Language as a Clue to Prehistory

[bilby]

Dogs are not usually called woof-woof or arf-arf, for instance.

The Russians say 's a barker

 

[lpetrich]

Puluga: The Most Widespread Name of the Divinity, by Alfredo Trombetti, translated and reprinted in Mother Tongue XVII

In addition to the Andaman Islands, the name occurs in the territory of Eurasia from the Baltic to Kamchatka, in Africa, especially West Africa, and lastly in New Guinea, Australia and Tasmania. Further research may add new connections and in particular, the presence of the name in Kamchatka makes it appear likely that it will also be found in the Americas, all the more so since the conception of the divinity as the personification of thunder or in general, of the frightening forces of nature, appears to be very widespread.

In other words, names of storm gods.

Andaman Islands:
Ongan: Onge: vluge, oluga
Great Andamanese: Bea, Bale: puluga - Jeru: biliku - Puchikwar: bilik - Chari: bilek - Juwol, Kol, Kede: bilak

Austroasiatic: Khasi–Palaungic: Khasi: blei, brei, blâi, prâi - Palaung: para, pra
Austroasiatic: N Aslian (Jehaic, Semang): pîe
Kra-Dai: Tai: Ahom: phrî - Khamti: phrî

Tasmanian: berik, burak
Austronesian: M-P: Oceanic: Keapara: palagu
Pama-Nyungan: (Walsh River) pirku-ir - (btwn Nicholson R & coast) pargi-gi

Niger-Congo (Togo): buruku - (Gold Coast): bluku - Sobo (Isoko? Urbobo?): blugwe
Niger-Congo: Bantu: Pol: bulgu
Afroasiatic: Cushitic: Oromo (Galla): bulgu

Chukchi-Kamchatkan: Itelmen (Kamchadal): biluka-i
Turkic: Common Turkic: Uyghur: buraxa-n
Mongolic: Buryat: burku-ng

But the last one of these I could identify.
Sanskrit Parjánya, Lithuanian Perkūnas "storm god" > Finnish perkele "devil" (under Christian influence)
From PIE *Perkwunos
Related to *perkwus "oak tree" > Latin quercus, Proto-Germanic *furhô "fir, pine tree" > English "fir"

There's a shared root shape: P - L/R - K

 

[lpetrich]

From that Trans-New-Guinean paper: *(m,mb)elak

It states Proto-Malayo-Polynesian *bilak means "lightning" but Austronesian Comparative Dictionary states "to shine, glitter", though some of its descendants do mean "lightning". The PMP word reconstructed as "lightning" is *kilat and for Proto-Austronesian *likaC.

Turning to Indo-European again, this root not appears as *perkw- but also as *bhel(g)-, *bhleyg- "shiny, white"

• *bhel- > Proto-Slavic *belu "white"
• *bhelg- > Latin flamma "flame" - fulgur "lightning" - fulgêre "to blaze, shine" > fulmen "lightning" - flagrâre "to burn, blaze"
• *bhleyg- > Proto-Germanic *blaikaz, *blîkanan, *blinkanan > *blankaz
  • *blaikaz > English "bleak", "bleach", German Blitz "lightning"
  • *blinkanan > English "to blink"
  • *blankaz > English "blank", French blanc, Spanish blanco, Italian bianco "white"

English "flash" - origin rather obscure.

This root is also present in Proto-Semitic: *barak- "lightning"

Alfredo Trombetti continued, citing plenty of Indo-European and Semitic examples, though with a curious lack of PIE and PS forms. He mentioned Uyghur buraxan (burakhan) as meaning "storm", and I checked Wiktionary. I found some Turkic langs having boran "blizzard, snowstorm", and from there borrowed into Russian as buran.

I checked in The Tower of Babel and I found Proto-Turkic *buruk "dust, smoke, to blow (snowstorm)". It is related to similar words for "dust, smoke" in Mongolic and Tungusic, and is likely a suffixed form of Eurasiatic *bur- "dust". I also find Uralic *purkV "snowstorm" < Eurasiatic *bur- "storm". Related meanings? Originally something like *buruk- ?

So we have a possible Proto-World root.

