Language as a Clue to Prehistory

[lpetrich]

Glottochronology provides a consistent -- albeit it unreliable for dating -- way for guesstimating kinship and separation time. Since there are SOME semantic items (who, me, hand, vagina) that are VERY resistant to change one would expect half a dozen or so ancient word matches, depending on separation time. (Icelandic and Sanskrit mutated only slowly over the centuries -- how well can ancient mutation rates be guesstimated?)

Et cetera. Expanding on this, it turns out that it is the vehement anti-lumpers whose positions are invalid, and who should be denounced at every opportunity.

I agree. It's remarkable how well-preserved 1x *m-, 2x *t-, "that" *t-, "who?" *k- is over northern Eurasia.

Yes, Basque and N. Caucasian (Dagestan?, "macro-Hurrian"?) are related. BUT we want to know WHEN the separation occurred relative to inventions, especially agriculture. Cognates for Certain vocabulary items -- e.g. barley, grain, irrigate -- would be of greatest interest, no?

BRIEFLY Skimming one Bengtson summary yielded only TWO possible "Neolithic" words: ‘threshing board/floor’ and 'vessel for keeping corn.'
Surely there must(???) be several other cognates relevant to Neolithic advances. Are there?

Indeed there are. JB lists a LOT of them in (4) The Anthropological Context of Euskaro-Caucasian | John D Bengtson - Academia.edu

• Domestic animals: bovine, sheep, goat, pig
• Dairying: milk, to milk, butter
• Grains and quasi-grains: wheat, barley, oats, millet, legumes (pea, bean)
• Agriculture: rake, (hoe, ax), pile of harvested grain, threshing floor/board, sieve, (mill, to mill, to grind), (flour, dough)

Among the cognates you mentioned was Basque *larain "threshing floor" ~ PEC *-VrtLV "to thresh" (> "threshing floor", "threshing board") ~ Greek halôs, halôa "threshing floor, garden" (Iliad), "halo around the Sun or the Moon", "disk of the Sun or the Moon", "disk-shaped shield". haloa- "to thresh" (Iliad). > English "halo"

 

[Swammerdami]

Indeed there are. JB lists a LOT of them in (4) The Anthropological Context of Euskaro-Caucasian | John D Bengtson - Academia.edu

• Domestic animals: bovine, sheep, goat, pig
• Dairying: milk, to milk, butter
• Grains and quasi-grains: wheat, barley, oats, millet, legumes (pea, bean)
• Agriculture: rake, (hoe, ax), pile of harvested grain, threshing floor/board, sieve, (mill, to mill, to grind), (flour, dough)

Among the cognates you mentioned was Basque *larain "threshing floor" ~ PEC *-VrtLV "to thresh" (> "threshing floor", "threshing board") ~ Greek halôs, halôa "threshing floor, garden" (Iliad), "halo around the Sun or the Moon", "disk of the Sun or the Moon", "disk-shaped shield". haloa- "to thresh" (Iliad). > English "halo"

This should extinguish any idea that proto-Basque was a pre-Neolithic language. Do the anti-lumpers object to the Euskaro-Caucasian hypothesis?

The anti-lumpers comprise one of the most vivid examples of a consensus of experts being quite wrong. The Stratford side of the Shakespeare Authorship controversy is another example. As are (though a minority except at THIS board) the Carrierites in the Jesus historicity question.

 

[lpetrich]

How about it? Can we guess the level of agriculture when/where Dagestan and Basque split?

Indeed one can.  Neolithic Europe - I'll normalize the dates to BP, where "present" is 2000 CE.

