Paleo-European languages

[Ares Land]

I thought that for a test using the 'scientific method', a prediction had to be right. I don't think 'X or not X' is falsifiable, except possibly when X is meaningless.

In terms of pure logic, yes, but there's really an unstated X, most of the time because there's almost always other factors at play than the one you're testing for.
It's not that linguistics is particularly soft or fuzzy: the same problem arises in physics.
(Orbital decay, for instance, doesn't falsify Newton's laws of motion: there's just another force arising from residual atmospheric friction)

But is this earlier settlement linguistically relevant?

Yep. Greenberg relied a lot on extra linguistic factors such as waves of migrations to the Americas, which he -- reasonably! -- assumed would have been limited in number.
Basically, his reasoning was, if you have three waves of migration, you get three language families.
Pre-Clovis human presence, though, suggests settlement patterns were more complex than usually assumed.

[Moose-tache]

When you meet someone who believes that Mandarin is derived form Basque, you may entreat them to explain what would disprove their theory, in order to see if it is falsifiable or not. This is a mistake. The Nylandian can simply say "Should anyone dig up a tortoise shell in Gansu that says 'we are definitely not Basque,' I will recant my theory fully." Now they have established that their theory is technically falsifiable and you are left no less frustrated. A smarter angle is to insist that the Nylandian ask what would be the consequences if their theory were wrong. For example, if Mandarin is not derived form Basque, what data remains unexplainable? What details jump out as obvious flaws in the mainstream theory? Probably nothing. Another example: if the word Shrdn is not cognate with Sardinia, the consequence is this: an ethnonym given to an unknown group by another group is coincidentally similar to a toponym halfway across the Mediterranean first attested several hundred years later. This is what we're dealing with here. What are the consequences of "Gorm the Celtic word for trash heap" and "Gram the Armenian word for goat testicles" being coincidentally similar? There are none. That's the problem here. It's finding faces on the surface of Mars, or Ley lines across the English countryside. The answer to "OK, but what if not that?" can only be a shrug.

As for "Amerindian," I think if you're trying to solve a linguistic problem, you need to apply linguistic methodology. Even if there was solid proof of one migration to the Americas, you would need to demonstrate a language family through regular sound correspondence. There is not even a hint of any such thing in Amerindian, besides some conspiratorial ravings over agreement affixes (basically the local chapter of the Mitians).

[Richard W]

Greenberg relied a lot on extra linguistic factors such as waves of migrations to the Americas, which he -- reasonably! -- assumed would have been limited in number.
Basically, his reasoning was, if you have three waves of migration, you get three language families.
Pre-Clovis human presence, though, suggests settlement patterns were more complex than usually assumed.

But if the pre-Clovis languages are all replaced, then the earlier settlement may not be linguistically relevant. For example, it now seems quite possible that the Palaeolithic and Mesolithic languages of Western Europe are irrelevant to what is attested for Western Europe. If the descendants of pre-Clovis were merely swamped, then we may get a few languages that don't fit, but that becomes an issue for mopping up, which is not where we're at.

I'm confused by the accounts of Amerindian genetics, but I get the impression that older populations were replaced or absorbed by newcomers.

[Richard W]

It's not that linguistics is particularly soft or fuzzy: the same problem arises in physics.
(Orbital decay, for instance, doesn't falsify Newton's laws of motion: there's just another force arising from residual atmospheric friction)

But orbital decay does falsify a purported list of all relevant forces that only lists Newtonian gravity. Practical long term prediction has to allow for it; I'm not sure how relevant it is for short term tracking when maintaining a catalogue of objects in low Earth orbit.

In experimentation, one is supposed to take error bars into account.

[Talskubilos]

The problem, in this case, is that having only three correspondences is nearly as useless as having only one: as zompist points out, this is easily within the bounds of chance resemblances. In fact, you can find more correspondences between Mandarin and Quechua than you have between Gaulish and Baltic!

Really, here is what we want to see from you: a list of sound correspondences, backed up with significant data (at least 20 words or so, if not more), preferably without making up unattested words. If you can do that, I suspect that you’ll start seeing a lot fewer complaints.

The thing is we're dealing with substrate loanwords, and unfortunately the material is scanty, so quality primes over quantity. On the other hand, these Gaulish words are reconstructed from its attested descendants.

[Ares Land]

Well, of course, but that's a pretty big 'if'. I can't say I've read anything really conclusive on the matter. IIRC there are claims of significant pre-Clovis ancestry, especially in South America.

Back to Europe...

