How Old Are the Language Families?

Indo-European and Sino-Tibetan, dated on written texts alone, come out 6,000 to 8,000 years old, with lower bounds older than every Masoretic Babel except one Sino-Tibetan bound. Deep families without old texts rest on radiocarbon. On conventional dates Egyptian and Sumerian are written before every Babel date in print; young-earth revisions remove this only by lowering all ancient history.

By AI (Claude Fable 5.1, xhigh thinking; sources and checks by Claude Opus 5.5, thinking level unavailable; split into parent and children by Claude Opus 5.5) Draft:  October 2026

Summary

If Babel made the language families, no family can be older than Babel and no language can be written down before it.  Both claims fail on conventional dates.  The families dated on written texts alone are the oldest, at 6,000 to 8,000 years for Indo-European and Sino-Tibetan, and two unrelated families are written down before every Babel date in print.  The parent article, Language Families and the Tower of Babel, states the young-earth model;  its other children ask whether language began once and whether there is time since Babel for the families.

Where language dates come from.  A family's age is inferred from how much its languages differ, scaled by a rate learned from calibration points:  dated texts, historical events, or radiocarbon-dated archaeology.1  A young-earth reader accepts the text chronology back to about 2000 BC but disputes radiocarbon before about 1000 BC.  The Babel dates in years before AD 2000 are 4,242 for Ussher, 4,358 for Morris and 4,767 for the Septuagint.

The text-dated families are the oldest.  Indo-European is 8,120 years old on 52 dated ancient languages,2 and 5,950 on a model that forces attested languages to be ancestors.3  Sino-Tibetan is 7,184 years on four dated texts.4  Every text these runs use is within the last 3,450 years.  Their lower 95% bounds are older than every Masoretic Babel in print, except Zhang and colleagues' Sino-Tibetan bound of about 2233 BC, which falls nine years after Ussher's 2242 BC and so takes in his date and the earlier Masoretic ones.5  Gray and Atkinson withheld each calibration in turn and recovered 13 of 14 within 390 years, so the rates are stable inside the texts' window.6  The projection two to three times beyond that window can be questioned, as R. M. W. Dixon did in 1997, though his own example runs the other way, to 10,500 years.77

The spread between models does not reach Babel.  One run by Heggarty's team on an older dataset gave 4,997 years, and they reject it as an artefact of excess synonyms.2  Even Chang's lower bound, about 2870 BC, is six centuries before Ussher's date.  The disagreement is between 6,000 and 8,000 years, not between 4,000 and 8,000.

Some families are young, and the deep ones rest on radiocarbon.  Japonic is about 2,200 years old, Turkic about 2,100, and Dravidian a median 4,433.8910  Bantu, Pama-Nyungan, Uto-Aztecan and Afroasiatic beyond Semitic are calibrated on archaeology, so their ages move with whatever a young-earth model does to radiocarbon.  The younger automated ASJP dates of Holman and colleagues rest on archaeology at the deep end too, and the Bantu specialist Koen Bostoen finds them too shallow.1111

Written before Babel.  On conventional dates Egyptian is written by rebus at Abydos about 3320 BC, and Sumerian is convincingly identified by about 2800 BC.1213  Elamite, written by about 2240 BC, and Hurrian names of the 23rd century BC bring the count to four before the later Masoretic dates.1214  Murray Adamthwaite's claim that these languages appear at the same time fails, since their first texts span about 2,500 years on his own chronology.15  The revised chronologies of Ashton and Down and of Osgood make all writing post-Babel by construction.  Their cost is set out in Egypt and the Flood date and Mesopotamian absolute chronology.1617

Verdict.  The families with no radiocarbon in their dates are the ones older than Babel, and every Bayesian estimate for Indo-European and Sino-Tibetan is older than every Masoretic date.  The written record puts Egyptian and Sumerian before every Babel date in print.  To keep Babel, a young-earth reader must lower the whole ancient chronology, at the price argued elsewhere, and compress the Indo-European and Sino-Tibetan ages, a question about rates of change that the sibling article tests.

Part 1: How old the families are, and how the dates are obtained

Where a language date comes from.  No word list carries a date.  A family's age is inferred from how much its languages differ, scaled by a rate of change, and the rate is learned from calibration points:  a language whose texts are dated, or a split tied to an event.  Hammarström states the rule:  "An estimate of time-depth is possible thanks to a number of calibration points, either historical, inscriptional or archaeological (in the rare event that one can convincingly argue a link between an archaeological entity and a (proto-)language)."1  The distinction matters to a young-earth reader more than to anyone else.  A text date rests on the chronology of the ancient Near East, Greece, India or China, which the young-earth side accepts back to about 2000 BC and disputes only at the Mesopotamian and Egyptian third millennium.  An archaeological date is a radiocarbon date, which the young-earth side disputes before about 1000 BC.  So the table below marks every calibration by type.

The table.  Ages are years before present;  the present is AD 1950 for some papers and AD 2000 for others, which moves nothing by more than fifty years.  For comparison, the Babel dates in the same units, before AD 2000:  Ussher's 2242 BC is 4,242;  Morris's 2358 BC is 4,358;  the Septuagint's 2767 BC is 4,767.  T is a written text or inscription;  H a historical event;  L a loanword stratum or other linguistic inference;  A archaeology dated by radiocarbon.  The last column is the automated ASJP date from Holman and colleagues 2011, for reasons given below.

Family Root age, 95% interval Calibrations Source ASJP
Indo-European 8,120 (6,740-9,610) 52 dated ancient languages, all T; oldest Hittite 3,450 Heggarty 20232 4,348
Indo-European 7,116-10,410 20 tips and 14 nodes, T and H; oldest Hittite 3,450 Bouckaert 201218
Indo-European 7,800-9,800 14 constraints, T and H Gray and Atkinson 20036
Indo-European 5,950 (4,870-7,190) T tips only, with 8 ancestry constraints Chang 20153
Indo-European 4,997 Heggarty's model on the older IELex data with Chang's constraints; a run its authors reject Heggarty 20232
Sino-Tibetan 7,184 (5,093-9,568) 4 tips, all T; oldest Old Chinese 2,800-2,300 Sagart 20194 5,261
Sino-Tibetan 5,871 (4,183-7,817) 8, T and H; oldest Chinese 2,700 ± 150 Zhang 20195
Sino-Tibetan 7,983 (4,778-11,285) 6, T and H Zhang 202019
Semitic 5,750 (4,400-7,400) 5 T tips, oldest Ugaritic 3,400; 5 nodes T and A; root floor 4,350, the first Akkadian Kitchen 200920 3,301
Afroasiatic first split before 11,000 BC A only: reconstructed vocabulary matched to dated cultures Ehret 200221 6,016
Afroasiatic ninth millennium BC A, with lexicostatistics Militarev 200222
Uralic about 5,300 (3,330-7,503) 3 nodes, L, H and A; oldest 2,630 ± 390 Honkola 201323 3,178
Dravidian median 4,433 (about 3,000-6,500) 5 node bounds, T; Tamil from 254 BC Kolipakam 20189 2,055
Austronesian 5,230 (4,750-5,800); 5,117 (4,660-5,680) 14: 10 nodes A and H, 4 tips T; then 5: Proto-Oceanic 3,200-3,600 A, plus 4 T tips Gray 2009; Greenhill 20102425 3,633
Bantu Narrow Bantu 4,847 ± 138; root 6,930 ± 419 4, all A Grollemund 201526 Benue-Congo 4,940
Bantoid 5,110 (4,640-5,770) 4, all A, from Grollemund Koile 202227
Pama-Nyungan mean 5,671 (4,455-6,966) 1 A prior; dates scale with it Bouckaert 201828 4,295
Uto-Aztecan about 4,100 (3,258-5,025) 5 nodes A, 1 T tip Greenhill 202229 4,018
Japonic median 2,182 (1,239-4,190) T tips and an H node Lee and Hasegawa 20118 1,564
Turkic 2,066 (1,517-2,755) 2 T tips, Old Turkic 1,150 and Cuman 700 Savelyev and Robbeets 202010 3,404
"Transeurasian", not an accepted family 9,181 (5,595-12,793) node priors from linguists' estimates, L and H; T tips Robbeets 202130

What the text-dated families show.  Five families are dated on written texts alone:  Indo-European, Sino-Tibetan, Dravidian, Japonic and Turkic.  Every text they use is within the last 3,450 years.  Japonic and Turkic come out younger than Babel on any account, and the Turkic root lands where Turkologists put it on other grounds, which shows the method works over short extrapolations.10  Dravidian's median of 4,433 years before present is about 2430 BC, about 190 years before Ussher's Babel, 75 before Morris's and three centuries after the Septuagint's, though its interval runs from about 3,000 to 6,500.  Indo-European and Sino-Tibetan come out at 6,000 to 8,000 years.  Sagart's calibrations are four dated texts, "the date of the earliest text rounded to the nearest century", and the root is 7,184 years with a lower bound of 5,093.4  Bouckaert's calibrations are "based on historical sources" throughout, the oldest being the Hittite tip at 3,450 years.18  Heggarty's 52 dated languages give the same answer, and removing Vedic and Avestan, whose dates rest on oral tradition, moved the root by 94 years.2  Gray and Atkinson withheld each of their 14 constraints in turn and recovered 13 of them within 390 years, the best evidence that the rates are stable inside the window the texts cover.6  The rate is learned inside about 3,500 years and projected two to three times further back.  That extrapolation is the one thing a young-earth reader can question, and it is the subject of Is There Time Since Babel for the Language Families?.

