Dan UngureanuRomână

The Dan Ungureanu dossier

Everything Dan Ungureanu has argued, claim by claim

The main page tells the story briefly. Here is all the material: every claim with its year, source and page, beside the criticism it drew and our own checks. Search for a word or pick a topic.

  • He says what Dan Ungureanu argues
  • Critic what other scholars objected
  • His reply how he answered
  • Check what the data show, checked by us

When did Romanian break away from Latin?

Ungureanu places the split much later than usually thought: after the year 500.

  1. He says Masticare appears in the relevant sense only after about 400 – one more argument for a late date.The Chronology of the Latin Vocabulary in Romanian
  2. He says In medieval Latin the names of the cardinal points become locally unstable; the north in particular gains many new names after about 700.The Names of the Cardinal Points in Late Latin
  3. 2017
    He says An earlier formulation had Romanian starting to break away after about 300; later work pushes the split later.The Earliest Dialectalization of Latin (2017)
  4. 2023
    He says Late and Christian Latin words, such as dominica ("Sunday") and paganus in its Christian sense, show that Romanian comes from an already evolved Latin.Istoria limbii române, precuvântare, Revista literară (2023), p. 24
  5. 2024
    He says The Reichenau glosses explain hard Latin words with words readers knew. Indirectly, they show what Latin was spoken at the time.Glosarul din Reichenau și limba română (2024), pp. 105–106
  6. 2024
    He says Of the Latin synonym pairs explained in the glosses, Romanian usually keeps or loses the same word as the glossators' spoken Latin.Glosarul din Reichenau și limba română (2024), pp. 105–115
  7. 2024
    He says Of 66 pairs, Romanian has the late word in 49 and the classical one in 17 – so it developed like late Latin in about three quarters of cases.Glosarul din Reichenau și limba română (2024), pp. 112–114
  8. 2024
    He says From several word-replacement scenarios he estimates the proto-Romanian split at around 540–600, give or take a few decades.Glosarul din Reichenau și limba română (2024), pp. 113–115
  9. 2024
    He says If the Latin of Dacia had been cut off before 250, there would have to be an unattested spoken Latin anticipating by centuries changes documented much later.Glosarul din Reichenau și limba română (2024), p. 115
  10. 2024
    He says The big shifts in Romance grammar – cases, plurals, verb endings – happened around 500–600, not in the second century.Glosarul din Reichenau și limba română (2024), pp. 115–116
  11. 2024
    He says The shift from "ke/ki" to "ce/ci", important for Romanian, is no older than about 500.Glosarul din Reichenau și limba română (2024), pp. 115–116
  12. 2025
    He says In 2025 he continues dating words and meanings such as cognatus, sera, collocare, hostis, fossatum, caballus and tornare.Colocviu, Academia Română (2025)
  13. 2024
    His reply Answering the 2024 SCL review, he states that his model requires sixth-century Latin, after 500, not fourth-century.O replică la recenzia lui Cezar Bălășoiu (2024)
  14. 2024
    His reply Eastern and Western Romance plurals parted after the Latin case system broke down, that is after 500, perhaps towards 600.O replică la recenzia lui Cezar Bălășoiu (2024)
  15. 2024
    Check If each pair were a coin toss, 49 out of 66 would be very hard to get by chance. The conclusion depends on those assumptions, though.Glosarul din Reichenau și limba română (2024), pp. 113–114
  16. 2024
    Check The article does not show why 50/50 is the right baseline. If the baseline rises to about 64%, 49 out of 66 is no longer significant.Glosarul din Reichenau și limba română (2024), p. 114
  17. 2024
    Check The four published dates cannot be reproduced consistently by one linear calculation, either with 49/66 or with the two-thirds approximation.Glosarul din Reichenau și limba română (2024), p. 114
  18. 2024
    Check The 66 pairs were chosen from a much larger, mixed body of glosses; a clear, repeatable selection rule is needed.Glosarul din Reichenau și limba română (2024), p. 107
  19. 2024
    Check In a provisional check on eight Romance languages, 205 of 231 cases follow the late word. But the languages are not independent observations.Glosarul din Reichenau și limba română (2024)
  20. 2024
    Check At least three pairs are not either-or in Romanian, since words from both survive: oaie/berbec, înjunghia/ucide, agru/câmp.Glosarul din Reichenau și limba română (2024)
  21. 2024
    Check Across the other Romance languages, pairs of languages agree in about 88% of cases (median; range 77–100%), so they cannot be counted as independent witnesses.Glosarul din Reichenau și limba română (2024)
  22. 2024
    Check Even grouped into five broad branches, Romance languages keep the late word in about 88% of cases. This is a provisional pointer, not a confirmed baseline.Glosarul din Reichenau și limba română (2024)
  23. 2024
    Check The 2024 dating (after 500) is in tension with the fourth-century split of 2016. The difference between a regional dialect and a final split needs explaining.Glosarul din Reichenau și limba română (2024)
  24. 2024
    Check Late Latin innovations can show how long Romanian shared a common history, but not where the second-century colonists came from.Glosarul din Reichenau și limba română (2024)