Basque:  Cardium pottery (N Mediterranean, 8,400 BP - 7,500 BP) <  Neolithic Greece (9,000–8,500 BP - 5,200 BP) ,  Pre-Pottery Neolithic B (Fertile Crescent, 10,800 BP - 8,500 BP)

 File:Expansion of farming in western Eurasia, 9600–4000 BCE.png

N Fertile Crescent: 10,500 BP > Armenia 9,500 - 8,500 BP > S Caucasus (Eurasian Georgia) 8,600 BP

N Fertile Crescent: 10,500 BP > S Anatolia 10,200 BP > W Anatolia 8,700 BP > Greece 8,500 BP > Italy 8,000 BP > S France 7,800 BP > E Iberia 7,600 BP > Pyrenees 7,400 BP

 List of domesticated animals -- dog 15,000 BP (terminal Pleistocene) (Eurasia) -- sheep 13,000 - 11,000 BP (Anatolia, Iran) -- pig 11,400 BP (W Asia) -- bovine 10,500 BP (W Asia) - goat 10,000 BP (Iran) -- so the Euskaro-Caucasian ancestral population had all these animals.

Fixed by adding "rye": Grains and quasi-grains: wheat, barley, rye, oats, millet, legumes (pea, bean)

 List of domesticated plants - not as convenient as that list of domesticated animals -- wheat 11,600 BP -- barley 11,000 BP -- rye 11,000 BP (?) -- oats 11,400 BP -- fava bean 10,250 BP -- so this ancestral population had all these crop plants.

 

[lpetrich]

This should extinguish any idea that proto-Basque was a pre-Neolithic language. Do the anti-lumpers object to the Euskaro-Caucasian hypothesis?

Neither Basque nor North Caucasian may be familiar to many of them, the same problem that Austro-Tai has. They also may have gotten annoyed at amateur long-distance comparisons.

The Anthropological Context of Euskaro-Caucasian | John D Bengtson - Academia.edu
One finds a vocabulary that reflects Neolithic technology and Middle Eastern domestic animals and plants.

Metals? JB and Alexei Kassian note at least 6 Proto-North-Caucasian terms: “*ɦĕrVcwĭ ‘silver’, *lŏʒV ̅ ‘a bright metal’, *rĕwcwi ‘red copper ; gold’, *riƛ(w)e ‘brass; gold’, *ṭɨš(w)ɨ ‘lead’, *ṭV ̆ tV(wV) ‘silver ; gold’. Note that none of them possesses Basque cognates”

Metal-smelting sites in the Caucasus Mountains start to appear around 7,000 BP (5,000 BCE).

For Indo-European: Loren Petrich's answer to Why aren't there common words in Indo-European languages for metals? Does that indicate that Indo-European languages arose in the Neolithic Saraswathi Valley? - Quora

 

[lpetrich]

Back to Response to comments "On the quantification of Euskaro-Caucasian lexical matches ..." (09/27/2022 | John D Bengtson - Academia.edu

Daniel França: Bad argument. Show me one other language family that we can reconstruct at the 13,000 year level? One!

JDB REPLY: Is it Afro-Asiatic? Furthermore, the idea that a language family has to be reconstructed before it is “proven” is a fallacy.

Then noting some early Indo-Europeanists stating that Indo-European was first recognized as a result of mass comparison, rather than explicit reconstruction.

JB closed by listing some Basque - North Caucasian sound correspondences.

As to AA, its uncontroversial reconstruction is very limited, with two big dueling ones: the Ehret one and the Orel-Stolbova one.

 

[lpetrich]

I checked on John Bengtson's magnum opus, "Basque and its Closest Relatives" (available online at academia.edu), for what it had to say about Kartvelian (South Caucasian).

Page 8 / 38: JB notes the work of CC Uhlenbeck back in 1924: Basque + ...

• NC: 42
• K+NC: 19
• K: 4

Georges Dumezil: Basque's grammar has a much better fit with NC's grammar than with K's grammar.

Page 484 (514) has a table of Proto-Kartvelian forms for JB's Basque-NC comparisons. Here is some of that table, with IE forms also:

• "dog": B hor, Avar hoy, Budukh qhor, PNC *qhhwey-rV- PK *zhaghl-, PIE *kwon-
• "ear": B belarri, Chechen lerg, Dargwa lihhi, PNC *leHi, PK *q'ur-, *stum-, PIE *h2ôws
• "tongue": B minhi, Andi mits':i, Tabasaran melz, Chechen mott, PNC *melci, PK *nena, PIE *dnghu-

So Kartvelian is not very close to either Basque or North Caucasian.