The model-dependence the young-earth reader can press.  The same data give different roots under different models.  Chang and colleagues forced eight attested languages to be the direct ancestors of their descendants, so that rates were learned from observed change rather than inferred, and got 5,950 years from their narrow dataset;  rerunning Bouckaert's own data without that constraint replicated Bouckaert's chronology, whose published median is 7,580.33  Heggarty's team answered with a model in which the data decide whether an ancient language is an ancestor, and found that the prior alone gave 5,815 years and the cognate data pushed it to 8,120.2  But they also reported that their model on the older IELex dataset with Chang's eight constraints gave 4,997 years, a 42% drop, which they attribute to excess synonyms in IELex inflating the rates from 0.0055 to 0.0132 changes per cognate set per thousand years.22  So a mainstream-published configuration returns an Indo-European root within about 750 years of a Masoretic Babel.  The young-earth reader who cites that number must also answer the authors' argument that it is an artefact, and must note that even Chang's lower bound, 4,870 years before AD 2000 or about 2870 BC, is six centuries before Ussher's date.  Sino-Tibetan shows the same spread:  5,871, 7,184 and 7,983 years from three teams on three datasets, the means 2,100 years apart.  Only Zhang 2019's interval reaches into a Masoretic window:  its lower bound falls nine years after Ussher's 2242 BC.5

Dixon's objection, stated correctly.  Gray and Atkinson open their paper by quoting R. M. W. Dixon that the age of Indo-European "could be anything – 4,000 years BP or 40,000 years BP are both perfectly possible".6  Dixon wrote those words about the hypothetical ancestor of all Australian languages, not about Indo-European.7  What he wrote about Indo-European, on the page before, is better for the young-earth reader than the misquotation:  "What has always filled me with wonder is the assurance with which many historical linguists assign a date to their reconstructed proto-language. ... We are told that proto-Indo-European was spoken about 6,000 years ago. What is known with a fair degree of certainty is the time between proto-Indo-Aryan and the modern Indo-Aryan languages - something of the order of 3,000 years. But how can anyone tell that the development from proto-Indo-European to proto-Indo-Aryan took another 3,000 years?"7  "The changes could have been rapid (as on Woodlark Island) or slow. We simply don't know."7  That is the extrapolation objection from a leading field linguist.  It cuts both ways, and Dixon's own example runs the other way:  "Why couldn't proto-Indo-European have been spoken about 10,500 years ago?"7

The deep families rest on radiocarbon.  Bantu's four calibrations are the Shum Laka rock shelter, the first farmers of southern Cameroon, the Obobogo village site and the Urewe pottery of the Great Lakes, "based on archaeological data" throughout;26 Koile's 2022 tree reuses them, so the two Bantu dates are not independent.27  Pama-Nyungan rests on one archaeological prior for the Western Desert, and its authors say the dates "scale approximately linearly with this calibration".28  Uto-Aztecan's five node calibrations are plant distributions and site changes.29  Ehret's Afroasiatic is reconstructed grass-processing vocabulary matched to the Cataract tradition "probably not long after 16,000 BCE".21  Whatever a young-earth model does to the radiocarbon timescale before about 1000 BC, it does to these dates.  That argument belongs to the Neolithic and the Flood date, and this article does not make it.  What this article says is that the text-dated families, which carry no radiocarbon, are the ones that come out oldest.

The world-wide automated dates.  Holman and colleagues dated nearly every family by a glottochronology-like formula on 40-word lists, fitted on 52 calibration points, with a mean error of 29% of the date.11  Their ASJP ages are younger than the Bayesian ones in every row of the table except Turkic and Bantu, where the ASJP figure is for the larger Benue-Congo group, by a factor of up to about 2.2:  Indo-European 4,348, Semitic 3,301, Dravidian 2,055.  Several sit at or after a Masoretic Babel.  A young-earth writer may cite this, with two cautions the data impose.  First, "the calibration dates older than 2500 BP are all archaeological", so the deep end of the ASJP clock is radiocarbon too.11  Second, the method is contested from the other side:  Koen Bostoen, a Bantu specialist, replied that its Niger-Congo and Benue-Congo dates "are significantly shallower than experts tend to assume" and that "All this does not rouse optimism about the possibility of a universally valid formula for calculating the rate of language change."11  An article that takes ASJP for some families and Bayesian dates for others is choosing its numbers;  this one prints both for every family that has both.

Why a young-earth chronologist accepts the text calibrations.  The tips in Heggarty's and Chang's runs are Hittite about 1450 BC, Mycenaean Greek about 1350 BC, Vedic no younger than 1000 BC, Avestan, Old Persian, classical Greek, Latin, Gothic and the medieval languages.  Their dates are the dates of their manuscripts and inscriptions, anchored on the Hittite and Egyptian synchronisms, Assyrian eponyms, Greek Olympiads and Roman consuls.  Mesopotamian absolute chronology shows the Middle Chronology fixed by tree rings from about 1900 BC.  No young-earth writer found disputes any of these dates for language, and the revised chronologies that would move the Late Bronze Age ones are tested as an escape route in Is There Time Since Babel for the Language Families? rather than assumed here.

Part 2: Families written down before each Babel date

If Babel made the families, no language can be written down before it.  Adamthwaite is the one young-earth writer who argues from the written record, and his claim is that the unrelated languages of the ancient Near East "all appear on the scene at about the same time—the mid-third millennium BC on conventional chronology", which is "precisely what we would expect from the story of the Tower of Babel."1515  The record, read from the specialists, does not say that.

What counts as attestation.  "Writing" is not the same as a text in an identified language.  Christopher Woods on the Uruk tablets of about 3200 BC:  "the evidence that the language represented by the proto-cuneiform script is, in fact, Sumerian is quite slim, being based on perhaps fifteen instances of phonetic writings", though "Most scholars today assume that Sumerian underlies the earliest texts from Mesopotamia."13  Piotr Michalowski:  the archaic texts from Ur, "dated approximately 2800 BC", carry "a few morphological elements and phonetic glosses" that provide "convincing evidence that the language of the texts is indeed Sumerian", and by the Fara and Abu Salabikh texts of about 2500 BC "for the first time we have evidence for an extensive written poetic tradition."12  Egyptian is secure earlier.  The tags from Abydos Tomb U-j, "which dates to ca. 3320 bc", are read by the rebus principle:  "several tags from Tomb U-j bear the image of an elephant on top of a mountain: Ꜣbw, Elephantine", a reading that ties the signs to the Egyptian language and no other.1313  Proto-Elamite of about 3100 BC is undeciphered and cannot be assigned a language.

The table.  "Secure" means a text whose language is identified.  Dates are conventional:  the Middle Chronology for Mesopotamia and Anatolia, radiocarbon and the lunar chain for Egypt.