Which Latin did Romanian come from?

His thesis: from a Latin spoken in northern Italy, later carried towards the Balkans.

  1. He says The area of closest affinity is Liguria – western Lombardy – Ladinia, together with related Alpine regions.Romanian and the Italian dialects II
  2. He says Latin recens meant "fresh, new"; the sense "cold" appears later, in the northern Romance world.Rece. Etimologia unui cuvînt
  3. He says He argues that the shift from 'fresh' to 'cold' happened in northern Italy and is shared with Romanian.Rece. Etimologia unui cuvînt
  4. 2016
    He says Romanian comes from a spoken Latin that already differed from region to region.The Geographic Origin of Romanian (2016), p. 125
  5. 2016
    He says Most of Romanian's affinities cluster in a small area: western Lombardy and the Arpitan and Vivaro-Alpine lands.The Geographic Origin of Romanian (2016), pp. 125, 130–133
  6. 2016
    He says The geographical argument rests on several kinds of innovation converging: in sounds, grammar, word formation and vocabulary.The Geographic Origin of Romanian (2016), p. 125
  7. 2016
    He says The comparison with northern Italy also covers the spread of prefixes like des-/dis-, not just lists of look-alike words.The Geographic Origin of Romanian (2016), p. 130
  8. 2016
    He says Romanian comes from a northern Italian Latin spoken in the second and third centuries, with innovations absent from Classical Latin and Tuscan.The Geographic Origin of Romanian (2016), p. 130
  9. 2017
    He says Romanian comes from a Latin dialect of western Cisalpine Gaul (north-west Italy), where the sound rules, words and forms leading to it are already found.The Earliest Dialectalization of Latin (2017)
  10. 2017
    He says He tries to measure the sound changes shared by Romanian and Western Lombard using Levenshtein distance.The Earliest Dialectalization of Latin (2017)
  11. 2017
    He says The sound-based arguments for a Celtic substrate are not specific enough; vocabulary weighs more.The Earliest Dialectalization of Latin (2017)
  12. 2023
    He says Two things must be kept apart: the source language, an already regional northern Italian Latin, and its speakers' later journey east.Istoria limbii române, precuvântare, Revista literară (2023), p. 24
  13. 2023
    He says Almost all the sound and grammar changes specific to Romanian are already found in the dialectal Latin of northern Italy.Istoria limbii române, precuvântare, Revista literară (2023), p. 24
  14. 2025
    He says His 2025 summary for the public: Romanian has Latin vocabulary from after 500 and many elements traceable to northern Italy.Gazeta de Sud, Craiova (2025)
  15. 2018
    Critic Ioana Vintilă-Rădulescu (2018) finds that the 2016 volume does not convincingly prove the northern Italian origin.Ioana Vintilă-Rădulescu, recenzie (2018), p. 157
  16. 2018
    Critic Ioana Vintilă-Rădulescu's main objection: there is no comparison showing the features cited occur only in northern Italy or are unusually dense there.Ioana Vintilă-Rădulescu, recenzie (2018), p. 157