Kartvelian's pronouns: 1s *me, 1p *chwen, 2s *shen, 2p *tkwen -- the 1s and 2s look like northern Eurasian 1s *m- and 2s *t- (*s-).

North Caucasian ones are more difficult, but they more-or-less fit. For the first person singular, one finds nominative *zo- and oblique *n- or *m-. Basque has ni, generalizing the oblique form.

 

[lpetrich]

The deep history of the number words | Philosophical Transactions of the Royal Society B: Biological Sciences

We have previously shown that the ‘low limit’ number words (from one to five) have exceptionally slow rates of lexical replacement when measured across the Indo-European (IE) languages. Here, we replicate this finding within the Bantu and Austronesian language families, and with new data for the IE languages. Number words can remain stable for 10 000 to over 100 000 years, or around 3.5–20 times longer than average rates of lexical replacement among the Swadesh list of ‘fundamental vocabulary’ items. Ordinal evidence suggests that number words also have slow rates of lexical replacement in the Pama–Nyungan language family of Australia. We offer three hypotheses to explain these slow rates of replacement: (i) that the abstract linguistic-symbolic processing of ‘number’ links to evolutionarily conserved brain regions associated with numerosity; (ii) that number words are unambiguous and therefore have lower ‘mutation rates’; and (iii) that the number words occupy a region of the phonetic space that is relatively full and therefore resist change because alternatives are unlikely to be as ‘good’ as the original word.

Mark Rosenfelder's The Numbers List has only 1 to 10, but also protoforms. IE and AN have protoforms for 1 to 10, but Bantu has only 1 to 5 and also 10.

Checking on this work's results for the most stable word forms, I find the usual suspects, like pronouns, "eye", "tongue", "to die", etc., though in different orders, and also some oddities, like Bantu having "hunger" and "elephant" as highly stable.

 

[Swammerdami]

It is interesting that the early Neolithic languages of Cardial Ware (surviving as Basque in the Pyrenees mountains), Fertile Crescent (surviving as Dagestan et al in the Caucasian mountains) and the Harappan civilization (surviving perhaps as Burushaski in the Hindu Kush) seem to be linked together. and are part of the large Dene-Caucasian family. What about the "Fertile Crescent language"? Could that be Sumerian?

I came across this paper, "Lexical Matches between Sumerian and Hurro-Urartian: Possible Historical Scenarios", which may appeal since it applies several ideas of applied mathematics. Since both languages are long extinct, only 65 pairs from a 110-word Swadesh list are available for comparison. Only 6 matches show up: words for dog, hand, liver, meat, rain, who). But the paper goes on and on in great detail and I haven't read it. I wish I had time to read it, as the paper offers linguistic insights completely unrelated to the two target languages.

The paper is by A. Kassian, an obscure Russian linguist (he did co-author an informal review with Starostin). Here's another paper he wrote:

A. Kassian Two Middle Hittite Rituals Mentioning Ziplantawij
describes a magic ritual performed by an Old Woman with her several female assistants in order protect the sitting king Tuthalija II.

The abstract to Lexical Matches between Sumerian and Hurro-Urartian contains a summary of Kassian's conclusions:

Four possible scenarios are evaluated from the typological, etymological and statistical points of view:
(1) chance coincidences;
(2) lexical borrowings from Sumerian into Hurro-Urartian or vice versa;
(3) genetic relationship between Sumerian and Hurro-Urartian;
(4) prehistoric language shift: adoption by a Hurro-Urartian (or closely related) group of the Sumerian language or vice versa.

Out of these four, two scenarios—lexical borrowings and genetic relationship—are typologically unlikely. The statistical probability of chance coincidences is low, although formally this explanation cannot be excluded. The fourth scenario—language shift—fits linguistic evidence and does not contradict archaeological data.

 

[lpetrich]

Bayesian phylogenetic analysis supports an agricultural origin of Japonic languages - PubMed - that's Japanese and the languages of the Ryukyu Islands

Farming/language dispersal theory: were some language families spread by farming?