For example, it now seems quite possible that the Palaeolithic and Mesolithic languages of Western Europe are irrelevant to what is attested for Western Europe.

That seems to be the majority view... and I don't understand it at all.
To me it's just a bad case of searching for our keys under the streetlight.

We've wiped out pre-Columbian America to an extent that would have been entirely out of reach of the Indo-Europeans. And still the influence of American languages is very noticeable in toponymy, and there are a solid amount of borrowings.

We can debate all day on whether Tiberis is Etruscan or Italic (and we did!) but at the end of the day I'm afraid what we're doing is debating whether 'Mississipi' is English or Spanish.

On other issues discussed here, positing lost IE languages is good fun, but the best option is probably to leave those poor Pelasgians alone.
There's nothing that surprising with the idea that the Germans and the Greeks borrowed a lot from whomever was there before there, and really no particular reason to claim those earlier people were Indo-European or even related to each other.

I'm also not convinced at all by the idea that a shared material culture necessarily entails linguistic uniformity. It's like claiming the French, Brits and Americans all speak the same language because we use the same Coke bottles.

Mesoamerica at the time of the Spanish conquest is another good example: the area shared a common culture with several unrelated languages. If we just had the ruins of Chichén Itzá and Teotihuacan with no other evidence, I'm not sure we'd have guessed that a different language from a different family was spoken in each place.

[Ares Land]

It's not that linguistics is particularly soft or fuzzy: the same problem arises in physics.
(Orbital decay, for instance, doesn't falsify Newton's laws of motion: there's just another force arising from residual atmospheric friction)

But orbital decay does falsify a purported list of all relevant forces that only lists Newtonian gravity. Practical long term prediction has to allow for it; I'm not sure how relevant it is for short term tracking when maintaining a catalogue of objects in low Earth orbit.

In experimentation, one is supposed to take error bars into account.

Again, I'm not sure what your point is.

It's exactly the same in linguistics: we really take into account a list of factors that include correspondances, borrowings, irregular change... You mentioned the Swadesh list earlier: the reason it's used so much is that we know, statistically, that genetic relationships are more likely and borrowing less likely for those words.

[Ares Land]

The thing is we're dealing with substrate loanwords, and unfortunately the material is scanty, so quality primes over quantity. On the other hand, these Gaulish words are reconstructed from its attested descendants.

Statistically, the odds that there will be chance resemblances between Baltic and Gaulish are very high, so there can be no conclusive proof that those are loanwords.

You've stated the existence of those loanwords as established fact, which is untenable. But you could restate your point as an hypothesis, or really as an invitation to keep looking for such similarities. And hopefully, you'll find enough to get your point conclusively proven.
This is exactly how, say, classifying Hittite as Indo-European started out. Or Déné-Yenisean (and it's a long process. Déné-Yenisean still isn't conclusively proven, and probably won't be for decades)
I mean, nobody's opposed in principle to the idea of an underlying 'satem' substrate, we only object to stating its existence as established fact.

Option B: since the evidence is unconclusive, we just tag these words 'etym. unclear' and leave at that. And really that option is perfectly acceptable and it really should be used more often, including in etymological dictionaries.

(For conlang lexicon building, I check Latin etymologies often and sometimes you just get stuff like bigusdickus < Proto-Italic *blahblah < PIE *ba, related to Tocharian *bwaaaghhh 'to step in a lot of horse manure' and Minoan A *gobble-gobble, 'fried ox testicle' and in all fairness to you, nobody seems terribly bothered by that.)

[Talskubilos]

There are (at least!) three sets of regular ("consistent") sound correspondences between English and French, in three different lexical strata:
- the vocabulary inherited from PIE in both English and French
- the vocabulary that was loaned from Latin into English and inherited from Latin in French
- the vocabulary that was loaned from French into English

I think "PIE" is a far stretch. For example, although French poisson and English fish are descendants from a common source, there's no guaranty it was PIE. On the other hand, we should also add the West Germanic (Frankish) superstratum in French.

[Ryusenshi]

On the other hand, we should also add the West Germanic (Frankish) superstratum in French.

Yeah. For instance, there is a regular correspondence between French words in g- and Germanic words in w-: gage/wage, garde/ward, guerre/war, gaufre/waffle, Guillaume/William, Galles/Wales

[Ares Land]

There are (at least!) three sets of regular ("consistent") sound correspondences between English and French, in three different lexical strata:
- the vocabulary inherited from PIE in both English and French
- the vocabulary that was loaned from Latin into English and inherited from Latin in French
- the vocabulary that was loaned from French into English

I think "PIE" is a far stretch. For example, although French poisson and English fish are descendants from a common source, there's no guaranty it was PIE. On the other hand, we should also add the West Germanic (Frankish) superstratum in French.

