Loading the map…

The Romanian linguistic atlas, 1929–1938

Where you are from, by your words

Between 1929 and 1938, two linguists went through hundreds of villages, in what is now Romania, the Republic of Moldova and Ukraine, and wrote down what people called maize, a grandfather, a bucket. The quiz is in Romanian; this edition explains the maps and how far they can be trusted.

Try the quiz (25 questions, Romanian words)See how people spoke

On the map: the crop polenta is made from, as 61 villages named it between 1929 and 1938. Villages east of the Prut, the river on today's border with Moldova, are missing from this volume.

The quiz

Which word would you use?#

The answers are Romanian dialect words; pick the one you grew up with, not the textbook one. You may pick two. The map redraws after every answer and lights the villages that spoke like you; words said everywhere change it little or not at all.

The atlas

An atlas made on foot#

In 1919 Sextil Pușcariu, a linguist at the university of Cluj in Transylvania, founded a Museum of the Romanian Language and began an atlas of its dialects. Two young linguists, Sever Pop and Emil Petrovici, then travelled from place to place with different questionnaires: Pop with a short one at 301 points, asking one person at each; Petrovici with a long one at fewer points and with more people. By 1938 they had made 394 surveys at 387 points between them, with almost 7,000 questions. Most points are villages.

They asked indirectly ("What do you call the thing you draw water from the well with?") and wrote each answer exactly as they heard it, in phonetic script. Each map of the atlas gives, for one question, the answer of every village. The maps were printed between 1938 and 1972; the Sextil Pușcariu Institute in Cluj has put them online in full.

For this page 65 maps were read point by point: 7,147 answers. 7,355 of them, on 65 maps, were read twice, independently, and the two readings give the same word in 97% of cases; where they differed, the scan settled it.

An example

One crop, four names#

In the 1930s maize had four main names: cucuruz in the west (20 villages), mălai in the north (12), păpușoi in the east (11) and porumb, today's standard word, in the south (19). The colour shows the word with the largest estimated share around each place, nearer villages counting more. The fuller the colour, the more often neighbours agreed; it fades where surveyed villages are sparse.

One word says little: porumb was heard from the Danube to the Carpathians. Words add up, though: someone who says păpușoi (maize), barabule (potatoes) and harbuz (watermelon) pulls the map towards Moldavia, the historical region east of the Carpathians.

Bessarabia and northern Bukovina

Grandfather: bunic, moș, bunel#

Sever Pop's 301 points include Bessarabia (most of today's Republic of Moldova) and northern Bukovina (now in Ukraine). In many Bessarabian villages a grandfather was bunel or bătrân ("the old one"); moș was said in northern Bukovina, Transylvania and the Banat; bunic, today's standard word, mostly in Moldavia west of the Prut.

These questions about family, the body and burial have answers recorded at places now in the Republic of Moldova and Ukraine, which are missing from the volumes published between 1956 and 1972. For a reader from the Republic of Moldova they are the most direct evidence.

The test

How well the map finds a village#

A map that guesses should show its error. The test used the atlas villages themselves: a pair of neighbouring villages, one from each survey and at most 20 km apart, was removed from the calculation, and the map then tried to find it from the two villages' answers to the quiz questions alone (between 22 and 25; usually 25). Distance was measured to Petrovici's village.

Over 50 pairs, the place it pointed to fell a median 33 km from the village; in nine cases out of ten, within 61 km. The zone the map calls "80% sure" contained the village 80% of the time, the 50% zone 52%. The overall weight given to each survey series was tuned on half the pairs and checked on the other half.

With only the family, body and burial questions, the only ones with villages beyond the Prut, villages are found a median 47 km away; for the 45 points now in the Republic of Moldova and Ukraine, 77 km.

How far the indicated place fell, over 50 pairs of hidden villages (km from the village)

The promised zone and how often it held, on hidden villages. Line: what the zone promises; bar: how often it held.

Today

What is left of the old words#

The test on people alive today comes from Bukovina, in Romania's far north-east: the Audiovisual Linguistic Atlas of Bukovina (A. Philippide Institute, Iași) filmed two young and two older speakers in each of ten villages of Suceava county in 2016 and 2017, answering the survey's questions. The calculation here keeps the 28 speakers whom ALAB's ethnolinguistic field labels "Romanian", from 7 villages; the 12 labelled "Polish" or "Hutsul" are left out. There are 15 questions shared with the atlas.

Many old words are still there: barabule for potatoes comes up with 14 of 14 older and 12 of 13 young speakers (some also say the standard cartofi). Others come up less often in the young speakers' recordings: ștergar for a towel (14 of 14 older, 7 of 14 young), cocostârc for a stork, păpușoi for maize.