Family First secure text Conventional date What anchors it Source
Afroasiatic, Egyptian Abydos Tomb U-j tags read by rebus about 3320 BC; Dynasty 1 from 3111-3045 BC archaeology and radiocarbon; Aha by 186-sample Bayesian radiocarbon MacArthur in Woods 2010; Dee 201331
Afroasiatic, Semitic Eblaite texts; Akkadian names, then Sargonic texts Eblaite "as early as 2500 BC"; Akkadian "mid-third millennium" king lists back from the Middle Chronology Michalowski; Huehnergard and Woods1212
Sumerian, isolate Uruk proto-cuneiform, probable; Ur archaic texts, convincing; Fara, certain about 3200 BC; about 2800 BC; about 2500 BC archaeology and radiocarbon for Uruk; king lists for Early Dynastic III Woods 2010; Michalowski1213
Elamite, isolate Linear Elamite; cuneiform Treaty of Naram-Sin about 2300 BC; about 2240/2230 BC Naram-Sin of Akkad on the Middle Chronology Desset 20221414
Hurro-Urartian Hurrian names in Akkade-period texts; Tišatal's inscription 23rd-22nd centuries BC; 21st-20th centuries BC Middle Chronology via king lists Wilhelm12
Indo-European Kanišite Hittite personal names 1931 BC; level II about 1970-1835 BC Old Assyrian eponym list, tree-ring-fixed Middle Chronology Kloekhorst 20193232
Hattic, isolate bilingual Hattic-Hittite texts in the Hittite archives 17th/16th to 13th centuries BC Hittite chronology, tied to Mesopotamia Watkins12
Sino-Tibetan Shang oracle-bone inscriptions from King Wu Ding about 1220-1190 BC Chinese dynastic chronology Shaughnessy in Woods 201013

Greek in Linear B, "Perhaps developed in the fifteenth century BC", and the Indo-Aryan gods and horse terms of the Mitanni dynasty add no family.1212  Kloekhorst records, and declines to rely on, the 2018 claim that personal names from Armi in the Ebla texts of the 25th and 24th centuries BC are Anatolian:  "at present I would therefore regard the linguistic status of these names as too uncertain to make any broad claims."32  If it were right, Indo-European would be written before Ussher's Babel;  it is not counted here.

How many unrelated families are written before each Babel date, on conventional dates.

Babel date Whose Families already written Possible additions
2767 BC, start of Peleg's life Septuagint 2: Egyptian, about 3320-3100; Sumerian, about 3200 probable, 2800 convincing Elamite, if Proto-Elamite is Elamite
about 2500 BC Cox 2: Egyptian; Sumerian, certain by Fara; Eblaite "as early as 2500" Elamite
2428 BC, end of Peleg's life Septuagint 2 Elamite; Indo-European via the Armi names
2358 BC Morris 2 Elamite
2242 BC Ussher; Answers in Genesis 3: Egyptian, Sumerian, Elamite from about 2300, with the Treaty of Naram-Sin at 2240/2230 borderline Hurrian, Akkade-period names
2178 BC, end of Peleg's life Seder Olam; Sibley 4: Egyptian, Sumerian, Elamite, Hurrian
about 2100 BC Ashton and Down; Courville 4
2008 BC Seder Olam at Ussher's Abraham; Sibley 4, Hurrian now with Tišatal's inscription Indo-European follows at 1931 BC

So on conventional dates every Babel date in print is preceded by at least two independently written, unrelated families, and the later Masoretic dates by four.  The point does not depend on dendrochronology:  the third-millennium Mesopotamian dates hang from the Middle Chronology by king lists, and the Egyptian ones from radiocarbon and one lunar date.  It is also not a problem for Babel alone.  Egyptian and Sumerian writing precede the Masoretic Flood, which is the subject of Egypt and the Flood date, and Sargonic Akkadian, the floor of Kitchen's Semitic tree at 4,350 years, is earlier than or contemporary with a 2242 BC dispersion.20

Adamthwaite's "same time", checked.  On his own conventional chronology the first written Egyptian, about 3320 BC, and the Uruk texts, about 3200 BC, are 600 to 800 years before the mid-third millennium.  Hattic is known only from second-millennium Hittite archives.  Kassite is second-millennium.  Etruscan is first-millennium.  The "same time" spans about 2,500 years.  His note that attestation can lag existence "by centuries" is right and cuts the wrong way for him:  it means the languages existed before their first texts, which pushes them further before Babel, not closer to it.  Two further errors should be noted so that nobody repeats them.  Adamthwaite writes of "Hittite words which passed into Greek and Latin, and from there to the languages of Western Europe, e.g. wātar : water",15 which mistakes an inherited cognate for a loan:  Hittite wātar and English water both descend from Proto-Indo-European.  And he calls Barry Fell's reading of the Indus script, as an alphabet writing an Indo-European ancestor of Sanskrit, "highly plausible", though his own paragraph opens by saying the script "remains undeciphered", which it does.15  His real point stands:  several unrelated languages appear in the earliest written record of one region.  Nichols says the same from the other side:  "From the ancient Near East alone, in just the last two to three thousand years Sumerian, Elamite, the Hurrian-Urartean line, and Hattic have vanished without descendants or demonstrable kin."33

What the young-earth revised chronologies do to the count.  The two published revisions take every written record after Babel by construction.  Ashton and Down put Menes in the "21st C B.C." and write that "The history of Egypt would begin soon after the dispersion from Babel, around 2100 B.C."1616  Osgood puts the dispersion at the end of Ubaid 2, before the Uruk culture, treats the first Uruk IV writing as a "reinvention" after Babel in which Enmerkar's rule was "a significant factor", and puts Hammurabi at "circa 1450 B.C.-revised".343434  On either, the count is zero for every Babel date, because writing itself is post-Babel.  What it costs is the subject of two sibling articles and is only summarised here.  Manning and colleagues' 2016 tree-ring sequence from Kültepe and Acemhöyük is compatible only with the Middle Chronology, which puts Hammurabi at 1792 BC, about 340 years above Osgood's figure, or the Low-Middle Chronology a few decades lower, and excludes even the Low Chronology.17  Ashton and Down's Egypt has to absorb Dee and colleagues' Bayesian radiocarbon model of 186 dates, which puts Aha's accession at 3111-3045 BC, and the Wenis lunar date of about 2321 BC;  Egypt and the Flood date shows neither dynastic overlap nor radiocarbon depletion can do it, and records that Creation Ministries International withdrew its endorsement of Down's chronology in 2022.31  So the attestation problem is real on conventional dates, and the young-earth answer to it is the same answer, with the same weaknesses, as its answer to Egyptian chronology.  A reader who accepts that answer gets a clean count of zero;  a reader who does not has two unrelated written languages before the earliest Babel in print.

Objections and responses

"The Bayesian dates are model-dependent.  Chang got 5,950 years and Heggarty's own sensitivity run got 4,997.  The field cannot agree, so the dates are worthless."  The spread is real and Part 1 prints it.  But every published figure is older than every Masoretic Babel:  Chang's lower 95% bound is about 2870 BC, six centuries before Ussher's date, and the 4,997 figure is one its authors reject as an artefact of excess synonyms.  The disagreement is between 6,000 and 8,000 years, not between 4,000 and 8,000.  And the constant-rate test of Is There Time Since Babel for the Language Families?, which has no Bayesian model in it, gives the same answer from the Bronze Age texts alone.

"The ancient Near East shows many unrelated languages appearing together in the third millennium.  Adamthwaite is right that this is what Babel predicts."  Several unrelated languages do appear in the earliest records of one region, and that is a fair point.  The dates are not simultaneous:  Egyptian and Sumerian are written 600 to 800 years before the mid-third millennium, Hattic and Kassite are second-millennium, Etruscan first-millennium.  On conventional dates the earliest two precede every Babel date in print.  Adamthwaite's own concession that attestation lags existence by centuries pushes them earlier still.

"Egyptian and Mesopotamian chronology is inflated.  Lower it, as Ashton and Down or Osgood do, and nothing is written before Babel."  True by construction, and Part 2 says so.  The question is what that lowering costs:  the Kültepe tree rings that fix the Middle Chronology, the 186-date radiocarbon model for the First Dynasty, the Wenis lunar date.  Those arguments are made in the sibling articles, and a reader who finds them persuasive should take the count of zero.  This article's point is narrower:  the language record does not add a separate problem beyond the chronological one, and it does not remove it.

"Dravidian, Japonic, Turkic and Uto-Aztecan are all younger than Babel.  Most families are."  Many are, and the article says so, though Dravidian's median is about two centuries older than the Masoretic dates and younger only than the Septuagint's.  The size of a family's internal diversity says nothing about when the family's first language was created;  a family can be young because it replaced its predecessors recently, as Pama-Nyungan did across most of Australia.  What a young-earth model needs is that no family be older than Babel, and the text-dated Indo-European and Sino-Tibetan roots are the ones that fail that test.

Issues and unanswered questions

No open issue bears on this article alone.  The ones it shares with its siblings, among them the unread source behind Dixon's Woodlark Island case and the upper-bound status of the family count, are listed in the parent, Language Families and the Tower of Babel.

Verdict

The families dated on written texts alone, with no radiocarbon anywhere in them, are the ones that come out older than Babel.  Indo-European and Sino-Tibetan are 6,000 to 8,000 years old, and their lower bounds are older than every Masoretic Babel but one, which reaches the window by nine years.  The spread between dating models is real, and every published Bayesian figure for these two families is still older than every Masoretic date;  the one run within about 750 years of it is rejected by its own authors.  Japonic and Turkic are younger than Babel on any account, Dravidian's median lies between the Masoretic and Septuagint dates, and the deep families without old texts move with whatever a young-earth model does to radiocarbon.