Diffusion/transformation theory: did people learn to farm from their farming neighbors?

The latter is a plausible scenario for Kartvelian speakers, for instance, learning to farm from Anatolian farmers who decided to settle north of the Caucasus Mountains.

So was Proto-Japonic brought to Japan by farmers around 1700 - 2400 years ago (YBP: "years before present")? Or was it brought by the first settlers around 12,000 - 30,000 YBP?

The authors used lists of words for body parts, kinship terms, basic verbs, numbers and pronouns -- Swadesh and Leipzig-Jakarta territory.

We made cognate judgements by (i) relying on previously identified sound correspondences that were used for reconstructing proto-Japonic ([25]; B. E. Riley 2003, unpublished PhD thesis; J. B. Whitman 1985, unpublished PhD thesis); (ii) working out systematic sound correspondences based on the comparative method [26]; and (iii) consulting previously published cognate judgements in glottochronological studies on Japonic languages [27,28].

The authors then used some phylogeny software developed for genes and proteins.

For Old Japanese, they used a date range of 1216 - 1300 YBP, and for Middle Japanese, 437 - 674 YBP. Another calibration point was for when the Tokugawa leaders moved thir capital city from Kyoto, capital since 1200 YBP, to Edo, now Tokyo, in 407 YBP.

They found a range of dates for the Japanese - Ryukyuan divergence: median 2182 YBP, mean 2398 YBP, standard error 47.21 y, 95% confidence limit: 1230 - 4190 YBP.

That fits the farmer theory very well.

 

[lpetrich]

To investigate the reliability of the result, another set of analyses was carried out with an independent set of Japonic language data compiled by Starostin.

Sergei Starostin's Altaic Etymological Dictionary

We acknowledge that Starostin's data are considered controversial by some scholars, as he made debatable reconstructions of proto-Japonic and used them to argue for genetic relationships to other equally debatable proto-languages (e.g. proto-Tungusic). However, this was not deemed to be a major problem for the current purpose, as those controversial reconstructed lexicons are excluded from our analyses.

Using 110-word lists for 9 Japonic languages, they found this range of dates: median 1976 YBP, mean 2080 YBP, standard error 9.13 y, 95% confidence limit: 1232 - 3279 YBP.

They use a date of 2400 YBP for the colonization of Japan and the Ryukyu Islands by Proto-Japonic speakers.

A potential problem with any phylogenetic approach to lexical data is a horizontal transmission of words between languages. This is particularly problematic since it is difficult for one to be absolutely sure that every single borrowed word has been detected. However, a recent simulation study suggests that the amount of undetected borrowing needs to be unrealistically high (greater than 20% per 1000 years) to invalidate time estimation or tree topology.

 

[lpetrich]

I like Alexei Kassian's work with George Starostin and Mikhail Zhivlov and others on macro-linguistics.

• Rapid radiation of the inner Indo-European languages: an advanced approach to Indo-European lexicostatistics
• Proto-Indo-European-Uralic Comparison from the Probabilistic Point of View | George Starostin, Alexei S . Kassian, and Mikhail Zhivlov - Academia.edu - success to within about 1% probability of coincidence
• Permutation test applied to lexical reconstructions partially supports the Altaic linguistic macrofamily | Evolutionary Human Sciences | Cambridge Core - Turkic, Mongolic, Tungusic: yes, Korean, Japonic: too weak
• Calibrated weighted permutation test detects ancient language connections in the Circumpolar area (Chukotian-Nivkh and Yukaghir-Samoyedic) [JHL 2023] | Alexei S . Kassian, George Starostin, and Mikhail Zhivlov - Academia.edu
• The “Nostratic” roots of Indo-European: from Illich-Svitych to Dolgopolsky to future horizons (2016) | Alexei S . Kassian, George Starostin, and Mikhail Zhivlov - Academia.edu - a critical look at macro-linguistics research

 

[lpetrich]

On Forni’s Basque–Indo-European Hypothesis [JIES 2013] | Alexei S . Kassian - Academia.edu - finding his proposed comparisons very weak.