Recordings of 2016–2017: how many informants use the word that, from the atlas, the model estimates was most common around their village Upper bar: older; lower bar: young.

Placed on the 1930s map from their words alone, these speakers land a median 111 km from their village; the young 156 km, the older 94 km. Where the atlas shows a local word in the village's area, 26% of the answers kept were the standard word alone. When the map treats the standard word as weak evidence, with its weight measured on the other villages, the error falls to 69 km (young 72, older 67) and the marked zone holds them 96% of the time. That is still worse than for the old villages of the same corner with the same questions (46 km, 7 villages), and the 80% zone is typically larger (81,550 against 39,800 km², medians). The quiz uses this rule and has 25 questions, not 15.

This is one corner of the atlas area and 28 people: a sign, not a measurement for the whole atlas area. The clips were transcribed three times by machine; 374 of 420 answers were kept, those where at least two transcriptions gave the same word.

Limits

Where the map stops#

The Prut. In the 1930s Petrovici also surveyed villages between the Prut and the Dniester, in northern Bukovina and south of the Danube; the 1940 volume has them. The volumes printed between 1956 and 1972 show only villages inside the Romania of the day. On their maps Bessarabia is empty. For those questions the quiz carries the word of the west-bank villages across the Prut; tested on Pop's questions, which have villages on both banks, that rule named the word recorded at points now in the Republic of Moldova and Ukraine 60% of the time, against 47% for the most common word at the points in present-day Romania (494 answers).

The textbook word. In the 1930s porumb or cartofi were the words of one region. Today they are the words of the standard language and can be said anywhere. For a present-day reader the map counts the standard word as weak evidence, the weaker the more often people who have another word at home use it anyway. A reader who gives only standard words gets the honest answer: the map has nothing to place them by.

Towns. Most of the atlas's points are villages. How well people who grew up in towns are placed has not been measured; someone who speaks the standard language gives few clues.

Points left out. The atlas also has answers in Hungarian, Transylvanian Saxon German, Serbian, Bulgarian and Ukrainian. The points surveyed for those languages, and four of Pop's points with bilingual speakers, are left out entirely, Romanian answers included.

All maps

How people said it, question by question#

Pick a thing and see its words, village by village. Tap a village to read its answer.

Family, body, burial (Bessarabia included)

House, field, kitchen, animals (Bessarabia missing)

Questions#

What is the Romanian linguistic atlas?

A dialect atlas made in Cluj under Sextil Pușcariu: between 1929 and 1938 Sever Pop and Emil Petrovici made 394 surveys at 387 points. Each map gives every point's answer to one question.

How sure is the result?

On 1930s villages hidden from the model, the indicated place fell a median 33 km away and the 80% zone contained the village 80% of the time. In a separate 15-question test on 28 speakers recorded in Bukovina in 2016–2017, the median error was 69 km; see "Today".

Why does the map place me far from home?

Because the atlas describes the villages of the 1930s. Words move, with school, radio and people moving house. A result far from home can mean your speech comes from parents or grandparents from another region.

Do I need to speak Romanian?

For the quiz, yes: the answers are Romanian dialect words. The maps and the method are explained in English.

Are my answers saved?

No. Your answers and your place never leave your phone or computer. A link you send carries only the words you chose.

Method#

The maps: 65 maps of the Romanian linguistic atlas, 15 from Sever Pop's series (ALR I, 1938 and 1942) and 50 from Emil Petrovici's (ALR II, 1956–1966). The map shows 288 of Pop's points and 61 of Petrovici's; the points of the dialects south of the Danube, in the map insets, and the points of other languages are not placed. Maps were read by vision models, 65 of them twice, independently; answers were reduced to words (lexical types) with a list written before any fitting, not by pronunciation.

The villages: coordinates from GeoNames, compared with where the atlas maps print each village number (median offset from that position 3.2 km; this does not measure how exactly each village is placed); where no name matched, the printed position was used.

The model: for each question, the share of each word at a place is computed from the villages around it, whose weight falls off with distance (on a scale of 35 km for Pop and 56 km for Petrovici), pulled towards the word's share among all points kept for that question. The clues from the answers multiply; because they are not independent, the product is raised to a power chosen on half of the village pairs and checked on the other half.

For a present-day reader, the standard word weighs what the Bukovina test shows; on questions that test lacks, the standard word weighs nothing. "Something else" weighs nothing. The word list was extended after the Bukovina transcripts were read (forms like ciucălău, a maize cob, or blit, a form of blid, a plate) and then frozen.

Today's test: the audiovisual atlas of Bukovina (A. Philippide Institute, Iași; fieldwork 2016–2017) has short recordings of answers to the same questions. Each clip was transcribed three times by machine; only a word that at least two transcriptions agreed on was kept.