On conventional chronology Egyptian and Sumerian are written down before every Babel date in print, and Elamite and Hurrian before the later Masoretic ones.  Adamthwaite is right that several unrelated languages appear in the earliest records of one region, but they do not appear at the same time.  The young-earth answer is to lower the whole ancient chronology, which removes the attestation problem by construction and carries the costs set out in Egypt and the Flood date and Mesopotamian absolute chronology.  Whether the text-dated ages could be compressed into the time since Babel is a question about rates of change, tested in Is There Time Since Babel for the Language Families?.

Companion pages

Sources

  1. 1.

    Hammarström, Harald.  “Linguistic diversity and language evolution.”  Journal of Language Evolution 1(1):19–29.  2016.  A review by the editor of Glottolog that counts 424 families including isolates and tabulates them by continent.  It explains why the comparative method cannot see past about 10,000 years and notes that the density of families does not follow the age of settlement:  South America is as diverse as New Guinea, and Africa much less so.

    1. 1a.

      Page 23:  “An estimate of time-depth is possible thanks to a number of calibration points, either historical, inscriptional or archaeological (in the rare event that one can convincingly argue a link between an archaeological entity and a (proto-)language).” ↩ ↩

  2. 2.

    Heggarty, Paul et al.  “Language trees with sampled ancestors support a hybrid model for the origin of Indo-European languages.”  Science 381:eabg0818.  2023.  A Bayesian dating of 161 Indo-European languages on the IE-CoR cognate database, coded by more than 80 specialists, with 52 ancient languages dated from their texts as calibrations and a model that lets the data decide which ancient languages are ancestors.  It finds a root about 8,120 years before AD 2000, and reports its own sensitivity runs, one of which, on older data with forced ancestry, gives 4,997.

    1. 2a.

      Section Results:  “Our main analysis (Fig. 2) produced an estimated date for the root of the Indo-European language family that is too early to be compatible with the Steppe hypothesis: ~8120 yr B.P., with a 95% credible region of 6740 to 9610 yr B.P. [Date estimates are reported here as a median date before present, followed by the 95% credible region (highest posterior density, or HPD), all rounded to the nearest decade, and taking the "present" for modern languages as 2000 CE.]” ↩ ↩

    2. 2b.

      Section Validation, and robustness analyses:  “enforcing direct ancestry does move the median root date estimate into a far more recent time frame, younger by 3632 years (42.1%), from 8629 yr B.P. (Fig. 4, SA8a) to 4997 yr B.P. (Fig. 4, SA8b).” ↩ ↩ ↩

    3. 2c.

      Section Validation, and robustness analyses:  “Vedic Sanskrit and Avestan are among the oldest languages in IE-CoR and thus offer especially deep calibration points. Their dating is controversial, however, because no original manuscripts survive. We therefore reran our main (M3) model with these two deep calibrations removed. The effect on the root date for Indo-European was negligible: just 94 years (1.16%) older, at 8214 yr B.P. (6785 to 9571 yr B.P.; Fig. 4, SA1).” ↩

    4. 2d.

      Section Results:  “In the prior, direct ancestry probabilities ranged from ~42% to 78% for all 27 potential ancestor languages, and the median root date estimate was 5815 yr B.P. (4149 to 8123 yr B.P.). Including the cognate data shifted the root date 2305 years earlier, to our result of a median age of 8120 yr B.P. in the posterior.” ↩

    5. 2e.

      Section Validation, and robustness analyses:  “This in turn inflates the estimates of rates of change across the tree [from a median of 0.0055 (0.0046–0.0066) to 0.0132 (0.0119–0.0145) changes per cognate set per thousand years], and these faster rate estimates result in younger root age estimates (12).” ↩

  3. 3.

    Chang, Will et al.  “Ancestry-constrained phylogenetic analysis supports the Indo-European steppe hypothesis.”  Language 91(1):194–244.  2015.  A Bayesian dating of Indo-European that forces eight attested ancient languages, such as Latin and Vedic Sanskrit, to be the ancestors of their 39 modern descendants, so the rate of change is learned from observed change.  Calibrated on dated texts only, it gets a root about 5,950 years before AD 2000;  on Bouckaert's data without the constraints it reproduces 7,580.  It is the clearest case of one modelling choice moving the root by 1,600 years.  Mirror:  Escholarship.org

    1. 3a.

      Page 220, figure 5:  “A1: Narrow dataset 5950 7190–4870”.  The median root age, then the 95% HPD interval, in years before AD 2000. ↩ ↩

    2. 3b.

      Page 194:  “Here we present a phylogenetic analysis in which ancestry constraints permit more accurate inference of rates of change, based on observed changes between ancient or medieval languages and their modern descendants, and we show that the result strongly supports the steppe hypothesis.” ↩

    3. 3c.

      Page 221, section 6.2:  “Using BEAST they [Bouckaert et al.] found a root age posterior distribution with median 7580 bp and 95% HPD interval 9350–5970.” ↩

  4. 4.

    Sagart, Laurent et al.  “Dated language phylogenies shed light on the ancestry of Sino-Tibetan.”  Proceedings of the National Academy of Sciences 116(21):10317–10322.  2019.  A Bayesian dating of 50 Sino-Tibetan languages on cognate-coded basic vocabulary.  Its four calibrations are all written texts no younger chronology disputes, Old Chinese, Old Tibetan, Tangut and Old Burmese, yet the root comes out about 7,184 years old.  The date is an extrapolation of a rate learned on branches no older than 2,800 years.  Mirror:  PMC

    1. 4a.

      Section Results, Tree Topologies and Dating:  “The mean root age estimated with the relaxed-clock model is at 7184 B.P., with 95% highest posterior density interval (HPD) [5093–9568] B.P.” ↩ ↩

    2. 4b.

      Section Materials and Methods, Phylogenetic Reconstruction and Dating:  “We specified calibrations as follows: Old Chinese, [2,800 to 2,300] yBP, in a uniform prior; Old Burmese, 800 yBP; Old Tibetan, 1,200 yBP; and Tangut, 900 yBP (the date range for Old Chinese corresponds to the period of the great Classical Chinese texts; the other dates correspond to the date of the earliest text rounded to the nearest century; see also SI Appendix, section 4).” ↩

  5. 5.

    Zhang, Menghan et al.  “Phylogenetic evidence for Sino-Tibetan origin in northern China in the Late Neolithic:  Supplementary Information.”  Nature 569.  2019.  The supplement to a Bayesian dating of 109 Sino-Tibetan languages on 949 root-meanings.  Its eight calibrations all come from written history or recorded migrations, none from radiocarbon, and the root comes out about 5,871 years before 1950.  Its lower 95% bound is the only dated family root found that reaches into a Masoretic Babel window.

    1. 5a.

      Page 9, section 2.4:  “The average time estimation for the divergence of the Sinitic and TB languages was 5,871 years BP (95% Highest posterior density (HPD): 4,183 – 7,817 years BP). In all the examined models, their estimations including average root ages and 95% HDP ranges were below 9,000 years (see Extended Data Fig. 5).” ↩ ↩ ↩

  6. 6.

    Gray, Russell D., and Quentin D. Atkinson.  “Language-tree divergence times support the Anatolian theory of Indo-European origin.”  Nature 426:435–439.  2003.  The first Bayesian dating of Indo-European, on 87 languages and 2,449 cognate sets, calibrated by 14 historically attested dates and run without assuming a constant rate.  It puts the root 7,800 to 9,800 years ago, and its leave-one-out test, which withholds each calibration in turn, is the best evidence that the rates are stable inside the window the texts cover.

    1. 6a.

      Section Abstract:  “our analysis of a matrix of 87 languages with 2,449 lexical items produced an estimated age range for the initial Indo-European divergence of between 7,800BP and 9,800BP. The results were robust to changes in coding procedures, calibration points, rooting of the trees and priors in the Bayesian analysis.” ↩

    2. 6b.

      Section Methods:  “Step-by-step removal of each of the 14 age constraints on the consensus tree revealed that divergence time estimates were robust to calibration errors. For 13 nodes, the reconstructed age was within 390 years of the original constraint range. … Reconstructed ages at the base of the tree ranged from 10,400BP with the removal of the Hittite age constraint, to 8,500BP with the removal of the Iranian group age constraint.” ↩ ↩

    3. 6c.

      Section Opening paragraph:  “Dixon epitomizes this view with his assertion that, based on the linguistic data, the age of Indo-European 'could be anything – 4,000 years BP or 40,000 years BP are both perfectly possible (as is any date in between)'.”  Dixon wrote these words about the ancestor of all Australian languages, not about Indo-European. ↩

  7. 7.