  17. 2018
    Critic Vintilă-Rădulescu criticises the use of Levenshtein distances without a systematic comparison with other Italian and Romance varieties.Ioana Vintilă-Rădulescu, recenzie (2018), pp. 128–129
  18. 2024
    Critic Rudolf Windisch (2024) is far more favourable: he considers the approach an acceptable method for rethinking the history of Romanian on firm grounds.Rudolf Windisch, recenzie (2024), p. 286
  19. 2024
    Critic Windisch sums up the model as Latin from Italy plus the later history of a Romanised population along the Via Egnatia.Rudolf Windisch, recenzie (2024), p. 274
  20. 2024
    Critic Dan Alexe criticises the word-and-geography argument for not taking enough account of the grammar Romanian shares with Balkan languages.Dan Alexe, Cabal in Kabul (2024), 13 Apr 2024
  21. Check The case of rece separates a change of meaning from the inherited Latin form, but alone does not show a northern Italian source without comparing the whole Romance world.Rece. Etimologia unui cuvînt
  22. 2013
    Check Historical sources show colonists of very diverse origins in Roman Dacia, not a population demonstrably from western Lombardy.Bryn Mawr Classical Review (2013)
  23. 2016
    Check The geographical argument has several steps: the feature is real, an innovation, has a spread, is independent, has a date, a migration, an area. Passing the first steps does not guarantee the conclusion.The Geographic Origin of Romanian (2016)
  24. 2016
    Check Several words altered by the same sound change are not separate evidence of migration. Features must be counted by historical change, not by word.The Geographic Origin of Romanian (2016)
  25. 2018
    Check The northern Italian model with the Via Egnatia cannot be presented as settled fact: there is one detailed critical review and a later favourable one.Ioana Vintilă-Rădulescu, recenzie (2018)
  26. 2018
    Check The incomplete-comparison objection can be tested: an inventory of the whole Romance world would show how exclusive and how dense the northern Italian likenesses are.Ioana Vintilă-Rădulescu, recenzie (2018), pp. 157–158
  27. 2024
    Check Windisch's favourable review is a genuine scholarly response; it stands beside the negative critiques without amounting to consensus.Rudolf Windisch, recenzie (2024), p. 286
  28. 2024
    Check The model shifts between 2016 and 2024: from a fourth-century split and a precise colonist source to a formation south of the Danube, on the Via Egnatia. The northern Italian source remains.Istoria limbii române (2024), recenzia din 2018, pp. 110–111; recenzia SCL 2024, p. 127
  29. 2025
    Check A recent survey of Balkan languages places eastern Balkan Romance over a wide area; continuity north or south of the Danube remains disputed for lack of documents.The Balkan Languages, Cambridge (2025)
  30. 2025
    Check A classification on 145 features of today's Romance dialects cannot alone prove an ancient origin: innovations after the split say nothing about the old source.Isogloss (2025)