For instance, for "tongue", Proto-Basque *minhi is much closer to Proto-North-Caucasian *melci than to Proto-Indo-European *dng'hwâ


Arnaud Fournet & Allan R. Bomhard - The Indo-European Elements in Hurrian (2010; corrected 31 January 2023) | Allan Bomhard - Academia.edu

Review of: The Indo-European Elements in Hurrian, by Arnaud Fournet & Allan R. Bomhard (2010) | Alexei S . Kassian - Academia.edu - finding that hypothesis weak, and instead proposing a Dene-Caucasian affinity: 1s, 2s, "black", "blood", "louse"

Also at [Review] A.Fournet, A.R.Bomhard. The Indo-European Elements in Hurrian (2010)


On the Burushaski-Indo-European hypothesis by I. Čašule | John D Bengtson - Academia.edu - finding it very weak and finding Dene-Caucasian affinity much stronger

Also at On the Burushaski-Indo-European hypothesis by I.Casule

Ilija Čašule responded:
The Indo-European Origin of the Burushaski Language and the Dene-Caucasian Hypothesis | Ilija Casule - Academia.edu

 

[lpetrich]

Hurro-Urartian from the lexicostatistical viewpoint [UF 42, 2010–2011] | Alexei S . Kassian - Academia.edu

Of his 110-word list, 65 are recorded for Hurrian, 22 for Urartian, and all but one of the Urartian ones are also in Hurrian. So the attestation of those two langs' words is not statistically independent (chi-square : 15.0427, p-value : 0.000105). So the attested words tend to be the same sorts of words in both langs.

Of the 21 in both, 15 are cognates, 1 doubtful, and 5 not cognates, at least by resemblance.

Looking for Dene-Caucasian cognates, AK found:

Overall 12, North Caucasian 9, Yeniseian 8, Sino-Tibetan 6, Burushaski 3, Basque 1

AK found three additional ones in the Nakh subfamily of the Northeast Caucasian langs (Nakh-Dagestanian).

All these cognates tend to be in the more stable parts of the Swadesh and Leipzig-Jakarta lists. It would be interesting to extend this work to the rest of the known Hurrian vocabulary. There is a curious shortage of Basque cognates, given all the Basque-NC ones, though some of this is from splitting the 1s and 2s pronoun declensions, with Hurrian getting the direct-case parts and with Basque getting the oblique-case parts.

Along the way, AK notes these resemblances:

Semitc 2, Uralic 1, Indo-European 2, Altaic 1, Omotic 1

Essentially random.

 

[lpetrich]

BASQUE AND HURRIAN | iurii mosenkis - Academia.edu - proposing some comparisons, with exact ones being for "dog", "hand", and "skin". I looked in Alexei Kassian on Hurrian and Sumerian, and John Bengtson's big book on Basque and North Caucasian:

• PBsq *hor "dog" - Su ur, Hu erw/bi, NC *qhhwei-
• Bsq *esku "hand" - Su su, Hu suni, NC *tsëgwV ~ *gwëtsV (NC: "arm")
• PBsq *asal "skin" - Hu asxe, NC *ʔwarcwä (NC: skin > color)

Lexical Matches between Sumerian and Hurro-Urartian: Possible Historical Scenarios - Cuneiform Digital Library Initiative - 5 or 6 cognates, including "dog" and "hand".


On Hattic,
Journal of Caucasian Studies » Submission » Northwest Caucasian Languages and Hattic [Kuzeybatı Kafkas Dilleri ve Hattice] - by Ayla Applebaum - no list of Hattic - NWC cognates, and only one example: "mountain". She found about 3 to 6 matches out of the 13 in her list.

She ought to have also compared Hattic to some reconstruction of Proto-West-Caucasian, because that would be the closest in time.