    Dixon, R. M. W.  The Rise and Fall of Languages.  Cambridge University Press.  1997.  A short book on how language families arise and how far they can be dated, by a field linguist of Australian languages who holds that whether all Australian languages share one ancestor cannot be answered.  Its pages 47 and 48 are a mainstream linguist's plain statement that the rate of change measured over the documented last 3,000 years of Indo-Aryan says nothing certain about the earlier, undocumented part.

    1. 7a.

      Page 47:  “What has always filled me with wonder is the assurance with which many historical linguists assign a date to their reconstructed proto-language. … We are told that proto-Indo-European was spoken about 6,000 years ago. What is known with a fair degree of certainty is the time between proto-Indo-Aryan and the modern Indo-Aryan languages - something of the order of 3,000 years. But how can anyone tell that the development from proto-Indo-European to proto-Indo-Aryan took another 3,000 years?” ↩ ↩

    2. 7b.

      Page 48:  “While working on the languages of Australia, over the past 30 years, I have at times experimented with the hypothesis that the languages are related as one family and then looked for a family tree model (an alternative scenario is discussed below). The question that then presented itself was: what could have been the date of the proto-language? After careful consideration, it seemed that the only honest answer is: it could be anything - 4,000 years BP or 40,000 years BP are both perfectly possible (as is any date in between).” ↩

    3. 7c.

      Page 47:  “Really, linguistic dating should be based on linguistic evidence. But languages are known to change at different rates. There is no way of knowing how long it took to go from the presumed homogeneity of proto-Indo-European to the linguistic diversity of proto-Indo-Iranian, proto-Celtic, proto-Germanic, etc. The changes could have been rapid (as on Woodlark Island) or slow. We simply don't know.” ↩

    4. 7d.

      Page 48:  “The later part of this expansion - the last 3,000 years, for which we have some written records - may have been very much quicker, or even very much slower, than the earlier part. Why couldn't proto-Indo-European have been spoken about 10,500 years ago?” ↩ ↩

  8. 8.

    Lee, Sean, and Toshikazu Hasegawa.  “Bayesian phylogenetic analysis supports an agricultural origin of Japonic languages.”  Proceedings of the Royal Society B 278:3662–3669.  2011.  A Bayesian dating of 59 Japonic languages and dialects, calibrated on Old and Middle Japanese and the historical move of the capital.  The root comes out about 2,182 years old, matching the arrival of Yayoi farming.  The same method that makes Sino-Tibetan 7,000 years old makes Japonic young, so it does not simply return old ages.  Mirror:  PMC

    1. 8a.

      Section 3(b), Date estimation:  “the median age of the root, 2182 YBP (the mean: 2398 YBP; the standard error: 47.21 years; 95% HPD: 1239–4190 YBP), is clearly in concordance with the age range of the farming/language dispersal theory.” ↩ ↩

  9. 9.

    Kolipakam, Vishnupriya et al.  “A Bayesian phylogenetic study of the Dravidian language family.”  Royal Society Open Science 5:171504.  2018.  A Bayesian dating of 20 Dravidian languages on basic vocabulary elicited from speakers, calibrated on inscriptions, literary first attestations and relative bounds.  The root comes out a median 4,433 years before AD 2000, which overlaps the Masoretic Babel window better than any other dated family.  Mirror:  PMC

    1. 9a.

      Section 4.3, Dating:  “We find that the root of the tree has a mean of 4650 years ago (median 4433), thus indicating that the ancestor of all Dravidian languages, Proto-Dravidian, may have been spoken around 4500 years ago.”  “the 95% HPD intervals on the root age range from approximately 3000–6500 years ago. Therefore, we cannot exclude the possibility that the root of the Dravidian language family is significantly older than 4500 years.” ↩ ↩

  10. 10.

    Savelyev, Alexander, and Martine Robbeets.  “Bayesian phylolinguistics infers the internal structure and the time-depth of the Turkic language family.”  Journal of Language Evolution 5(1):39–53.  2020.  A Bayesian dating of 32 Turkic languages on 254 meanings, calibrated only on Old Turkic runic inscriptions and the Codex Cumanicus.  The root comes out about 2,066 years old, where Turkologists put it on other grounds, which is a check that the method works over short extrapolations.

    1. 10a.

      Section 5, Results:  “The time-depth of the Turkic family on the maximum credibility tree is estimated to be around 2,066 years BP (median height of the node), with a 95% highest posterior density between 1,517 and 2,755 years BP.” ↩ ↩

    2. 10b.

      Section 6, Discussion:  “The age of 2,066 years BP (66 BC) lies within the bounds traditionally discussed as the probable time-depth of the Turkic language family (2,000–2,500 years BP). However, the size of the credible intervals exceeds one thousand years for the root of the family, allowing for any time between 755 BC and 483 AD.” ↩

  11. 11.

    Holman, Eric W. et al.  “Automated Dating of the World's Language Families Based on Lexical Similarity.”  Current Anthropology 52(6):841–875.  2011.  The ASJP project's dating of nearly every language family with a glottochronology-like formula on 40-word lists, fitted on 52 calibration points and published with critical comments.  Its mean error is 29% of the date, and every calibration older than 2,500 years is archaeological, so the deep part of its clock rests on radiocarbon.

    1. 11a.

      Page 841:  “Automated judgments of lexical similarity for groups of related languages are calibrated with historical, epigraphic, and archaeological divergence dates for 52 language groups. The discrepancies between estimated and calibration dates are found to be on average 29% as large as the estimated dates themselves, a figure that does not differ significantly among language families.” ↩

    2. 11b.

      Page 853:  “Although the calibration dates older than 2500 BP are all archaeological, figure 2 shows that younger dates of all four types are about equally close to the regression line.” ↩ ↩

    3. 11c.

      Page 866:  From the published comment of Koen Bostoen, a Bantu specialist at the Royal Museum for Central Africa.  “the calculated time depths for Niger-Congo and Benue-Congo—that is, 6227 BP and 4940 BP, respectively—are significantly shallower than experts tend to assume. The time depth generally associated with Bantu is 5000 BP, while Niger-Congo is estimated at 12,000–10,000 BP. All this does not rouse optimism about the possibility of a universally valid formula for calculating the rate of language change.” ↩ ↩

  12. 12.

    The Cambridge Encyclopedia of the World's Ancient Languages.  Cambridge University Press.  Edited by Roger D. Woodard.  2004.  The standard reference on the languages of the ancient world, one chapter per language by a specialist:  Piotr Michalowski on Sumerian, Gernot Wilhelm on Hurrian, John Huehnergard and Christopher Woods on Akkadian and Eblaite, Calvert Watkins on Hittite, Woodard on Greek.  Each states when its language is first attested and how securely.

    1. 12a.

      Page 20:  “We have to wait for about four hundred years for our next archaic texts from Ur, dated approximately 2800 BC … the notation of a few morphological elements and phonetic glosses provides convincing evidence that the language of the texts is indeed Sumerian.”  “The next larger groups of texts from Sumer are the Early Dynastic III texts from Fara (ancient Shuruppak), Abu Salabikh, Nippur, and Adab from around 2500 BC. … now for the first time we have evidence for an extensive written poetic tradition.”  From Michalowski's chapter on Sumerian. ↩ ↩ ↩

    2. 12b.

      Page 95:  “Hurrian is first attested in a few words and personal or place names mentioned in Akkadian texts of the Akkade period (twenty-third to twenty-second centuries BC). The term Old Hurrian (herein abbreviated OH) has been coined for the language of a royal inscription most likely to be dated to the Ur III period (twenty-first to twentieth centuries BC)”.  Page 96:  “The oldest Hurrian text is the royal inscription of Tišatal, endan of Urkeš.”  From Wilhelm's chapter on Hurrian. ↩ ↩

    3. 12c.

      Page 26:  “Such a full syllabary is known for Eblaite as early as 2500 BC, but the first adaptation of cuneiform to Semitic must have taken place somewhat earlier.”  From Michalowski's chapter on Sumerian. ↩

    4. 12d.

      Page 218:  “Old Akkadian, the language spoken by the Sargonic kings, refers collectively to the texts from the earliest attestation of Akkadian (mid-third millennium) to the beginning of the second millennium.”  “Contemporaneous with the date of Old Akkadian materials, c. 2400 BC, are the thousands of texts excavated at Ebla in northwest Syria.”  From Huehnergard and Woods's chapter on Akkadian and Eblaite. ↩

    5. 12e.