 

[lpetrich]

Looking in Tower of Babel Databases for Dravidian, to help out Wiktionary, I find

father: *appa *tantay - mother: *amma *taḷḷay
older brother: *aṇṇa - older sister: *akka
younger brother: *tampa - younger sister: *tamkay
son: *makanṯu - daughter: *makaḷ

The younger siblings' words look somewhat similar, like the children's words.

Colors:
red: *kem-
green, yellow *pac-
blue: borrowed from Sanskrit nîla "dark blue, dark green, black"

Some words for yellow come from words for turmeric, a plant whose rhizomes are ground up to make an orange-yellow powder that is valued as a dye and as a coloring and flavoring agent for food.

 

[Swammerdami]

Puluga: ...
In other words, names of storm gods.
...
Great Andamanese: Bea, Bale: puluga - Jeru: biliku - Puchikwar: bilik - Chari: bilek - Juwol, Kol, Kede: bilak
...
Austronesian: M-P: Oceanic: Keapara: palagu

Any chance Ba'al is cognate?

Baal or Ba'al: A storm and fertility god of the Phoenician and Canaanite pantheons, reckoned as chief of the gods by the 1st millennium BC.
Etymology
From Late Latin Baal (as in the Vulgate) and Ancient Greek Βάαλ (Báal); from Hebrew בַּעַל‎ (bá`al, “lord, husband, owner”), Phoenician 𐤁𐤏𐤋‎ (bʿl, “lord, master, owner”) and Ugaritic 𐎁𐎓𐎍 (baʿlu, “lord, owner”), all from Proto-Semitic *baʿl- (“owner, lord, husband”).

...
Sanskrit Parjánya, Lithuanian Perkūnas "storm god" > Finnish perkele "devil" (under Christian influence)
From PIE *Perkwunos

Storm God Perkwunos > thunder > percussion?
Not according to https://www.etymonline.com/word/percussion which makes this interesting comment:
"the modern English word [quash] is a merger of two words, both in Middle English as quashen, from two unrelated Latin verbs."

 

[lpetrich]

percussion - Wiktionary, the free dictionary
From Old French < Laitn percussiō "striking" < percutiō "I strike" < per- + quatiō "I shake, smash, wield, move, harass"

per- makes adjectives and verbs that are "very" something.

per "through" < PIE *per- "before, in front" > English "for", Greek peri, ...

quatiō < PIE *kweh1t- "to shake"

Looking in TOB Databases and looking for "shake" in the Indo-European one, I find

*(s)kwat-, *(s)kAut-

with quatiō for it.

In that database, I found PIE *perkw- by looking for Latin quercus "oak".

I expanded the references and I found *perkwus (u-stem) with English meaning "oak" and German meaning Eiche (cognate).

Derivatives: daraus vielleicht `Stärke, Kraft, Leben, Weltenbaum' und `Baum des Donnergottes Perkʷuno-s'

Google Translate from German: from this perhaps `Strength, Power, Life, World Tree' and `Tree of the Thunder God Perkʷuno-s'

 

[lpetrich]

Turning to the TOB Databases about PIE *bhel(g)- *bhleyg- I find
*bhela- "light, bright"
*bhalge-, *bhlage- /-e- "to flame"

Under "Nostratic etymology" I find

Eurasiatic: *belV

Uralic *peĺkkä (?) "clean"
- Ostyak (Khanty) paɣǝl "lightning"

Altaic *bĕ̀ló "pale"
- Mongolic: *balaj "blind, dark"
- Tungusic: *beli "pale, to whiten"
- Korean: *pằrk- "bright"
- Japonic: *pàrá- "to clear up (of sky, weather)"

Kartvelian *berc̣q̇ / brc̣q̇- "to glitter"

Dravidian *veḷ- "white"

Comments: Cf. also forms under *ṗaĺV 'burn' and Mong. *büli-ɣen 'warm' (see ND 195 *bil̄V 'warm') - a great deal of confusion. In ND 199 *bal̄[ʕ]V 'blind' the Mong. form is separated and compared with East Cush. and Eg. *balʕ- 'blind' (?), as well as some IE forms (contaminated within *bhlendh-?).

Mentions alternate possibilities *bal̄[ʕ]V, *bal̄(i)ka.