Do we have Dacian words?

Words inherited from before Latin are few and hard to pin on the Dacians.

  1. He says Communist nationalism shaped which explanations of the substrate were considered acceptable.Substratul dacic, ideologia comunistă și metodologia lingvistică
  2. He says A Romanian–Albanian likeness alone does not tell substrate from loan, wandering word or coincidence. Aromanian can serve as a third witness.Common Lexic in Romanian and Albanian, p. 1
  3. He says In hard cases, when Aromanian cannot decide, he accepts coincidence and keeps the traditional etymologies.Common Lexic in Romanian and Albanian, p. 1
  4. 2015
    He says Substrate words are not scattered at random: they cluster in the same few areas of meaning.Loanwords and Substrate in Romance Languages (2015), pp. 133–143
  5. 2015
    He says Some meanings stay tied to a place even when its people switch language; that is why some old words survive more often.Loanwords and Substrate in Romance Languages (2015), p. 143
  6. 2015
    He says Other meanings are easily replaced and keep acquiring synonyms or loanwords.Loanwords and Substrate in Romance Languages (2015), pp. 142–143
  7. 2015
    He says Verbs for unintended perception are often stable; those for deliberate perception, like "to look", are replaced more easily.Loanwords and Substrate in Romance Languages (2015), p. 143
  8. 2015
    He says French and Romanian borrowed in similar areas – feudal, social, emotional – despite different histories of contact.Loanwords and Substrate in Romance Languages (2015), pp. 142–143
  9. 2015
    He says Some substrate etymologies used in the comparison are conjectures, not generally accepted, or could be old loanwords.Loanwords and Substrate in Romance Languages (2015), p. 143
  10. 2017
    He says About a third of the words counted as substrate have parallels in Italy or France, which weakens the idea that they are specifically Dacian.Cuvinte de substrat? Da. Cuvinte dacice? Nu. (2017)
  11. 2019
    He says A Romanian word resembling an Albanian one does not automatically prove it was inherited from the Dacians, Thracians or another old Balkan people.Cît din substratul lexical al limbii române e dacic? (2019), p. 362
  12. 2019
    He says A good share of the words usually put down to the substrate might have Romance, Celtic, Alpine or Italic links.Cît din substratul lexical al limbii române e dacic? (2019), p. 362
  13. 2019
    He says When the source language cannot be proven, it is more honest to say "substrate" than "Dacian".Cît din substratul lexical al limbii române e dacic? (2019), p. 362
  14. 2019
    He says Among the words analysed he finds no certain Dacian word, with the possible, and debatable, exception of brusture ("burdock").Cît din substratul lexical al limbii române e dacic? (2019), p. 362
  15. 2019
    He says The vocabulary evidence offered so far does not convincingly prove that Dacian influenced the Latin that became Romanian.Cît din substratul lexical al limbii române e dacic? (2019), p. 362
  16. 2019
    He says Contact between Latin and Dacian may be historically plausible without being provable through the etymologies cited.Cît din substratul lexical al limbii române e dacic? (2019), p. 362
  17. 2019
    He says Many words often given as Dacian are not Dacian: among them brad (fir), buză (lip), gușă (goitre), mânz (foal), fluier (flute) and târg (market).Cît din substratul lexical al limbii române e dacic? (2019), p. 362
  18. 2019
    Check The Romance, Celtic, Germanic or Italic explanations he offers must pass the same tests of sound, meaning, date and contact as the Dacian ones he criticises.Cît din substratul lexical al limbii române e dacic? (2019)
  19. 2019
    Check Even brusture, the only possible Dacian candidate he allows, is not listed as certainly Dacian in Romanian etymological dictionaries.Cît din substratul lexical al limbii române e dacic? (2019), pp. 3
  20. 2019
    Check The 2019 article leaves many questions open; turning all its comparisons into certain etymologies would distort it.Cît din substratul lexical al limbii române e dacic? (2019), pp. 1–13
  21. 2019
    Check For îngâna ("to mimic"), his Romance explanation matches the etymological dictionary DER, which proposes Latin gannare/*ingannare, with descendants in other Romance languages.Cît din substratul lexical al limbii române e dacic? (2019), p. 7
  22. 2019
    Check For țap ("billy goat"), his proposed Italic origin clashes with Romanian dictionaries that call it Thraco-Dacian and compare it with Albanian. The question stays open.Cît din substratul lexical al limbii române e dacic? (2019), p. 12
  23. 2019
    Check His explanation of traistă ("shoulder bag") through medieval Greek is backed by DER, which points to tagistron and related Greek and Calabrian forms.Cît din substratul lexical al limbii române e dacic? (2019), p. 12

When is it kinship, and when is it chance?

Any two languages have look-alike words; the question is how many would turn up by pure chance.