Nortaneous specifically mentions these three strata to make a point. (You didn't quote his question: what does that imply about the classical model of Germanic? which IMHO is the important bit)

Indeed we could add quite a few strata to that list: poisson and fish can't be traced back to PIE. (But pied / foot can!). Then there are the Frankish loanwords, the Francien loanwords as opposed to the Norman loanwords, then the strata within French (Francien also borrowed from Norman, and that's just the tip of the iceberg. Old French is an ungodly mess.), then the Latin reborrowings in French, then dialectal borrowings within English... And the list still isn't exhaustive.

The thing is, none of these really affect much what we know of proto-Germanic.

[bradrn]

The thing is we're dealing with substrate loanwords, and unfortunately the material is scanty, so quality primes over quantity.

Hmm, I suspect we’re getting to the root of the disagreement here. In this particular case, I don’t agree that quality is better than quantity: as I said, chance resemblances are very common, so having just a few correspondences doesn’t necessarily prove anything — regardless of whether we’re looking at a poorly-attested substrate or not. (And what do you mean by ‘quality’ anyway?)

There are (at least!) three sets of regular ("consistent") sound correspondences between English and French, in three different lexical strata:
- the vocabulary inherited from PIE in both English and French
- the vocabulary that was loaned from Latin into English and inherited from Latin in French
- the vocabulary that was loaned from French into English

I think "PIE" is a far stretch. For example, although French poisson and English fish are descendants from a common source, there's no guaranty it was PIE. On the other hand, we should also add the West Germanic (Frankish) superstratum in French.

I know we’ve already talked about this, but I’m still curious to know: what evidence, exactly, made you reject the usual PIE model? And, given that you reject the existence of a single ‘PIE’ language, how do you explain the many sound correspondences between words in the various IE branches?

[Talskubilos]

The thing is, none of these really affect much what we know of proto-Germanic.

So what?

Statistically, the odds that there will be chance resemblances between Baltic and Gaulish are very high

Really? Surely you'll be astonished by the list of Gaulish-Latin chance resemblances made by Delamarre (e.g. Gaulish markos 'horse' ~ Latin Marcus).

You've stated the existence of those loanwords as established fact, which is untenable.

Not more "established fact" that the existence of PIE, which is a theory.

But you could restate your point as an hypothesis, or really as an invitation to keep looking for such similarities. And hopefully, you'll find enough to get your point conclusively proven.

As I said before, the hypothesis of a Baltoid substrate language wasn't originally mine but Coromines', and followed his steps in my own research.

This is exactly how, say, classifying Hittite as Indo-European started out. Or Déné-Yenisean (and it's a long process. Déné-Yenisean still isn't conclusively proven, and probably won't be for decades)

I don't think Swadesh list and genealogical trees will be useful for that. Proposed macro-families are plagued with the same defects than the current PIE theory, although with less data available.

I mean, nobody's opposed in principle to the idea of an underlying 'satem' substrate, we only object to stating its existence as established fact.

Actually, this other "satem" substrate shouldn't be confused with Baltoid and could be a lost branch of Indo-Iranian, as suggested by Richard.

[Ares Land]

OK, we're officially into 'it's just a theory!!!" territory... I'm outta here. Good luck everyone.

[bradrn]

You've stated the existence of those loanwords as established fact, which is untenable.

Not more "established fact" that the existence of PIE, which is a theory.

The problem here is that there are different levels of theory-hood. On the one hand, you have things like the General Theory of Relativity, every single prediction of which has been borne out by experimental reality. On the other hand, you have pure crackpottery like our friend Edo Nyland (or, if you prefer to stick with physics, stuff like the Flat Earthers). It should be patently obvious that, even though they are both theories, one is probably closer to reality than the other.

My point here is that, on the scale of theory-hood from crackpot to practically-the-truth, PIE is far closer to the latter than the former. Although we can’t directly confirm it unless we get a time machine, since its proposal PIE has been poked and prodded at by numerous experienced linguists — and I can’t help but think that they surely would have noticed by now if there was something very wrong with that theory, no? I mean, it is within the realms of possibility that there’s something they’ve all missed — and by all means, please do tell us if you think you’ve found such a gaping hole! — but I have a very strong hunch that there isn’t. On the other hand, while I certainly wouldn’t call your theories crackpottery, they do have far less support than the consensus, and I find it suspicious that you are unable to give more examples than a few instances of what could well be chance resemblances.

