent a series of complex changes that ultimately resulted in the appearance of the definite article”
Ivancu 2019 (after Kuteva, Heine 2008) ATTESTED
The chain, form by form
-ul, -l
from Lat. illum, via Rom. lu: omul, lupul, băiatul.
-i
from Lat. illi: băieții, ochii.
-a
from Lat. illa, via eua, eau: fata.
-le
from Lat. illae: fetele, tablourile.
-lui
from Lat. illui: băiatului; proclitic before proper nouns (lui Mahamet).
-lor
from Lat. illorum: băieților, turcilor.
-ei
from Lat. illaei: fetei.
Aldea 2004, table on p. 24 ATTESTED
Mind the -u-: on masculine nouns ending in a consonant, the article is preceded by the linking vowel -u- (băiat-u-l). The classic mistake is taking it for part of the article. The machine below follows Ivancu 2019's segmentation; other descriptions notate the bound morpheme as -(u)l (Cornilescu 2016).
Romanian has one more definite article: proclitic cel, from the demonstrative acel, grammaticalized in Old Romanian. This page's tail is the other one: the enclitic morpheme -(u)l. (Cornilescu 2016)
The page's machine
Strip the article
Tap each word: it splits into stem, linking vowel and article, following Ivancu 2019's segmentation. The -u- is NOT the article — it is the glue.
omul = om + u + l (Lat. illum)
lupul = lup + u + l (Lat. illum)
băiatul = băiat + u + l (Lat. illum)
băieții = băieți + i (Lat. illi)
fetele = fete + le (Lat. illae)
tablourile = tablouri + le (Lat. illae)
Feminine singulars (fata) and genitive-datives (fetei, băiatului) fuse or shift the ending — they do not split cleanly, so they stay in the chain above.
Who else glues the tail
Three branches, one tail
The postposed article spans three Indo-European branches: Romance — Romanian (om-ul), Aromanian (om-lu), Meglenoromanian (om-u); Albanian (kal-i); Slavic — Bulgarian, Macedonian and Torlak. Greek, surrounded by tail-gluers, keeps its article up front.
Albanian
mal → mali (the mountain), zog → zogu (the bird), veturë → vetura (the car). Suffixes: -i/-u masculine, -(i/j)a feminine.
Bulgarian
човек → човекът (the man), добър човек → добрият човек (the good man) — the article sits on the adjective, as in înaltul. Articles: -ът/-ят, -а/-я, -та, -то, -те in the plural.
Macedonian
Three articles by distance: човекот (the man), човеков (this man), човекон (that man). No parallel in any other standard Balkan language.
Greek is missing
ο άνθρωπος — the article stands up front, as it always has in Greek. Greek still counts as Balkan through most of the other key features. (Friedman & Joseph 2025, ch. 1 and 6)
Friedman & Joseph 2025, ch. 6: the article, branch by branch →
How the tail became a Balkanism
Four witnesses
Leake spots the tail back in 1814, in Albanian and Wallachian; Kopitar gives it a programme in 1829. Two centuries later, in the Cambridge synthesis (2025), the curiosity has become Balkanism number one.
Kopitar, 1829
In short: Romanian, Balkan Slavic and Albanian — three lexically different but grammatically identical languages — united by a Thraco-Illyrian substratum; his one concrete example: the postposed article.
via Friedman 2006; the German wording, in Friedman & Joseph 2025, ch. 2
The formulation that launched the Balkanist programme — after Leake's similar 1814 observations.
Miklosich, 1862
In short: the first explicit list of Balkanisms (1862: 6–8); the 'un-Romance' postposing of the article is item 4 — Bulgarian, Romanian, Albanian.
translated after Friedman & Joseph 2025, ch. 2
Sandfeld did not list it first: Sandfeld himself (1930: 12) points to Miklosich.
Sandfeld, 1930
In short: the canon — the postposed article opens, at no. 1, the list of specifically Balkan features (ch. 4).
Linguistique balkanique, via Friedman & Joseph 2025, ch. 2
The canonization: the tail becomes Balkanism number one.
Friedman & Joseph, 2025
In short: Leake's and Kopitar's recognition of the area 'stands up to scrutiny' — the postposed article is a Balkanism, and language contact 'was certainly involved'.
ch. 6, §6.1.2.2–§6.1.2.2.1.5
The modern verdict: the pattern is solid; the exact historical route stays disputed.
Friedman & Joseph 2025, ch. 2: history of Balkan linguistics (PDF) →
Where we stand
What we know and what we reconstruct
Four drawers for the whole page: the solid material, the mechanism, why it holds, and the hole in the middle.
What we know
The material comes from ille, grammaticalized in Late Latin; the fixing of the Romanian tail is preliterary. By 1521 the system is fully formed (Neacșu). Three branches share the tail — Romance, Albanian, Slavic; Greek does not.