  1. He says A basic-vocabulary list should be chosen by stability measured across several language families, not copied mechanically from an established list.A list of 100 best kept words across linguistic families
  2. He says Expressive words with unstable meanings leave plenty of room for comparison; likenesses between dialects there may be accidental.Common Lexic in Romanian and Albanian, p. 1
  3. 2010
    He says When comparing very distant languages you must estimate explicitly how often likenesses arise by chance. Otherwise ancient kinship and coincidence cannot be told apart.Long-range comparisons and word roots decay (2010), p. 123
  4. 2010
    He says If a very ancient kinship is real, the surviving likenesses should cluster in the meanings that change least, not be scattered across the vocabulary.Long-range comparisons and word roots decay (2010), pp. 123–130
  5. 2010
    He says Roots used in distant comparisons should pass strict filters: close sounds, a precise meaning, rare replacement and wide attestation across the reconstructed family.Long-range comparisons and word roots decay (2010), pp. 129–135
  6. 2010
    He says Only about 2–3% of reconstructed roots might be robust enough for very distant comparisons.Long-range comparisons and word roots decay (2010), pp. 129–135
  7. 2010
    He says The pronoun roots *me- and *ne- are stable, precise in meaning and rarely borrowed.Long-range comparisons and word roots decay (2010), p. 136
  8. 2010
    He says Comparisons between very distant languages can become testable hypotheses, not just collections of look-alike words.Long-range comparisons and word roots decay (2010), p. 123
  9. 2010
    Check His method keeps returning to alternative explanations: set a baseline, look for evidence that separates hypotheses, and drop labels more precise than the data allow.Long-range comparisons and word roots decay (2010)
  10. 2010
    Check If his estimate for the pronoun "we" is read as 15 matches out of 50, with a 10% chance rate, such a result would arise by luck about once in 13,500 tries.Long-range comparisons and word roots decay (2010), p. 131
  11. 2010
    Check The 10% chance rate is not calibrated on pronouns. If the true rate were about 19.5%, the result would no longer be significant.Long-range comparisons and word roots decay (2010), p. 131
  12. 2010
    Check The idea that pronouns are never borrowed is too absolute: studies of language contact document borrowed pronouns and even whole pronoun sets.Long-range comparisons and word roots decay (2010), p. 136
  13. 2010
    Check The sample for "we" mixes overlapping families and subgroups, inclusive and exclusive forms, possessives and person endings. These are not independent observations.Long-range comparisons and word roots decay (2010), pp. 132–136
  14. 2010
    Check The strongest idea of the 2010 article is that likenesses should cluster in meanings shown separately to be stable. It still lacks an explicit model of chance.Long-range comparisons and word roots decay (2010), pp. 123–125
  15. 2010
    Check The rule that few, scattered likenesses prove there is no kinship is too categorical: a lack of evidence does not prove kinship is absent.Long-range comparisons and word roots decay (2010), pp. 131–132
  16. 2018
    Check A more rigorous version would test recurring sound correspondences against random shuffles, instead of scoring look-alike words one by one.Journal of Language Evolution (2018)
  17. 2026
    Check Some consonants, nasals among them, are common in pronouns worldwide and can spread regionally without any distant kinship.WALS, capitolul 136

Is there an Altaic family?

Ungureanu moved from accepting a very ancient kinship to rejecting the Altaic family.

  1. He says Comparing basic vocabulary, Turkic, Mongolian and Manchu do not form a single Altaic family.There is no Altaic linguistic family
  2. Check Rejecting the Altaic family contradicts his 2010 conclusion, which favoured an Indo-European–Uralic–Altaic kinship. It is a change of mind, not a consensus of the field.There is no Altaic linguistic family, p. 130
  3. Check At a 1% chance rate and independent words, 22 matches out of 600 are too many for chance. The arithmetic does not support calling the number trivial.There is no Altaic linguistic family
  4. Check Despite its categorical title, the Altaic paper does not claim to rule out all kinship; it leaves a very ancient relation open.There is no Altaic linguistic family
  5. Check His criticism holds as a mechanism: the more synonyms and near meanings you accept, the more chances for accidental matches.There is no Altaic linguistic family
  6. 2011
    Check If the chance rate is 2.5%, 22 matches out of 600 are borderline significant; at 3%, they can be pure chance. The chosen baseline changes the conclusion.Relațiile lexicale… (2011), verificat cu: Recenzie în Revue roumaine de linguistique (2013), pp. 76–77
  7. 2021
    Check Recent statistical studies of Altaic ("Transeurasian") reach no common verdict. Some find signals, but language contact remains a possible explanation.Diachronica (2021)
  8. 2021
    Check A 2021 Nature study argues for a shared Transeurasian ancestry, using 3,193 cognate sets, 254 concepts and 98 languages. The question remains disputed.Robbeets et al., Nature (2021)
  9. 2022
    Check A 2022 preprint by Tian and colleagues challenges the linguistic, genetic and archaeological basis of the 2021 study. It is a critique, not a collective verdict.Tian et al., bioRxiv (2022)

How reliable are reconstructed languages?

Two teams reconstructing the same parent language can reach very different results.