      Page 552:  “The Hittite language is preserved for us on clay tablets … the tablets themselves written over the period from the seventeenth/sixteenth to the end of the thirteenth century.”  Page 573:  “settled in their homeland of central Anatolia when it was already populated by urbanized non-Indo-European Hattic speakers … Our documents include many bilingual Hattic-Hittite texts”.  From Watkins's chapter on Hittite. ↩

    6. 12f.

      Page 653:  “The earliest of these is the syllabic script of the Mycenaeans called Linear B … Perhaps developed in the fifteenth century BC and based upon the Minoan Linear A script … Linear B script ceases to be attested after the downfall of Mycenaean society in the twelfth century BC.”  From Woodard's chapter on the Greek dialects. ↩

    7. 12g.

      Page 96:  “preserved some archaic Indo-Aryan traditions of unknown origins (dynastic names, some gods known from Vedic sources, hippological terms)”.  From Wilhelm's chapter, on the Mitanni dynasty, which made a peace treaty with Egypt shortly after 1400 BC. ↩

  13. 13.

    “Visible Language:  Inventions of Writing in the Ancient Middle East and Beyond.”  Oriental Institute Museum Publications 32.  Oriental Institute of the University of Chicago.  Edited by Christopher Woods.  2010.  The catalogue of a museum exhibition on the first writing systems, with chapters by specialists:  Woods on Mesopotamia, Elise V. MacArthur on Egypt, Edward L. Shaughnessy on China.  It separates the first writing from the first text securely in a known language, which is the distinction the attestation count turns on.

    1. 13a.

      Page 119:  “Tomb U-j, which dates to ca. 3320 bc, is a rectangular, mudbrick tomb with twelve interconnected chambers”.  From MacArthur's chapter on the Egyptian writing system. ↩ ↩

    2. 13b.

      Page 120:  “writing on the tags shows that the Egyptian writing system had adopted the rebus principle … several tags from Tomb U-j bear the image of an elephant on top of a mountain: Ꜣbw, Elephantine … Using the rebus principle, the place indicated by the signs is not 'Elephant (Ꜣbw) + Mountain/Highlands/Mound (ḫꜢs.t),' but rather the city Elephantine (Ꜣbw).”  From MacArthur's chapter. ↩

    3. 13c.

      Pages 44–45:  “In sum, the evidence that the language represented by the proto-cuneiform script is, in fact, Sumerian is quite slim, being based on perhaps fifteen instances of phonetic writings, of varying degrees of certitude that have been discovered to date. However, the circumstantial evidence provided by the cultural continuity between the late fourth and the mid-third millennium, at which time there is no doubt the language represented is Sumerian, bolsters this argument. Most scholars today assume that Sumerian underlies the earliest texts from Mesopotamia.” ↩

    4. 13d.

      Page 35:  “the script and tablets of the earlier phase are labeled Uruk IV (ca. 3200 bc), while the script and tablets from roughly one hundred years later are labeled Uruk III (also known as the Jemdet Nasr period, ca. 3100 bc) … Approximately five thousand proto-cuneiform tablets were unearthed”. ↩

    5. 13e.

      Page 216:  “It is not until the oracle-bone inscriptions of the late thirteenth-century bc that we find grammatically connected marks that certainly record language … These inscriptions on turtle-shell and ox bone … derive from divinations performed at the royal court of the last nine kings of the Shang dynasty (ca. 1220–1050 bc).”  Page 217:  “the reign of King Wu Ding (reigned ca. 1220–1190 bc)”.  From Shaughnessy's chapter on China. ↩

  14. 14.

    Desset, François et al.  “The Decipherment of Linear Elamite Writing.”  Zeitschrift für Assyriologie 112(1):11–60.  2022.  A decipherment of Linear Elamite as a phonetic script for Elamite, from bilingual silver beakers and royal inscriptions.  It restates that before the decipherment the oldest Elamite text was the cuneiform Treaty of Naram-Sin, about 2240 BC on the Middle Chronology, and that Elamite is an isolate.  Its dates hang from the Mesopotamian chronology.  Mirror:  Hdl.handle.net

    1. 14a.

      Page 44:  “the number of Elamite texts known for the earliest documented phase of this language (in the so-called 'Old Elamite' period, from the 23rd to the 16th century BCE) was extremely limited, only including: 1) the so-called 'Treaty of Naram-Sin' from Susa …, dating to ca. 2240/2230 BCE. It was written without any Akkadian loanword in an almost completely phonographic writing”.  “the Elamite language remains an apparent linguistic isolate, preventing any linguistic comparisons, despite hypotheses trying to connect it to Dravidian languages of India, the Afro-Asiatic linguistic group, or the Caucasian languages.” ↩ ↩

    2. 14b.

      Page 11:  “Linear Elamite writing was used in southern Iran in the late 3rd/early 2nd millennium BCE (ca. 2300–1880 BCE). … A full description and analysis of Linear Elamite of writing, employed for recording the Elamite language, is given here for the first time”.  Page 15:  “All dates are given according to the so-called 'Middle Chronology' (Sallaberger/Schrakamp 2015).” ↩

  15. 15.

    Adamthwaite, Murray.  “The languages of Babel.”  Creation 42(1):52–55.  Nov 11, 2019.  A popular article by a young-earth writer with a doctorate in ancient Near Eastern studies, the only young-earth piece found that uses the written record of languages.  It argues that many unrelated, complex languages appear in the Near East at about the same time, the mid-third millennium BC, as Babel predicts.

    1. 15a.

      Section The first phenomenon:  “Two phenomena stand out; first, all these languages appear on the scene at about the same time—the mid-third millennium BC on conventional chronology. For secular scholars, their origins remain obscure.” ↩ ↩

    2. 15b.

      Section The second phenomenon:  “some of the vocabulary of these ancient languages passed into later languages, notably Hittite words which passed into Greek and Latin, and from there to the languages of Western Europe, e.g. wātar : water.” ↩

    3. 15c.

      Section The first phenomenon:  “Indus Valley language: the script of this very early culture remains undeciphered, and the underlying language remains unknown. Linguist Barry Fells attempted a decipherment in the 1970s … While not all have accepted Fells' decipherment, the Sanskrit connection makes it highly plausible.” ↩

    4. 15d.

      Section A polyglot of unrelated languages:  “Indeed, what is striking about the linguistic landscape of the Ancient Near East is the plethora of unrelated but highly complex languages. … Rather, what we find is precisely what we would expect from the story of the Tower of Babel.” ↩

  16. 16.

    Ashton, John, and David Down.  Unwrapping the Pharaohs:  How Egyptian Archaeology Confirms the Biblical Timeline.  Master Books.  2006.  A popular young-earth book that lowers the whole of Egyptian chronology so that Egyptian history begins after Babel, putting King Menes in the 21st century BC against a conventional 2920 to 2770 BC.  Creation Ministries International withdrew its endorsement of Down's chronology in 2022.

    1. 16a.

      Page 8, caption:  “REVISED DATES: 21st C B.C.   TRADITIONAL DATES: 2920-2770 B.C.   KING MENES.” ↩

    2. 16b.

      Page 200:  “The history of Egypt would begin soon after the dispersion from Babel, around 2100 B.C.” ↩ ↩

  17. 17.

    Manning, Sturt W. et al.  “Integrated Tree-Ring-Radiocarbon High-Resolution Timeframe to Resolve Earlier Second Millennium BCE Mesopotamian Chronology.”  PLoS ONE 11(7):e0157144.  2016.  A tree-ring sequence with radiocarbon wiggle-matching from timbers at Kültepe and Acemhöyük, buildings tied to named Old Assyrian officials.  It fixes the early second millennium BC to the Middle or Low-Middle Chronology and excludes the lower chronologies, which anchors the Old Assyrian and Kanišite dates used here. ↩ ↩

  18. 18.

    Bouckaert, Remco et al.  “Mapping the origins and expansion of the Indo-European language family.”  Science 337:957–960.  2012.  A Bayesian dating and mapping of 103 ancient and modern Indo-European languages on basic vocabulary.  Its 20 dated ancient languages and 14 dated splits all come from historical sources, the oldest being Hittite at about 3,450 years, and its root falls 7,116 to 10,410 years ago in Anatolia.  Mirror:  PMC

    1. 18a.

      Section Main text, the paragraph after Table 1:  “The inferred location (Figure 1) and timing (7,116–10,410 BP 95% highest posterior density [HPD]) of Indo-European origin is congruent with the proposal that the family began to diverge with the spread of agriculture from Anatolia 8,000–9,500 years BP (11).” ↩

    2. 18b.

      Section Supplementary Materials, p. 7, Calibrating rates of change:  “The rate of cognate replacement per unit time is calibrated by constraining lineage divergence times (at internal nodes) and ancient language ages (at the tips) based on historical sources after ref. (1, 16) as listed in Table S2. We used probability distributions rather than point estimates for the age of each internal node and ancient language, in order to account for uncertainty in the historical data.” ↩

  19. 19.