What we reconstruct
The road of postposed ille to glued article: homo ille → omul, after Graur 1967 (via Cornilescu 2016).
Why it holds
The forms are alive: illum→-l, illae→-le, illorum→-lor. Three branches with the same tail, each from its own native material — shared inheritance is ruled out (F&J 2025, ch. 3).
What stays disputed
WHY it glued itself at the tail: calque on a native model, pure Latin, or mixed. We have no direct record of the decisive moment; the evidence is indirect, and its reading stays disputed.
Three stops
Three dates
Three stops, one storyline: the letter, the list, the canon.
Neacșu
1521
The Câmpulung letter: lucrul, ochii, locul cela strimtul — the article fully formed.
stop 1 of 3
Miklosich
1862
The first explicit list of Balkanisms; article postposing is item 4.
stop 2 of 3
Sandfeld
1930
The canon: postposed definite article, Balkanism number one.
stop 3 of 3
1780 Elementa at Vienna · 1829 Kopitar · 1967 Graur.
The road of the tail
ille
the Latin demonstrative
-l, -a, -i
the article glued at the tail
omul
the whole word
Why at the tail
Three answers, no winner
All three start from the same ille — and the article cannot be shared inheritance: Proto-Indo-European had no article. They fight over WHY it glued itself behind. The page's verdict: DISPUTED
The calque
Kopitar 1829 (via Friedman 2006): the tail copies a Thraco-Illyrian model. The potential exception is Drobeta: Hamp (1982, via Friedman & Joseph 2025) reconstructs the name as *druṷā-tā — a phrase with a postposed definitizing element, a highly suggestive piece of evidence, if the reconstruction is accepted.
Pure Latin
Graur 1967 (via Cornilescu 2016): the article comes out of Latin postposed ille (homo ille), by reanalysis. Renzi 1993 (in the same report) argues the rival hypothesis, with a preposed demonstrative.
The mix
Brâncuș 2004 (via Cornilescu 2016): the Latin postposed construction, reinforced by contact with Dacian idioms, which preferred postposing.
The consensus is stronger on the pattern than on its origin: the postposed article is treated as a Balkanism, and language contact was certainly involved (Friedman & Joseph 2025). Beyond the possible Drobeta exception, no appeal to substratum rests on any concrete evidence. Apart from that exception, the old Balkan languages are too weakly attested to demonstrate such morphosyntactic convergences.
Nor is 'independent drift' a school: the fixing of postposing right in the contact period, in three branches, is too much for mere coincidence (F&J 2025, ch. 3). Pre-Latin non-Hellenic is too meagerly attested for serious speculation beyond the lexicon (Friedman 2006).
Friedman & Joseph 2025, ch. 1: Drobeta and the dead languages →
Friedman & Joseph 2025, ch. 6: the verdict on the pattern →
Test yourself
Which is older?
Six landmarks from the page, shuffled. Put them from oldest to newest.
Move each card to its position, 1 to 6. Keys 1–6 move the chosen card; cards drag with a finger too.
- 1930
Sandfeld: the canon
Position - 1521
Neacșu: the article fully formed
Position - 1967
Graur: postposed ille
Position - 1780
The Vienna grammar (Elementa)
Position - 1862
Miklosich: the first list
Position - 1829
Kopitar: the substratum
Position
Score: 0 of 6
The right order, oldest first:
- 1521 — Neacșu: the article fully formed
- 1780 — The Vienna grammar (Elementa)
- 1829 — Kopitar: the substratum
- 1862 — Miklosich: the first list
- 1930 — Sandfeld: the canon
- 1967 — Graur: postposed ille
The settled row: every landmark in its place.
Sources
The Cambridge synthesis first, then the grammars, then the texts. Every risky claim points at its source.
The reference synthesis
- Friedman & Joseph, The Balkan Languages, Cambridge 2025 (open access)
- ch. 6: the postposed article, branch by branch
- ch. 3: why the pattern is no coincidence
- ch. 1: the dead languages, Drobeta
- ch. 2: history of Balkan linguistics (PDF)
- Introduction: the first list (PDF)
Grammars and linguists
- Ivancu 2019: the Romanian article after GALR (redalyc)
- Aldea 2004: the ille→article table, p. 24 (PDF)
- Cornilescu 2016: Graur 1967, Brâncuș 2004, Renzi 1993 (PDF)
- Friedman: the Balkans as a linguistic area (PDF)
- Friedman 2006: Kopitar, Leake, the state of knowledge (PDF)
The neighbours' exact forms (usage sources)
- Albanian: the postposed article paradigms (grammar summary)
- Bulgarian: the postposed article paradigms (grammar)
- Macedonian: the three articles (grammar summary)
The texts
Catalogue neighbours
Type: Fraunces and Source Serif 4, under the SIL Open Font Licence.