  1. He says Published comparative dictionaries should themselves be checked statistically for reliability.Nilo-Saharan: a statistical assessment
  2. 2014
    He says The unwritten methods of five etymological dictionaries can be spelled out as procedures and applied to other data, so they can be checked.Reconstructing Languages (2014)
  3. 2014
    He says Between two randomly chosen languages, about 1–4% vocabulary matches appear as pure noise, and more if looser meanings are accepted.Reconstructing Languages (2014)
  4. 2014
    He says Reconstructing languages is exposed to mental traps: seeing patterns that are not there and choosing the target after the result.Reconstructing Languages (2014)
  5. 2024
    Critic Alexandru Nicolae criticises reconstructions that cannot be checked by the comparative-historical method and the excessive precision given to undocumented periods.Interviu cu Alexandru Nicolae, HotNews (2024)
  6. 2024
    His reply In "Istodor", Ungureanu defends an approximate reconstruction of Proto-Romanian but admits that very exact reconstructions of an undocumented period risk becoming imaginary.Istodor (2024)
  7. Check The Nilo-Saharan paper points to a real problem: short roots and loose meanings offer many chances for accidental likenesses.Nilo-Saharan: a statistical assessment
  8. Check With a 1-in-60 chance, the exact probability of at least one match in 13 tries is about 19.6%, not the 21.7% from simple multiplication.Nilo-Saharan: a statistical assessment
  9. Check The strongest contribution is comparing independent reconstructions: very different results for the same family show how much the reconstructed ancestor depends on method.Nilo-Saharan: a statistical assessment
  10. Check The criticism of picking rules after seeing results can become a test: fix the rules for meaning shifts in advance and check them on held-out data.Nilo-Saharan: a statistical assessment
  11. 2026
    Check Roger Blench (2026) describes large differences between Bender's and Ehret's Nilo-Saharan reconstructions, including family structure and the number of reconstructed forms.Roger Blench, Arguments for the Coherence of Nilo-Saharan (2026), pp. 202–204
  12. 2026
    Check The Bender–Ehret disagreement does not by itself prove there is no Nilo-Saharan family. Blench defends it in 2026 with other kinds of evidence too.Roger Blench, Arguments for the Coherence of Nilo-Saharan (2026)

What does the document actually say?

Before interpreting, check the paper, the watermark and the reading of the manuscript.

  1. He says The widely shared poster about the sale of Roma slaves is a forgery, not a genuine notice.A forged poster about the sale of Gypsy slaves
  2. He says On typological grounds he argues that the coins attributed to Sponsianus are fakes. That is his position, not a settled verdict.The Sponsianus coins is still false
  3. He says Rumeni, read as the first record of the name "Romanians", is a modern transcription error: the manuscript would read cumeni, i.e. Cumans.Prima atestare a românilor sub numele de români
  4. He says The context supports cumeni: the passage would describe two peoples and pagans, and the parallel copies keep no stable rumeni.Prima atestare a românilor sub numele de români
  5. 2003
    He says The Rohonczi Codex is not from the 11th–12th centuries: the paper's watermark is northern Italian, about 1529–1540.„Nu trageți în ambulanță”, Observator cultural (2003), 6 May 2003
  6. 2003
    He says The manuscript is missing from catalogues until its donation in 1838, a further problem about its provenance.„Nu trageți în ambulanță”, Observator cultural (2003)
  7. 2015
    He says About thirty Romanian river names may be Cuman or Old Turkic in origin; the argument rests on meaning, not just on sound.Turkic Hydronyms in Romania (2015)
  8. 2024
    He says A map depicting an imagined past can be mistaken for a document from that past – a classic error in reading sources.History as gesunkenes Kulturgut (2024)
  9. Check A recurring thread in his work: historical objects with political or emotional pull are checked technically first, interpreted second.A forged poster about the sale of Gypsy slaves

What changes a society?

His essays on mentalities, schools, the church and technology.