    Zhang, Hanzhi et al.  “Dated phylogeny suggests early Neolithic origin of Sino-Tibetan languages.”  Scientific Reports 10:20792.  2020.  A third Bayesian dating of Sino-Tibetan, on 131 languages and 1,726 cognate sets, with six calibrations taken only from written records.  It gets a root about 7,983 years old, and its lower 95% bound is older than every Masoretic Babel date.  Mirror:  PMC

    1. 19a.

      Section Results:  “We estimate the mean root age at 7983 years BP, with a 95% highest posterior density interval of 4778–11,285 BP.”  “We calibrated the tree using historical records (Table S1), including extinction times for the historical languages (Old Chinese, Padam, Shaiyang). … Unlike in one previous phylogenetic study of the Sino-Tibetans, our calibrations are based solely on these empirical records rather than traditional linguistic theories.” ↩

  20. 20.

    Kitchen, Andrew et al.  “Bayesian phylogenetic analysis of Semitic languages identifies an Early Bronze Age origin of Semitic in the Near East.”  Proceedings of the Royal Society B 276:2703–2710.  2009.  A Bayesian dating of 25 Semitic languages on 96-word lists, calibrated on five dated ancient languages and five node bounds drawn from epigraphy and archaeology.  The root comes out about 5,750 years old, and its lower bound is fixed at the first Akkadian writing, so the analysis cannot return a younger root.  Mirror:  PMC

    1. 20a.

      Section 3(b), Semitic language divergence dates:  “Our phylogeny indicates the most basal divergence within Semitic occurred at 5750 YBP (HPD: 4400–7400 YBP), suggesting an origin of Semitic during the Early Bronze Age”. ↩

    2. 20b.

      Section 2(b), Phylogenetic analysis and divergence date estimation:  “An additional constraint (v) was placed on the time of the most recent common ancestor of the included Semitic languages to 4350–8000 YBP (the lower date is based on the earliest known epigraphic evidence of Akkadian; Buccellati 1997). An analysis was also performed without the constraint on the age of the root, which returned an estimate of 4300–7750 YBP for the root, i.e. almost exactly our constraint range.” ↩

  21. 21.

    Ehret, Christopher.  The Civilizations of Africa:  A History to 1800.  University Press of Virginia.  2002.  A history of Africa by a historian and comparative linguist who calls the Afroasiatic family Afrasan.  It ties reconstructed proto-Afrasan words for flour and grindstones to the Cataract Tradition of wild-grass collectors, so its dates for the family are radiocarbon dates of archaeological cultures matched to vocabulary, not a linguistic clock.

    1. 21a.

      Page 38:  “One of these, the Southern Afrasan peoples, have long been inhabitants of the Ethiopian Highlands. It was they who most likely adapted the ideas of wild grain collection to the cooler grasslands of that region sometime before 11,000 BCE.” ↩

    2. 21b.

      Page 35:  “Across an expanse of lands running from the region of the Nubian Nile eastward through the Red Sea hills to the northern Ethiopian Highlands—and probably not long after 16,000 BCE—an entirely new tack in gathering and hunting, unknown previously in the world, was taken by some of the peoples living in those regions.” ↩

  22. 22.

    Militarev, Alexander.  “The Prehistory of a Dispersal:  the Proto-Afrasian (Afroasiatic) Farming Lexicon.”  Examining the Farming/Language Dispersal Hypothesis, 135–150.  McDonald Institute for Archaeological Research.  Edited by Peter Bellwood and Colin Renfrew.  2002.  A chapter by a Moscow long-range comparativist that dates the breakup of Proto-Afroasiatic by lexicostatistics on loan-corrected 100-word lists and ties its speakers to the Natufian culture of the Levant through more than 30 reconstructed farming words.  Its method is glottochronological and its homeland date rests on the radiocarbon chronology of the Natufian.

    1. 22a.

      Page 135:  “The Proto-Afrasian language, on the verge of a split into daughter languages, should be roughly dated to the ninth millennium BC.”  “The Proto-Afrasian speakers can be identified, with high probability, with the creators of the Natufian and Post-Natufian Early Neolithic archaeological culture of the Levant.” ↩

  23. 23.

    Honkola, T. et al.  “Cultural and climatic changes shape the evolutionary history of the Uralic languages.”  Journal of Evolutionary Biology 26(6):1244–1253.  2013.  A Bayesian dating of Uralic basic vocabulary calibrated on three splits dated from loanword strata, the movements of neighbouring peoples and archaeology, none from a text of the family itself.  The root comes out about 5,300 years old, but its lower 95% bound, about 3,330 years, is well after any Babel date.

    1. 23a.

      Section Results:  “The Proto-Uralic phase diverged ca. 5300 YBP [HPD (= 95% highest probability density): 7503–3330 YBP] with the separation of the Finno-Ugric and Samoyed clades (Fig. 2).”  Section Materials and methods:  “We calibrated our topology based on the estimated times of divergence of three subgroups of the Uralic languages. These dates can be considered fairly well supported by several fields of study, that is, loanword studies, history and archaeology.” ↩

  24. 24.

    Gray, Russell D. et al.  “Language phylogenies reveal expansion pulses and pauses in Pacific settlement.”  Science 323:479–483.  2009.  A Bayesian dating of 400 Austronesian languages from the Austronesian Basic Vocabulary Database.  It puts the root about 5,230 years ago in Taiwan, calibrated on settlement dates from archaeology and on four languages of known date.

    1. 24a.

      Page 481:  “We calibrated 10 nodes on the trees with archaeological date estimates and known settlement times (28). The languages Old Chinese and Old Javanese were calibrated to the ages when they were spoken, and Favorlang and Siraya were calibrated to their data collection times.”  “The estimated root age of Austronesian across all the post–burn-in trees has a mean of 5230 years [95% highest posterior density (HPD) interval, 4750 to 5800 years B.P.].” ↩

  25. 25.

    Greenhill, Simon J. et al.  “How accurate and robust are the phylogenetic estimates of Austronesian language relationships?”  PLoS ONE 5(3):e9573.  2010.  A robustness test of the 2009 Austronesian dating by the same team, which re-dates the tree on only the five least controversial calibrations:  four historical languages and the Proto-Oceanic split, which rests on the radiocarbon-dated Lapita horizon.  The root barely moves, to about 5,117 years.  Mirror:  PMC

    1. 25a.

      Section Results:  “The estimated age of the trees had a mean of 5,117 years, and a 95% highest posterior density interval of 4,660 to 5,680 years BP”.  Section Materials and Methods:  “To assess the robustness of the date estimates on the Austronesian language phylogenies we used the 5 least controversial calibrations. Proto-Oceanic (3,200–3,600 BP). Old Javanese (700–1,200 BP). Old Chinese (2,300–2,900 BP). Favorlang (346–384 BP). Siraya (346–384 BP).” ↩

  26. 26.

    Grollemund, Rebecca et al.  “Bantu expansion shows that habitat alters the route and pace of human dispersals.”  Proceedings of the National Academy of Sciences 112(43):13296–13301.  2015.  A Bayesian dating of about 400 Bantu languages on 100 meanings.  All four of its calibrations are radiocarbon-dated archaeological sites, so its root of about 6,900 years and its Narrow Bantu date of about 4,800 are radiocarbon dates carried onto a language tree.  Mirror:  PMC

    1. 26a.

      Section Supplementary Information, Fig. S1 caption:  “On the basis of four calibration ranges supported by archaeological studies … the root of the tree estimates a common ancestor with the outgroup Grassfields speakers at ∼6,900 B.P. (node 0, Fig. 1; age = 6,929.7 ± 418.6 B.P.), a date considerably older than the 5,000-B.P. younger limit suggested by our calibration range. The tree then dates the remaining Bantu in-group (node b) to ∼4,800 B.P. (4,846.5 ± 138.1)”. ↩

    2. 26b.

      Section Supplementary Information, Materials and Methods, Calibration Ranges:  “We calibrated four parts of our trees: three using date ranges and one as a fixed point. These calibrations are based on archaeological data and are labeled in Fig. 1: (a) 5,000+ Bantoid, non-Bantu; (b) 4,000–5,000 Narrow Bantu; (c) 3,000–3,500 Mbam-Bubi ancestor; and (d) 2,500 Eastern Bantu.” ↩

  27. 27.