Seems like it could be *balak

 

[lpetrich]

Borean (approx.) : PVLV

Afroasiatic *bVlVg- / *bVlVḳ- "to shine"
Semitic *bVlug- "to shine, dawn"

However, in TOB, Semitic *bariḳ- "lightning" is listed as a separate root.
Some of its possible Afroasiatic cognates are noted to be at least possible borrowings from Arabic.

Back to that putative Borean root.
Austric: Austronesian: *balar 'pale, albino', *balaR 'pale, white'
However, the Austronesian Comparative Dictionary has
Austronesian *puNi "whiteness"
Malayo-Polynesian *balaR "pale, unnaturally white" *budaq, *bulan, *bulaR "unnaturally white; albino", *(ma-)putiq "white"

Also lists Amerind *pali "Sun, to shine", *pole "white" (couldn't interpret the abbreviations of families - nothing matching in Wiktionary for Guarani, Nahuatl, O'odham, Ojibwe, Quechua)

Also Bantu *-bád- "to shine" (no Bantu examples in Wiktionary)

So there seems like some widespread root p/b - l/r - (k), though I concede the risk of coincidence and borrowing. "The Most Widespread Name of the Divinity"? I'll settle for "bright, white, to shine".

"White" is in the Swadesh list of highly-conserved meanings, and it's also in the longer Leipzig-Jakarta lists, even if not in the 100-word one or the Dolgopolsky lists.

 

[lpetrich]

Cognate maximization versus cognate minimization: in search of a "golden middle" for Altaic etymology | Studia Uralo-altaica - by George Starostin - "in search of a "golden middle" for Altaic etymology" - PDF file

GS praised his late colleague Eugene A. Helimski for "his critical evaluation of overpermissive semantic standards in these studies," referring to macro-linguistics, saying that "(his) extremely sober, yet generally sympathetic attitude to such hypotheses as Altaic and Nostratic was almost unique among his colleagues, most of whom tended to sway too far in either of the two opposite directions,"

He notes that reports of the death of Altaic in mainstream linguistics are somewhat exaggerated, pointing to the recent work of Martine Robbeets. "Yet it may be argued that few, if any, significant methodological breakthroughs in Altaistics have been achieved over the past 15 years — breakthroughs, that is, which could decisively turn the wheel in one or the other direction and result in either abandoning the hypothesis for good or, conversely, producing a unified working model of the Altaic theory which could serve as a base reference point for generations of future scholars to come."

Then discussing the Etymological Dictionary of Altaic Languages (EDAL) by Sergei Starostin, Anna Dybo, and Oleg Mudrak in 2003, with its 2800 proposed etymologies. That's huge. For instance, it has 1115 matches between Turkic and Japonic, while Julius Pokorny's 1959 IE etymological dictionary contains 600 - 650 for Sanskrit and Latin.

But that work has many critics who consider many of its etymologies weak. Like "jerboa" ~ "otter" ~ "dolphin". I'd expect dolphins to be considered a kind of fish, because of their very fishlike body shape and habits. Consider English "whale" ~ Latin squalus "big sea fish (shark?)"

EDAL seems to have cognate maximization, while opponents of Altaic do cognate minimization, or cognate elimination.

Arguably the most focused and successful work in establishing such a “golden middle” in the post-EDAL era has been conducted by Martine Robbeets, a firm adherent of the Altaic hypothesis who has, nevertheless, been openly and highly critical of the cognate maximization strategy.

GS asks where one draws the line.

 

[lpetrich]

George Starostin continues with "onomasiological reconstruction" - semantic reconstruction, attempting to reconstruct which words have which meanings, rather than being imprecise about the meaning of a reconstructed form.

One can then look for matches, and also mismatches that are related with plausible semantic shifts.

GS demonstrates this procedure for words for "head" and "ear", finding not only "head" but also "forehead" and "brain", and not only "ear" but "to hear". The latter is *k'ûylu- and it looks much like Proto-Indo-European *klew- and Proto-Uralic *kuwli- "to hear".

But the very fact that onomasiological scenarios like these appear to be working for Altaic, and allow for the proposed etymologies to be presented not as random, disconnected bunches of phonetically and semantically similar comparanda, but as parts of a coherent network, within which we try to account not only for potential archaic retentions, but also for innovations which replace them, should be enough to significantly boost confidence in the hypothesis.