Also
Hattic as a Sino-Caucasian language [UF 41, 2009–2010] | Alexei S . Kassian - Academia.edu

Most of the Northwest Caucasian comparisons also work for North Caucasian and even Dene-Caucasian, though he compares only North Caucasian, Yeniseian, and Sino-Tibetan as Sino-Caucasian. He finds a firm relationship with NC and Yen, and with those two together, but not as strong with ST. For the Swadesh-Starostin 110-word list,

North Caucasian 18, Yeniseian 14, Sino-Tibetan 12, Burushaski 9, Basque 4.

 

[lpetrich]

Someone else did a statistical analysis of Altaic and came up with different results from Kassian, Starostin, et al. DSignificance testing of the Altaic family at Andrea Ceolin

Furthermore, I decided to make the unconventional choice of focusing on modern languages rather than reconstructed proto-forms. The main reason for this methodological choice is to avoid the problems of significance tests as the one presented in Kassian et al. (2015), who argued in favor of an Indo-Uralic macrofamily on the basis of phonetic similarities in the basic vocabulary of Proto-Indo-European and Proto-Uralic. The paper shows there are some degrees of freedom in the selection of reconstructed proto-forms, even in a well-studied domain like Indo-European.

Dumb decision. It increases the risk of false negatives because of language changes. A way to get around that is to use multiple attested languages, but if one does one's comparison work carefully, one ends up rediscovering protolanguages.

Thus, he decided to compare Turkish, Mongolian, and Manchu, and no other members of Turkic, Mongolic, or Tungusic.

What he found:

The results only partially point towards an Altaic family: Mongolian and Manchu show significant sound correspondences, while Turkish and Mongolian show some marginally significant phonological similarity, that might however be the consequence of areal contact. Crucially, Turkish and Manchu do not test positively under any condition.

However, Permutation test applied to lexical reconstructions partially supports the Altaic linguistic macrofamily | Evolutionary Human Sciences | Cambridge Core finds much better significance, and did so using protolanguages.

 

[lpetrich]

Let's consider "water" and "name" in both Indo-European and Uralic. I'll use Wiktionary as my main resource.

Indo-European:

• Present-day: English water, name - German Wasser, Name - Swedish vatten, namn - French eau /o/, nom /noN/ - Spanish agua, nombre - Italian acqua, nome - Romanian apa, nume - Irish uisce /iSki/, ainm /anyamy/ - Welsh dwr, enw - Polish woda, imie (stem imien-) - Czech voda, jmeno - Serbo-Croatian voda - ime (stem imen-) - Bulgarian voda, ime - Russian voda, imja (stem imjen-) - Lithuanian vanduo, vardas - Greek nero, onoma (stem onomat-) - Persian ab, nâm - Hindi pâni, nâm - Punjabi pânî, nâ - Bengali pani, nam - Sinhalese jalaya, nama
• 1000 CE: Old English waeter, nama - Old High German wazzar, namo - Old Norse vatn, namn - Gothic watô, namô - Old French ewe, nom - Old Spanish agua, nomre - Old Irish uisce, ainmm - Old Church Slavonic: voda, ime (stem imen-) - Greek neron, onoma (stem onomat-) - Aprabhamsa paniu, nâm(a)
• 1 CE: Latin aqua, nômen (stem nômin-) - Greek hudôr (stem hudat-), onoma (stem onomat-) - Middle Persian ab, nâm - Prakrit pânîa, nâm(a)
• 1000 BCE: Old Persian ap-, nâma - Sanskrit ap, jala, pânîya, , nâma (stem nâman-)
• 2000 BCE: Hittite wâtar, (stem witen-), lâman

Uralic: Present-day: Finnish vesi, nimi - Mari vüd, lüm - Hungarian viz, név

Let's see what using protolanguages will get us:

• 1 CE: Proto-Germanic *watôr (stem *watin-), *namô (stem *namn-) - Proto-Celtic *udenskyos, *anman - Proto-Slavic *voda, *jime (stem *jimen-)
• 1000 BCE: Proto-Balto-Slavic *wandô, *in'men
• 2000 BCE: Proto-Indo-Iranian *ap-, *nâma
• 4000 - 3000 BCE: Proto-Indo-European *wodr (stem *wedn-, *uden-), *h1nomn (*ënomn)

Proto-Uralic: 3000 BCE: *wete, *nime

Much easier. The original Indo-European word for water - *wodr - got replaced several times: Latin aqua, Medieval Greek neron, Proto-Indo-Iranian *ap, Sanskrit pânîya, and Sinhalese jalaya. The original IE word for name - *h1nomn - got replaced only once, in Baltic: Lithuanian vardas, Latvian vards, cognate with English "word", German Wort, Latin verbum.