  1. 2017
    He says Trust polls on institutions say more as a trend over time than as a ranking of very different institutions.Biserica ortodoxă română: trei decenii de statistici, Vatra (2017)
  2. 2017
    He says Institutional change is better tracked through long-run statistics than through the flow of news.Biserica ortodoxă română: trei decenii de statistici, Vatra (2017)
  3. 2021
    He says Roads, infrastructure and the media change mentalities even when people stay put.Marginalia la Florian Znaniecki, Vatra (2021), 1 Sep 2021
  4. 2022
    He says Cultural magazines stay useful for finding and selecting what is worth reading in a fragmented market, even after moving online.Internetul nu iubește literatura, Vatra (2022), 24 Feb 2022
  5. 2022
    He says Strictly literary magazines should open up to wider social and cultural debate.Internetul nu iubește literatura, Vatra (2022), 24 Feb 2022
  6. 2023
    He says Social mobility through school went hand in hand with urbanisation and changing occupations.Învățămîntul ca ascensor social, CriticAtac (2023), 24 Apr 2023
  7. 2023
    He says After 1990, emigration partly replaced the move to the city, weakening the link between school, city and social rise.Învățămîntul ca ascensor social, CriticAtac (2023), 24 Apr 2023
  8. 2024
    He says He explains Dacomania partly through the contrast between written and oral culture.History as gesunkenes Kulturgut (2024)
  9. 2026
    He says Technology changes how things are produced and, with it, the structure of society.Tehnologii ale emancipării, Vatra (2026), 16 Jul 2026
  10. 2026
    He says Technology can first reduce the need for skills and then demand new ones, such as computer-aided design and machining.Tehnologii ale emancipării, Vatra (2026), 16 Jul 2026
  11. 2008
    Check The weakness of Zidul de aer (2008): "mentality" is treated as invisible and very stable, but many conclusions lack measurable indicators or conditions under which they could be refuted.Zidul de aer (2008)
  12. 2008
    Check The Romanian individualism scores cited from Trompenaars vary – 81, 52 and 64 – by question. There is no single stable score for the whole population.Zidul de aer (2008), verificat cu: ASE București, studiu despre Trompenaars (2013), p. 97
  13. 2008
    Check The label "particularist" fits poorly with the Trompenaars results cited: 88, 68 and 44 on three universalism tests.Zidul de aer (2008), verificat cu: ASE București, studiu despre Trompenaars (2013), pp. 97–98
  14. 2008
    Check Different cultural models describe Romania differently; Hofstede and Trompenaars scores are not interchangeable measurements.Zidul de aer (2008), verificat cu: Cross-cultural management, ResearchGate
  15. 2014
    Check A national average does not automatically describe a person or a firm. Jumping between levels risks the ecological fallacy.Organization Studies (2014)
  16. 2023
    Check The 2023 essay on schools asks that reforms be evaluated after they are applied: did they work, and what unintended effects did they have?Învățămîntul ca ascensor social, CriticAtac (2023)
  17. 2024
    Check The 2024 article shows he has not dropped "mentalities": the oral/written contrast remains beside explanations through mechanisms, indicators and feedback.History as gesunkenes Kulturgut (2024)
  18. 2026
    Check The intuition that trust in some institutions is low fits the OECD data. The data do not, however, prove an explanation by an unchanging mentality.OECD Trust Survey
  19. 2026
    Check In an April 2026 poll, Romanians back both self-reliance and the common good, both freedom and order. Overall percentages do not describe each respondent.Barometrul INSCOP, aprilie 2026
  20. 2026
    Check His later texts lean more on infrastructure, urbanisation, migration, incentives and feedback loops. These add to the talk of mentalities rather than replacing it.Marginalia la Florian Znaniecki, Vatra (2021)
  21. 2026
    Check His diagnosis of rural school disadvantage is broadly backed by OECD data; that does not automatically confirm all his explanations.OECD Economic Surveys: Romania
  22. 2026
    Check The 2026 technology essay warns that rules built on proxy indicators can produce the very pattern they then use as proof.Tehnologii ale emancipării, Vatra (2026)

How does he answer his critics?

His 2024 replies ask critics for precise examples.

  1. 2024
    His reply He asks critics to point to concrete data they think wrong, not just general complaints about method or competence.Peer review și mentalitatea dîmbovițeană, Contributors (2024), 19 Nov 2024
  2. 2024
    His reply He answers bibliography criticisms by saying part of the Romanian tradition of language history took shape under communist ideological constraints.Peer review și mentalitatea dîmbovițeană, Contributors (2024)
  3. 2024
    His reply In "Istodor" he asks for checkable points: dated words, inscriptions, atlas forms, exact pages.Istodor (2024)

Two ideas, in depth

How do you check whether a reform worked?

Outside linguistics, Ungureanu asks the same rigour of institutions. In a 2023 essay about schools he proposes a simple loop:

  1. Name what is wrongWhich concrete problem do you want to solve? For example: village children finish school with far fewer chances.
  2. Change the ruleThe reform itself: a law, a programme, a new exam.
  3. Check afterwardsA few years on, go back to the problem in step 1. Is it solved? What unwanted effects did the change have?