GS proposes

1. Start out with some list of meanings whose words are relatively stable
2. Do semantic reconstruction on well-defined subfamilies
3. Do semantic reconstruction using these reconstructions

What would be good evidence?

"1. No matter the size of the original list (as long as it is not specifically dominated by unstable cultural lexicon), many, if not most, of its elements have to be represented by a hypothetical Proto-Altaic reconstruction."

Example of plausible semantic shifts: "if a certain etymon means ‘tree’ in one proto-language and ‘wood/k. of wooden object’ in all other languages, semantic typology indicates that ‘tree’ must probably have been the original meaning)"

"2. Not a single element of the list must be represented by more than one hypothetical Proto-Altaic reconstruction, at least not without a substantial reason (though the opposite is certainly permissible, since merging similar meanings such as ‘hand/arm’, ‘body hair/head hair’, ‘eat/drink’ etc. in one lexical root is a very common practice)."

WALS Online - Feature 129A: Hand and Arm - Same word or different words?

WALS Online - Feature 130A: Finger and Hand - Same word or different words?

WALS Online - Feature 132A: Number of Non-Derived Basic Colour Categories and WALS Online - Feature 134A: Green and Blue and WALS Online - Feature 135A: Red and Yellow

He concludes "... but unless the next Etymological Dictionary of the Altaic Languages strives to become the Onomasiological Dictionary of the Altaic Languages, I am afraid that comparative Altaic studies have little hope of getting unstuck from the relative rut in which they find themselves today."

 

[lpetrich]

The Global Lexicostatistical Database

400-item basic lexicon wordlist for potentially "Nostratic" languages of Eurasia

Optimized for semantic and lexicostatistical analysis.

Indo-European, Uralic, Altaic, Eskimo-Aleut, Kartvelian, Dravidian

This database represents the current (intermediate) stage of research on a unified and formalized wordlist of 400 basic lexical items, reconstructed for several non-controversial protolanguages of Eurasia. Its primary purpose is to serve as a test base for a full-scale lexicostatistical verification of the "Nostratic" hypothesis (V. M. Illich-Svitych, A. B. Dolgopolsky and others), according to which all or most of these protolanguages go back to a single "Proto-Nostratic" ancestor. It can and should, however, be used (when completed) for additional purposes as well, including etymological research on individual families, general typology of semantic shifts, and areal linguistics.

It's an expanded version of the 100-word and 200-word Swadesh lists. That is done because long-range linguistics requires a large comparison vocabulary for getting good statistics.

Part of that large vocabulary is semantically-related sets of words, like "eye : to see", "black : night", "earth : dust : sand".

The work will be done by giving the list items precise semantics and compiling plausible semantic shifts. There are lots of weird semantic shifts, but they are not very common, and one should not count on them to demonstrate a relationship.
"fig" > "liver of fig-fed goose" > "liver"
"Rome" > "language descended from Latin" > "book-length fiction" > "love affair"
"new" > "book-length fiction"

Doing so is Johannes Becanus Goropius territory; he had etymologies like
"protector from cold" > "oak tree"

Not much different from "fig" > "liver" if one omits the intermediate, "liver of fig-fed goose", I must concede.

 

[lpetrich]

Stability? From "how many different roots in the family correspond to this particular meaning?" and "in what percentage of the languages belonging to this family has this root been retained in the same meaning that it had in the protolanguage?"

Good measures. It would also be interesting to see how well they are correlated.

They use North Caucasian (Lezgian), Altaic (Turkic, Mongolic, and Tungusic), Indo-European, Dravidian, Zhuang-Tai, and Khoisan (Khoe), with data from the Tower of Babel etymological database project.


"The procedure was not automated, and a slight element of personal bias may have been involved in the process, but this will most likely be corrected over the course of currently ongoing and future revisions."

Likewise, the semantic connections between the different entries on the list have been elicited empirically, based on observed polysemies as well as undebatable etymological connections between words with different, but similar meanings. The typical rule of thumb is that the same semantic shift has to be observed independently at least twice, preferably in geographically remote areas, to be considered "trivial"; as a rule, they are observed much more often than twice. A database of semantic shifts elicited from the various language families of Eurasia as their data are being entered into the main database is forthcoming.

Has both English and Russian versions of the meanings, for improved specification.