So I think I'll stick to comparing protolanguages.

 

[lpetrich]

I'll fold attested langs and protolangs together:

Indo-European:

• Present-day: English water, name - German Wasser, Name - Swedish vatten, namn - French eau /o/, nom /noN/ - Spanish agua, nombre - Italian acqua, nome - Romanian apa, nume - Irish uisce /iSki/, ainm /anyamy/ - Welsh dwr, enw - Polish woda, imie (stem imien-) - Czech voda, jmeno - Serbo-Croatian voda, ime (stem imen-) - Bulgarian voda, ime - Russian voda, imja (stem imjen-) - Lithuanian vanduo, vardas - Greek nero, onoma (stem onomat-) - Persian ab, nâm - Hindi pâni, nâm - Punjabi pânî, nâ - Bengali pani, nam - Sinhalese jalaya, nama
• 1000 CE: Old English waeter, nama - Old High German wazzar, namo - Old Norse vatn, namn - Gothic watô, namô - Old French ewe, nom - Old Spanish agua, nomre - Old Irish uisce, ainmm - Old Church Slavonic: voda, ime (stem imen-) - Greek neron, onoma (stem onomat-) - Aprabhamsa paniu, nâm(a)
• 1 CE: Latin aqua, nômen (stem nômin-) - Greek hudôr (stem hudat-), onoma (stem onomat-) - Middle Persian ab, nâm - Prakrit pânîa, nâm(a) - Proto-Germanic *watôr (stem *watin-), *namô (stem *namn-) - Proto-Celtic *udenskyos, *anman - Proto-Slavic *voda, *jime (stem *jimen-)
• 1000 BCE: Old Persian ap-, nâma - Sanskrit ap, jala, pânîya, , nâma (stem nâman-) - Proto-Balto-Slavic *wandô, *in'men
• 2000 BCE: Hittite wâtar, (stem witen-), lâman - Proto-Indo-Iranian *ap-, *nâma (stem nâman-)
• 3000 BCE: Proto-Indo-European *wodr (stem *wedn-, *uden-), *h1nomn (*ënomn)

Uralic:

• Present-day: Finnish vesi, nimi - Mari vüd, lüm - Hungarian viz, név
• 3000 BCE: *wete, *nime

I've thought some more, and using protolanguages can get us far into the past - Proto-Indo-European is reasonably well dated to around 6,000 - 5,000 BP (4000 - 3000 BCE) with the help of archeological evidence. That's in the middle of the Holocene, and around when the Sumerians and Egyptians invented writing.

 

[lpetrich]

I checked on the inflection of the Uralic words, and here is for Finnish:

• nom. sg. vesi, nimi
• nom. pl. vedet, nimet
• obl. sg. vede-, nime-
• obl. pl. vedi-, nimi-

Hungarian: viz pl. vizek, név pl. nevek -- (Mordivinic) Erzya ved' pl vedt', lem pl. lemt -- E Mari vüd pl. vüdvlak, lüm, lümvlak -- (Permic) Komi va pl. vajas, nim pl. nimjas -- E Khanty jënk (cognate with words for "ice"), näm -- N Mansi vit du. vityg pl. vityt, nam du. namyg pl. namyt -- N Sami tSatsi obl tSaz- pl tSazit pbl tSaz-, namma obl nama- pl namat obl namai-

j is the "y" sound here: yet, not jet.