    Koile, Ezequiel et al.  “Phylogeographic analysis of the Bantu language expansion supports a rainforest route.”  Proceedings of the National Academy of Sciences 119(32):e2112853119.  2022.  A Bayesian phylogeographic dating of 419 Bantu and Bantoid varieties that reuses Grollemund and colleagues' four archaeological calibrations, so the two Bantu dates are not independent tests.  Its root comes out a median 5,110 years old.  Mirror:  PMC

    1. 27a.

      Section Results, Part 2:  Geographic Model:  “The median age of root is estimated to be 5,110 BP (95% HPD 4,640 BP to 5,770 BP), while the split between narrow Bantu and the northwestern branch Mbam–Bubi dates to 4,420 BP (95% HPD 4,040 BP to 5,000 BP).” ↩

    2. 27b.

      Section Materials and Methods, Calibration Points:  “We considered the following calibration points drawn from ref. 3: 1) 5,000+ Bantoid, non-Bantu (92), 2) 4,000 to 5,000 Narrow Bantu (2, 29, 30, 93–96), 3) 3,000 to 3,500 Mbam–Bubi ancestor (97), and 4) 2,500 Eastern Bantu (74).”  Reference 3 is Grollemund and colleagues 2015. ↩

  28. 28.

    Bouckaert, Remco R. et al.  “The origin and expansion of Pama–Nyungan languages across Australia.”  Nature Ecology & Evolution 2:741–749.  2018.  A Bayesian phylogeographic study of 306 Pama-Nyungan languages that places the family's origin south of the Gulf of Carpentaria in the mid-Holocene, a rapid replacement across most of Australia.  Its dates rest on one archaeological calibration, and the authors say they scale with it.

    1. 28a.

      Section Results:  “This reveals clear support for a homeland at the base of the Gulf of Carpentaria, with an estimated age of 4,455–6,966 years (95% highest posterior density; mean = 5,671 years).” ↩

    2. 28b.

      Section Results:  “One factor that makes inferences about Australian prehistory both difficult and consequential is the scarcity of established events that can be used to calibrate rates of change (see Supplementary Note 1). While the fit between the rapid-replacement hypothesis and our inferred location and date of origin is compelling, our time estimates are based on a single age calibration of the Western Desert subgroup and scale approximately linearly with this calibration.” ↩

  29. 29.

    Greenhill, Simon J. et al.  “A Recent Northern Origin for the Uto-Aztecan Family.”  SocArXiv.  2022.  The preprint of a Bayesian dating of 34 Uto-Aztecan varieties, later published in Language 99(1), 2023.  It puts the family's origin about 4,100 years ago among foragers near southern California, calibrated on five nodes from archaeology and dated plant distributions and one historical language.

    1. 29a.

      Page 2:  “We infer the age of Proto-Uto-Aztecan to be around 4,100 years ago (3,258 - 5,025 years), and identify the most likely homeland to be near what is now southern California.” ↩

    2. 29b.

      Page 8, section 2:  “To calibrate these clocks and to estimate divergence times, we used archaeological evidence, reconstructed plant names, and dated plant distributions to calibrate the date ranges for five nodes and one ancient language”.  “Both Fowler (1983) and Hill (2015) argue that the dispersal of Proto-Northern Uto-Aztecan speakers is unlikely to have occurred before the earliest attested dates of Mesquite (Prosopis spp.) in the area ~3,490 BP.” ↩

  30. 30.

    Robbeets, Martine et al.  “Triangulation supports agricultural spread of the Transeurasian languages.”  Nature 599:616–621.  2021.  A Bayesian dating of 98 Japonic, Koreanic, Tungusic, Mongolic and Turkic varieties as one "Transeurasian" family, combined with archaeology and genetics.  It gets a root about 9,181 years old, but it assumes the grouping it dates, which Glottolog and most specialists do not accept, and its calibrations are linguists' estimates of proto-language ages rather than texts.  Mirror:  PMC

    1. 30a.

      Section Main text, Linguistics:  “Our results indicate a time-depth of 9181 bp (5595–12793 95% highest probability density (95% HPD)) for the Proto-Transeurasian root of the family”. ↩

  31. 31.

    Dee, Michael et al.  “An absolute chronology for early Egypt using radiocarbon dating and Bayesian statistical modelling.”  Proceedings of the Royal Society A 469:20130395.  2013.  A Bayesian model of 186 radiocarbon dates, 74 of them new measurements on short-lived samples, from Predynastic and First Dynasty Egypt.  It dates the accession of King Aha, the start of the First Dynasty, to 3111 to 3045 BC at 68% probability, independently of the king lists.  Mirror:  PMC ↩ ↩

  32. 32.

    Kloekhorst, Alwin.  “Kanišite Hittite:  The Earliest Attested Record of Indo-European.”  Handbuch der Orientalistik 1.132.  Brill.  2019.  A study of the 271 assured and 81 possible Hittite personal names in about 23,000 Old Assyrian merchant tablets from Kaniš in Anatolia, which shows them to be an early dialect of Hittite and so the oldest Indo-European words anywhere.  The tablets are dated on the Middle Chronology through the Old Assyrian eponym list.

    1. 32a.

      Page 1:  “The earliest dated tablet from Kültepe dates to ca. 1935 BCE, and the youngest one to 1717 BCE (dates according to the Middle Chronology).”  Page 3:  “these Hittite names would form the oldest witnesses of an Indo-European language anywhere in the world.”  Page 3, note 9:  “The earliest attested Kanišite name is ḫazuala (g.u.) as found on AKT 6a, 1 (dated to REL 42 = 1931 BCE, cf. Barjamovic, Hertel & Larsen 2012: 92), although it is not fully clear whether this name is really Hittite or not”. ↩

    2. 32b.

      Page 8:  “Of the in total 23,000 Old Assyrian tablets that are excavated, the vast majority stem from the archaeological stratum Kaniš Lower Town level II (ca. 1970–1835 BCE), whereas Lower Town level Ib (ca. 1835–1710 BCE) has yielded only some 450 tablets.” ↩

    3. 32c.

      Page 3, note 9:  “Kroonen, Barjamovic & Peyrot (2018) have recently claimed that a number of personal names that are recorded in texts from Ebla, dated to the 25–24th centuries BCE (Bonecchi 1990), and which refer to individuals said to be from the state of Armi … belong to one or more languages 'that clearly fall within the Anatolian Indo-European family' (2018: 6). … However, Kroonen, Barjamovic & Peyrot do not offer a detailed analysis of this material, and at present I would therefore regard the linguistic status of these names as too uncertain to make any broad claims.” ↩

  33. 33.

    Nichols, Johanna.  “The Origin and Dispersal of Languages:  Linguistic Evidence.”  The Origin and Diversification of Language, Memoirs of the California Academy of Sciences 24, 127–170.  California Academy of Sciences.  Edited by Nina G. Jablonski and Leslie C. Aiello.  1998.  A chapter by a leading typologist that defines the "stock" as the oldest grouping the comparative method can prove, puts the oldest stocks at about 6,000 years, and explains that ceiling by vocabulary loss.  It then models how fast stocks multiply and finds that one ancestor would need at least 72,000 to 132,000 years to yield the world's stocks.  Every input of that model is calibrated on the mainstream ages of known families.

    1. 33a.

      Page 128:  “From the ancient Near East alone, in just the last two to three thousand years Sumerian, Elamite, the Hurrian-Urartean line, and Hattic have vanished without descendants or demonstrable kin.” ↩

  34. 34.

    Osgood, A. John M.  “The Ark Landing, the Migration, and the Early Settling of Noah's Sons.  Part 1:  The Beginning and the Sons of Shem.”  Answers Research Journal 16:501–528.  2023.  A young-earth reconstruction of Mesopotamian prehistory in Answers in Genesis's research journal.  It makes the Ubaid cultures the sons of Shem before Babel and the Uruk culture the people of Cush and Nimrod, and lowers the second-millennium dates by several centuries.

    1. 34a.

      Section A Detailed Look at the Sons of Shem:  “That the dispersion from the time of Babel occurred at the end of Ubaid 2, where we see the beginning of the northern Ubaid expansion, gains traction with the findings of Hajj Muhammad pottery (Ubaid Phase 2) … at Tell Brak”. ↩

    2. 34b.

      Section The Uruk Period:  “It is in the Late Uruk Period (Uruk IV) that early writing first appears … This is almost certainly not the first writing of mankind but the reinvention of a written form after the Flood and particularly after the dispersion from Babel.”  “as we have placed Nimrod (Enmerkar) in the same period (Uruk IV), it is highly likely that his rule was a significant factor in the early development of writing in Mesopotamia.” ↩

    3. 34c.

      Section The Uruk Period:  “Hammurabi of the Old Babylonian Period (circa 1450 B.C.-revised) did some building there, and it was a functioning center during the Ur III Period before 1660 B.C. (revised).” ↩