The database currently has IE, Uralic, Yukaghir, (Altaic: Turkic, Mongolic, Tungusic, Korean, Japanese), Dravidian, Kartvelian, (Eskimo-Aleut: Eskimo, Aleut), (Chukotko-Kamchatkan: Chukchee-Koryak, Itelmen), Nivkh

I looked in it, and it has 1, 2, 3 but no other numerals.

 

[lpetrich]

KAUFMAN’S BASIC CONCEPT LIST ON HISTORICAL PRINCIPLES - 1973

It has 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 100

I looked for databases of phonetic and semantic shifts, and I found these:

PHOIBLE 2.0 - - phonological inventories of many languages

Searchable Index Diachronica - list of known sound shifts

The Database of Semantic Shifts in the languages of the world

 

[lpetrich]

Stochastic_approach_to_worldwide_languag.pdf

Stochastic approach to worldwide language classification: the signals and the noise towards long-range exploration
Beaufils Vincent1, Tomin Johannes2,
1 Independent researcher, Vienna, Austria
2 Independent researcher, Vienna, Austria

The authors used the  Automated Similarity Judgment Program (ASJP) in the work. Since it has only 40 meanings, it has a strong risk of small-number statistics. One should use more meanings.

They decided on 27 consonant classes and also on scoring similar but different consonants by amount of similarity, as a measure of the likelihood of sound changes between them. After discussed how they matched words, they turned to their choice of words to compare.

Throughout our research, it became clearer and clearer that more words do not mean better classifications. As observed in other researches, very few words have a high degree of stability against borrowing, semantic shift and phonological erosion over a longer time frame.

That's only for relatively close languages, because a short list means difficult long-range comparison.

They started with the 40 words in the  Automated Similarity Judgment Program list, and removed some.

They first removed "to come", "liver", "path", "person", and then "fish", "mountain", "stone", leaving 33 words.

 

[lpetrich]

Why did they remove "liver"?

“liver” is certainly not a word suited for medium- and long-range classification, as it implies some kind of stable knowledge of anatomy at an equal level in various parts of the world.

That's a dumb argument, because this internal organ is a universal vertebrate feature, and most animals that we use for food are vertebrates, meaning that those ones have livers.  Liver Then looking in liver - Wiktionary, the free dictionary reveals the word to be fairly stable.

English "liver", like other Germanic words, are from Proto-Germanic librô, from Proto-Indo-European *leyp- "to stick, something sticky, like fat" > Greek lipos "fat (substance)"

Celtic words go back to Proto-Celtic *awV- origin obscure.

Latin iecur, iecinor- or iecor-, Greek hepar, hepat-, Sanskrit yakrt, yakn-, Persian jigar (dZ), Latvian aknas < PIE *yekwr, *yekwn- (r/n declension)

Romance forms come from the fīcātum part of Latin iecur fīcātum, "liver of fig-fed goose", lit. "figged liver".

Balto-Slavic is somehwat complicated.
Russian pechen', Slovak pečeň ~ Polish pieczeń "roast meat"
Polish wątroba < Proto-Slavic *otroba "internal organs"
Czech játra, Slovene jẹ́tra, Serbo-Croatian jȅtra < *etro
Bulgarian cheren drob lit. "black lung", drob < PSl *drobu "internal organs"
Lithuanian kẽpenys lit. "baked goods" < kèpti "to bake"

PSl *etro < PIE *h1enteros (>*enteros) < *h1en "in" + -teros "especially something, more something, ..."
> Greek enteron "iinternal organs" > "intestines" (that word itself is from Latin intestinum "internal organs")

Outside of Indo-European,

Proto-Uralic *mëksa

Proto-Turkic *bagïr
Its Turkish descendant - bağır - means "breast, bosom"
Turkish karaciğer < kara "black" + ciğer "lung, liver" < Persian jigar "liver"

Proto-Kartvelian *q'wiz1-

Proto-Semitic *kabid

Proto-Sino-Tibetan *m-sin "liver, bitter, bite"

North Caucasian:
Proto-Abkhaz-Adyghe *ch'wa
Proto-Nakh *daHVx (x = kh)
Proto-Lezgian *lätl'

Proto-Austronesian *qaCay > Proto-Malayo-Polynesian *qatay

Proto-Tai *tap(C)

Proto-Mon-Khmer *tlëëm

So words for "liver" are often well-preserved.