Proto-Uralic had dual *-k plural *-t oblique *-i- . Most Uralic langs dropped the dual, and they usually also kept the plural, but Hungarian speakers turned the dual ending into a plural ending. Some langs have different plural forms: Mari -vlak, Komi -jas.

 

[lpetrich]

The deep history of the number words | Philosophical Transactions of the Royal Society B: Biological Sciences by Mark Pagel and Andrew Meade

The most stable word forms that they found in these families:

• Indo-European (both): 2, 3, 5, who?, 4 - 1SG, 1, 1PL, when?, tongue - name
• Bantu (both): to eat, tooth, 3, eye, 5 - hunger, elephant, 4, person, child - 2
• Austronesian (paper): child, 2, to pound/beat, 3, to die - eye, 4, 10, 5, tongue - 8
• Austronesian (supp.): 2, 3, to die, eye, 4, - 50, 10, 7, 5, tongue - 8
  • 
    where (paper) is from the main body of the paper, and (supp.) from the supplementary data. However, "child" and "to pound/beat" are also in the supplementary data.
    
    Rather mixed bag. Outside the numerals, most of these words are in lists of highly-stable words like the Swadesh and Leipzig-Jakarta lists, but their order is different, and some members are not in such lists.
    
    Their main result is that the words for the lowest positive integers, words for 1 to 5, are super stable. My experience is that 1 to 10 are often very stable, 11 to 99 often have special forms with limited stability, 100 is also stable, and 1000 is somewhat less stable.
    
    The history of number words in the world's languages—what have we learnt so far? | Philosophical Transactions of the Royal Society B: Biological Sciences
    

    In contrast with low-limit numbers, the higher numbers are characterized by a rapid and morphologically consistent pattern of expansion, and behave like grammatical phrasal units, following language-internal rules.

    Mostly formed as compounds or other derivations (PIE *kmtom 100 as "big 10": *(d)km-tom < *dekm + *-tom). Words for powers of the base are often lexically separate, however.

 

[lpetrich]

Reconstructing prehistoric languages | Philosophical Transactions of the Royal Society B: Biological Sciences - 22 March 2021

The vocal tract as a time machine: inferences about past speech and language from the anatomy of the speech organs | Philosophical Transactions of the Royal Society B: Biological Sciences - about a test run on some human skeletons from medieval Netherlands.

Preferred sound groups of vocal iconicity reflect evolutionary mechanisms of sound stability and first language acquisition: evidence from Eurasia | Philosophical Transactions of the Royal Society B: Biological Sciences

From consonant strength to sonority:

voiceless stops (p, t, k), voiced stops (b, d, g), voiceless fricatives (ph/f, th, kh), voiced fricatives (bh/v, dh, gh), nasals (m, n, ng), lateral liquids (l), central liquids (r), glides/approximants (y, w), high vowels (i, u), mid vowels (e, o), low vowels (a)

We conclude that the stability values of sound groups are highly diverging, from almost complete stability (labials, laterals, nasals) to almost complete instability (central vowels, palatals, glottals). The most striking result is the difference in stability between consonants and vowels, where consonants have higher stability values.

Consonants more stable than vowels - a common finding in historical linguistics.

The sounds of prehistoric speech | Philosophical Transactions of the Royal Society B: Biological Sciences - basically collecting a lot of previous work.

Inferring recent evolutionary changes in speech sounds | Philosophical Transactions of the Royal Society B: Biological Sciences

We suggest that the large-scale borrowing of phonologically contrastive speech sounds (also known as ‘phonological segments’), through lexical borrowing, may have made some speech sounds more prevalent in the world's languages, thereby levelling to some extent earlier areal-specific profiles.

Nothing that both very rare sounds and very common sounds are seldom borrowed, and that one should look in between.

First, we find substantial differences between ancient and reconstructed languages, on the one hand, and present-day languages, on the other. Specifically, the distributions of specific speech sounds, in particular fricatives and affricates, have increased over time.

Second, and crucially, the greatest disparities between earlier and present-day distributions turn out to be largely owing to those sounds that have spread in the world's languages owing to relatively recent borrowing events.