[WeepingElf]

Amerind did seem to make sense as long as the "Clovis first" model was not disproven by archaeological finds that showed that humans had been in the Americas earlier, but this only shows how hazardous it is to build a language family hypothesis on extralinguistic evidence.

But is this earlier settlement linguistically relevant?

At least we don't know that it isn't. What reasons are there to assume that the language(s) of the Clovis people obliterated those of the earlier immigrants completely? Such a scenario may be conceivable (Europe seems to have gone through two waves of language replacement, leaving only few remains of the Neolithic languages - Basque and the three Caucasian families - and probably none of the Paleolithic ones), but so far we AFAIK have no proof. Are there any paleogenetic studies which reveal that the present-day indigenous Americans all descend from Clovis or later immigration waves and none from pre-Clovis people?

[Ares Land]

Are there any paleogenetic studies which reveal that the present-day indigenous Americans all descend from Clovis or later immigration waves and none from pre-Clovis people?

As I recall, Anzick-1 (the one 'Clovis person' for which DNA analysis has been done) is related to all Native Americans. But I don't think any paleogenetic studies have been undertaken on pre-Clovis DNA.
So we just don't know.

There are further unknowns and unknown unknowns.

Greenberg posited 1 family where Americanists posited 300. Greenberg quipped about traffic controllers on the Bering strait, but I believe we have quite a wrong view of what Beringia was like.
We think of it as a 'land bridge', but it wasn't so much a bridge as a fairly large area of permanent settlement. There's really nothing precluding it being similar to Australia -- 300 languages and 20 families...
Further unknown unknowns: a fairly solid attack on the traditional linguistic tree model is that it sometimes inexplicably breaks down with hunting-gathering populations... I'm thinking of Australian languages again. You can't really draw anything resembling a tree, and languages share similarities that can't be satisfactorily explained with traditional models.

So it's really possible that an entirely different model applies and that Greenberg and his opponents were both right. Or both wrong.

[Richard W]

Actually, this other "satem" substrate shouldn't be confused with Baltoid and could be a lost branch of Indo-Iranian, as suggested by Richard.

----
I've found the paper that concluded that there were Indo-Iranian loans into Celtic. It's

Alix Boc, Anna Maria Di Sciullo and Vladimir Makarenkov. Classification of the Indo-European languages
using a phylogenetic network approach. DOI: 10.1007/978-3-642-10745-0_71, January 2010

It's available from several sites - Makarenkov allegedly (I can't verify it was actually him) uploaded it to https://www.researchgate.net/publicatio ... 6/download .

There is a presentation that gives some of the algorithms at https://slideplayer.fr/slide/5230520/ - pseudo-code starts at 9:26.
----
As I read the paper, it uses the Dyen, Kruskal and Black IE Swadesh list database. The method builds word phylogenies using the Levenshtein distance metric and the neighbour joining method, which each give me the heebie-jeebies. I'm not at all sure I can repeat the calculations, and I fear the 'loans' may turn out to be an artefact of the analysis. The problems I see are:

• The Levenshtein distance metric can't handle sound shifts. If one didn't transliterate, words in different scripts would have the maximal normalised difference!
• The neighbour joining method works particularly badly when generating phylogenetic trees. Taking more members of a group makes the group more distant from everything else. One attempt to use it failed to recover Indo-Iranian, and the Iranian language Wakhi wound up as the outgroup for IE.

I fear we need to run the analysis to see what it gives, and carefully check whether what might actually be retentions really look like loans.

[WeepingElf]

So it's really possible that an entirely different model applies and that Greenberg and his opponents were both right. Or both wrong.

Yes. The traditional comparative method has been developed in order to explain what had happened in the Indo-European family, which appears to have spread rapidly over a large area. IE probably emerged when an expansive culture with a military edge (horses!) conquered large swaths of Europe, Central Asia and South Asia. An alternative, but IMHO much less likely, explanation is that the IE family was spread by Neolithic farmers expanding into hunter-gatherer areas. Neither of the two is the norm in human population dynamics; nor are they the same thing as hunter-gatherers expanding into hitherto uninhabited areas. What happens on a continent that is inhabited by various groups side by side with none ever becoming dominant is simply not known well.

[keenir]

You've stated the existence of those loanwords as established fact, which is untenable.

Not more "established fact" that the existence of PIE, which is a theory.

So is gravity!