 

[lpetrich]

Another location of the paper: OSF | Stochastic approach in worldwide language classification: the signals and the noise towards long-range exploration at SocArXiv
Complete with supplementary data.

I'm a bit annoyed at seeing "Type I" and "Type II" for errors when "false positive" and "false negative" are completely good terms for them.

After deciding on which languages, which word list, and how to do matching, the authors then considered how to set up a null hypothesis for matching. They settled on randomly selecting from their full set of of langs.

There is a very curious omission: protoforms. The authors should have compared those instead of individual langs in a family. Doing so saves a lot of work, because the necessary comparison work was already done in finding the protoforms.

So between a short word list and the lack of use of protoforms, I conclude that its results may not be very reliable, though their methods are otherwise reasonable.

What they found support for:

• Altaic: Turkic, Mongolic, Tungusic (Narrow Altaic) - not Korean, Japanese
• Eurasiatic: Indo-European, Uralic, Altaic - not Chukotko-Kamchatkan, Eskimo-Aleut
• Nostratic: Eurasiatic, Kartvelian, Afroasiatic - not Dravidian
• North Caucasian: Northeast, Northwest Caucasian
• Austric: Austro-Tai, (Austroasiatic, Hmong-Mien) with (Hmong-Mien, Sino-Tibetan, Japonic)
• Almosan-Keresiouan: Algonquian, Wakashan, Saliahan, Iroquoian
• Mayan, Totozoquean: (Mixe-Zoque, Totonacan)
• Mataco–Guaicuru: Matacoan, Guaicuruan
• Trans-New-Guinean, West Bomberai, Timor-Alor-Pantar
• Niger-Congo: Atlantic-Congo, Mande

From their discussion,

The system cannot differentiate the nature of the signals behind the inferred long-range connections: relatedness or early language contact. However, as we can see throughout the research, the word list in use is highly resistant to borrowing.

With a longer word list, it can be easier. Are cognates concentrated in the more stable part of the list or the less stable part?

Browsing the worldwide phylogeny in the online Supporting Information S5 File we see that well documented languages of Eurasia and Northern Africa (at the bottom of the file) lead to cleaner results. More noise interferes in the classifications of less documented languages like those of New Guinea, Australia or South America. Better quality of the data in use leads to a better differentiation between signals and noise.

Like using a longer word list and using protoforms as far as is reasonably feasible.

This is something like Automated Dating of the World’s Language Families Based on Lexical Similarity | Current Anthropology: Vol 52, No 6 with PDF file Automated Dating of the World’s Language Families Based on Lexical Similarity - shh768.pdf using the ASJP with its 40-word list.

But that latter one had a big collection of calibration points, and I tried fitting their ASJP similarity values to an exponential curve in time. I found that the decline rate slowed for times greater than about 2500 BP. Could this be an effect of differences in stability? The less stable part decaying before the more stable part.

 

[lpetrich]

Probabilistic Evaluation of Comparative Reconstructions by Andrei Munteanu

"This dissertation introduces Wordlist Distortion Theory, a framework for the probabilistic evaluation of comparative reconstructions in historical linguistics."

Essentially trying to find the best-fitting sound correspondences for going from an earlier language to a later one. This is done by automated trial and error, by trying out some random change to the correspondences and then accepting one that reduces the error. Accepting only for reducing the error is "hill climbing" and it can get stuck in local optima. So what they did instead was "simulated annealing", accepting increased error with a probability depending on how much increase, thus letting the solution wander near local optima in the hope of finding one with long slopes around it. How much increased error to accept is gradually reduced over each run, to try out smaller and smaller regions of parameter space.

A motivation for this research was to enable testing of long-range hypotheses like Altaic, Eurasiatic, Nostratic, Borean, and Proto-World. An illustration of this problem is the case of dueling reconstructions of Afroasiatic: Ehret and Orel-Stolbova. On calculating the reliability of the comparative method at long and medium distances: Afroasiatic comparative lexica as a test case - only 6% match between the two.

AM decided to work on Austronesian and Ongan, of the Andaman Islands, testing the Austronesian-Ongan hypothesis.

