Romanian words in the dialects of Italy: a map of Dan Ungureanu’s “Istoria limbii române”

Romanian and the dialects of Italy descend from the same language, Latin, so many shared words are to be expected. A more precise place of origin can be argued only from shared, exclusive innovations: changes Romanian shares with one area and with no other. The German linguist Karl Brugmann stated the rule in 1884. A Latin word kept in two places and lost in a third shows nothing, however many such words are collected.

The author himself writes that he counts only innovations, and Rudolf Windisch’s review notes the same. The checks below, made on the linguistic atlas of Italy and southern Switzerland (AIS, compiled by Karl Jaberg and Jakob Jud from fieldwork in the 1920s) and on northern Italian texts of the 12th and 13th centuries, show how hard it is for any one word to meet that condition.

On Saenko’s list of 110 basic meanings (2015), Romanian shares a root with:

Aromanian, spoken in the Balkans, is Romanian’s closest relative. Piedmontese and Neapolitan score almost the same: 75–78 and 79. Matches Romanian has with one group only: two with the Iberian Peninsula, one with the Rhaeto-Romance area of the Alps, one with Sardinia, none with Piedmont or Liguria. The list includes no Lombard dialect and no French, so it cannot test the very region with the most words in the book.

Not all three words are from the vocabulary: “după” has no entry in it, and for “numai” and “vreun” the book itself gives, in other chapters, the forms outside northern Italy listed below. Villages are classified by rules on the root of the word, not by a dialectologist.

Forms of the “numai” type (Romanian for “only”, from Latin “non magis”) and the change of l to r occur together in 12 villages, nine of them in Ticino. Neighbouring villages are not independent evidence, so the 12 count as one piece of evidence, for one area. Ten of them have the form with d- (“doma”), whose link to “non magis” is assumed, not demonstrated. Chance alone would give about five shared villages, not 12. The same check keeps, as innovations worth following up, the change of l to r, the metathesis “tulbur” (11 villages), the feminine gender of “miere” (honey) and “fiere” (bile), and “gula” with the meaning “mouth”.

The atlas has a limit of its own: it shows what was left after more than six centuries in which Tuscan replaced local forms. The poet Bonvesin da la Riva, writing in Milan around 1280, has both “verun” (any) and “dra” (Tuscan “della”), with r for l; in the atlas of the 1920s the two features no longer occur together in any village.

In “Istoria limbii române” (Cartier, Chișinău, 2024), Dan Ungureanu argues that Romanian descends from a late Latin spoken after the year 350 in north-western Italy, in an area he bounds by Milan, Lugano, Grenoble and Genoa, from which its speakers would later have left for the Balkans. The usual account places that Latin in or near the Roman province of Dacia, in today’s Romania, or south of the Danube. As evidence the book offers the changes that, according to the author, Romanian shares with the dialects of that part of Italy: l turned into r (“soare”, sun; “moară”, mill); metatheses, sounds that have swapped places, as in “tulbure” (cloudy), from Latin “turbulus”, and “întreg” (whole), from “integru”; and new meanings such as “gură” (mouth), from Latin “gula”, throat.

In the preface the author reconstructs a sentence as it might have sounded in northern Italy between 600 and 700. It begins “Eo ao beut vin tulbure” (I have drunk cloudy wine); in Tuscan the same words would have been “Eo ho bevuto vino torbido”. It is the author’s reconstruction, not a surviving text.

Two reviews appeared in 2024. The Romance scholar Rudolf Windisch, in “Zeitschrift für Balkanologie” (60/2), summarises the book chapter by chapter and calls the comparison of Romanian forms with those of northern Italy and beyond the Alps “eine akzeptable Methode”, an acceptable method; he also notes that the book’s three overlaid maps are “kaum erkennbar”, barely discernible. Cezar Bălășoiu, of the University of Bucharest, in “Studii și cercetări lingvistice” (LXXV/1), concludes that the book’s method is “profund eronată”, deeply mistaken, because it starts from the premise that Dacia was completely abandoned at the end of the third century, and objects, among other things, that it never says which linguistic features the La Spezia–Rimini line rests on.

The objection goes beyond the geography of words. In the usual classification, which Bălășoiu recalls, the line separates Western Romance (Portuguese, Spanish, Catalan, French, Occitan, Rhaeto-Romance and the dialects of northern Italy) from Eastern Romance, which takes in the dialects of central and southern Italy, Dalmatian and Romanian. One of the features that define it is what happened to voiceless consonants between vowels: Latin “rota” (wheel) gave “roată” in Romanian and “ruota” in Italian, but “rueda” in Spanish and “roue” in French.

The thesis in full, with the author’s replies to his critics, is on the page about Dan Ungureanu and in the dossier of 165 claims. The atlas village by village, for 19 words: The dialects of Italy.

The book’s preface sets the Latin word, the Romanian word and a dialect form side by side. Below are 12 rows of that table, with the place the author gives. The dotted line on the map is the La Spezia–Rimini line, drawn here after the map in the book.

Two of these rows were checked in old texts. “Reo” is also literary Tuscan, so it is not confined to the north; and in Bonvesin “brosco” means a toad, where Romanian “broască” is a frog.

The ledger, that is the list of forms taken from the book, was extracted automatically from the manuscript in two independent passes: one by rules (the region label, then the form printed in italics), the other by a language model reading each paragraph. The first produced 2,165 rows, each with one region and one form, the second 2,249; they agreed on 1,808, including 55 rows where one pass also put the form under an extra label. The 820 cases on which the passes disagreed were checked against the text in a third pass: 485 accepted, 335 rejected. Every form was found letter for letter in its paragraph. The ledger has 2,269 rows; one accepted case can give several rows, when the same form stands under two labels. Apart from eight rows (four keyed by hand from the preface table and four added in the third pass), no row entered the ledger on the strength of a single pass. None of the passes was done by a dialectologist.

In an independent check of 40 words, 374 rows were compared with the book’s text: 15 were wrong, most of them with the place attached to the neighbouring form, and no form the book clearly gives was missing. Since that check, a place receives a point only if it is printed right beside the form. Errors certainly remain. No paragraph of the chapter was left unattached to an entry.

On the map, 185 places have a point. A town or village received a point only if the GeoNames list or the list of AIS atlas villages has exactly one place of that name in the region the book indicates; the other names stay in the ledger without a point. Valleys, old provinces and departments have their point at their centre, and old texts in their author’s town, from tables drawn up by hand. The Ladin area on the map is the outline drawn around the places the book itself calls Ladin.

The chapter refers 64 times to numbered maps of the AIS atlas. Eight references were set aside because the title given in the book is missing or does not match the title of the map with that number. For the rest, villages and forms come from the atlas’s digital edition.

The author’s text, his examples from old texts and the book’s images are not reproduced here. The page gives only the data of each entry: the word, the Latin word cited, the region, the place and the form. The book’s synthesis maps are by Paul Șteț and Ligia Pop; the one that overlays phonetics, morphology and vocabulary was drawn by Ligia Cornelia Pop under the supervision of Professor Alina Satmari.