↺ If not, back to step 1.

In the same essay he explains why school no longer lifts people socially as it used to: after 1990, emigration partly replaced moving to the city. Recent OECD data broadly support his diagnosis of the disadvantage of village schools. Source: "Învățămîntul ca ascensor social – experiența unui profesor", CriticAtac, 24 April 2023.

Two reconstructions of the same language family

Ungureanu asks that linguistic reconstructions be checked like any scientific result. His example: the Nilo-Saharan family of Africa, reconstructed by Lionel Bender and by Christopher Ehret. They worked on largely the same material and ended up in very different places:

AspectBender (1997–2000)Ehret (2001)
Which languages belongabout 12 branches, including Kadu (Chabu later dropped)without Kadu or Chabu
Shape of the treecautious, with parallel branchesstrictly in pairs
Reconstructed words173 for Proto-Nilo-Saharanmore than 1,600
Sound correspondenceslittle usedused extensively
Justifying the branchesat lengthmissing
Earlier workdiscussed at lengthselective
Meanings allowed in comparisonsbroad, with short rootsvery broad, with short roots

A nearly tenfold difference in the number of words does not show which reconstruction is right. It shows that the method leaves a lot of room for choice, and Ungureanu wants that room measured. Short roots and broad meanings are exactly the conditions in which chance matches appear. The comparison: Roger Blench (2026); last row from Ungureanu's critique.

What he wrote, and when

The books, papers and essays used in the dossier, in order.

  1. 1999
    Originile grecești ale culturii europene carte

    A book on the Greek origins of European culture.

  2. 2002
    Originea limbajului și primul om carte

    A book on the origin of language and the first humans.

  3. 2008
    Zidul de aer. Tratat despre mentalități carte

    A treatise on mentalities: how institutions' written rules square with how people actually behave.

  4. 2010
    Long-range comparisons and word roots decay articol

    Asks explicitly how many likenesses between languages would arise by chance, and which words last longest.

  5. 2011
    Relațiile lexicale între indo-europeană și familiile uralică și altaică carte

    A book on the Nostratic hypothesis, with stricter filters for the roots compared.

  6. 2014
    Reconstructing Languages carte

    A book that puts the methods of reconstructing languages themselves to the test, looking for noise and mental traps.

  7. 2015
    Loanwords and Substrate in Romance Languages articol

    Compares substrate words and loanwords across Romance languages: some meanings hold, others are easily replaced.

  8. 2016
    Româna și dialectele italiene; The Geographic Origin of Romanian carte

    His central thesis appears: Romanian comes from a Latin of northern Italy.

  9. 2017
    The Earliest Dialectalization of Latin; Cuvinte de substrat? Da. Cuvinte dacice? Nu. articol

    Argues that Latin had split into dialects early, and that a substrate word is not automatically Dacian.

  10. 2019
    Cît din substratul lexical al limbii române e dacic? articol

    Goes word by word through the vocabulary considered Dacian and finds almost no certain Dacian word.

  11. 2021
    Marginalia la Florian Znaniecki eseu

    An essay on how roads, the press and television change the way people think, even when they stay put.

  12. 2023
    Precuvântarea la „Istoria limbii române” și eseul despre școală eseu

    Separates the northern Italian source language from the later journey to the Balkans; separately, asks that school reforms be judged after they are applied.

  13. 2024
    Istoria limbii române carte

    The mature version of the thesis: Latin from northern Italy, then formed in the Balkans along the Via Egnatia. Critics are divided.

  14. 2024
    Glosarul din Reichenau și limba română articol

    Counts words in medieval glosses and proposes around 540–600 for Romanian's split, a rough estimate from his own model.

  15. 2024
    Replicile către critici articol

    States that his model requires Latin from after 500 and admits that undocumented periods cannot be reconstructed exactly.

  16. 2024
    History as gesunkenes Kulturgut articol

    On pseudo-history and Dacomania, explained through the contrast between written and oral culture.

  17. 2025
    Când și unde apare limba română? conferință

    A lecture in Craiova linking the post-500 date to the northern Italian traces in the vocabulary.

  18. 2026
    Tehnologii ale emancipării. Tehnologii ale reacțiunii eseu

    An essay on how technology changes work and society, and on rules that produce their own evidence.

← Back to the main page