AM wanted to test his method with a control language, so he invented a very fake one, by taking each protoform's entry number and using which number of word in JRR Tolkien's "Lord of the Rings", counting from the beginning of that fantasy novel. He used the English spelling of each word as its "phonetic transcription".

He found that this control language gave randomness probabilities of 1 or close to it: 0.181, well above his 10^(-5) acceptance limit.

For the Philippine langs that he looked at, he found Proto-Austronesian, Proto-Malayo-Polynesian, Proto-West-Malayo-Polynesian, and Proto-Philippine, all their ancestors, and not so good but still above significance for Proto-Oceanic.

For the Oceanic langs, PAN was not very good but significant, PMP was more significant, and POC was very significant. However, PWMP and PPh were not significant.

For the western ones, PAN, PMP, and PWMP were all significant, POC was somewhat significant, and PPh not significant.

For the central ones, PAN was significant, PMP was more siginficant, POC was somewhat significant, and PPh and PWMP not significant.

For the Formosan ones, PAN was sigificant, PMP less significant, POC a little significant, and PWMP and PPh not significant.

All these results are what one would expect from their ancestry, and the significant results for some non-ancestral protolanguages was from them sharing some recognizable word forms.

Using the proposed Proto-Ongan-AN word list gave much worse results, with borderline significance. Most of that list's word forms were fairly close to the PAN ones, so the poor results were likely due to not having much to work from.

 

[lpetrich]

What my luck. Allan Bomhard has also uploaded to archive.org Joseph Greenberg's book on Eurasiatic.

• Greenberg - Indo-European and Its Closest Relatives - The Eurasiatic Language Family, Vol. 1 - Grammar (2000) : Allan R. Bomhard : Free Download, Borrow, and Streaming : Internet Archive
• Greenberg - Indo-European and Its Closest Relatives - The Eurasiatic Language Family, Vol. 2 - Lexicon (2002) : Allan R. Bomhard : Free Download, Borrow, and Streaming : Internet Archive

It covers Indo-European, Uralic, Narrow Altaic, Korean, Japanese, Chukotko-Kamchatkan, Eskimo-Aleut, and Nivkh (Gilyak).

I now have all three reconstructions of Nostratic: Muscovite, Bomhardian, and Greenbergian.

JG proposes that some of the Indo-European laryngeals are descended from velar stops: /k/ - I recall Fredrik Kortlandt proposing something similar in one of his papers, though I've been unable to track it down: in verbs, Uralic -k- ~ IE -H-.

That is something like PIE *ghans- "goose" ~ PIE *Hanats "duck" (water bird: Latin anas) that I mentioned earlier.

But I must note that JC doesn't try to find sound correspondences.

 

[lpetrich]

I've found Indo-European secondary products terminology and the dating of Proto-Indo-Anatolian | Scholarly Publications with PDF file

In this paper, an attempt is made to date the ancestor of all IndoEuropean languages on the basis of the Indo-European terminology for the exploitation of animals for products that do not require killing the animal (the "secondary products" revolution). It is argued that this terminology is compatible with a society that made use of animal traction, but that did not necessarily practice dairying or use wool for textile production. This is compatible with a date at the beginning of the fourth millennium BCE and with the hypothesis that this ancestor language was spoken by people of the so-called Khvalynsk culture on the Volga River.

Eggs are a secondary product, but are not listed here, because the PIE speakers did not have domestic birds.

 List of domesticated animals

• Chicken: 6000 BCE - Southeast Asia, in Middle East around 2000 BCE
• Duck: 4000 BCE - China
• Pigeon: 3000 BCE - Mediterranean basin
• Goose: 3000 BCE - Egypt

 

[lpetrich]

Nomenclature for early subfamilies of IE is in a somewhat confused state, with overall IE itself sometimes being renamed.

• Overall: Indo-Anatolian, Indo-Hittite, Basal IE -- Early Proto-Indo-European
• After Anatolian splits off: Indo-Tocharian, Nuclear IE -- Late PIE
• After Tocharian splits off: Core IE, Inner IE -- Late PIE

This core has split

• Albanian (difficult to place)
• Greek, Phrygian, Armenian
• Italic, Celtic, Germanic
• Balto-Slavic, Indo-Iranian

Rapid radiation of the inner Indo-European languages: an advanced approach to Indo-European lexicostatistics

Indo-European cladistic nomenclature - paywalled