The oddities of Basque

[Travis B.]

is this one of those times where "primitive peoples don't need as many words" as some critics say?

It should be remembered that oftentimes "primitive people" use far more words about things in the natural environment than are typically part of the active vocabularies of people from developed societies. When you're a hunter-gatherer, it matters to know that plant A is good to eat whereas plant B is deadly poisonous.

[Richard W]

It might exist, but its lexicon would only be only a subset of the +2000 items commonly reconstructed for PIE.

Who reconstruct 2000+ items? That's the count for Pokorny, and many of his entries are rejected as lacking adequate evidence.

[keenir]

It might exist, but its lexicon would only be only a subset of the +2000 items commonly reconstructed for PIE.

Who reconstruct 2000+ items? That's the count for Pokorny, and many of his entries are rejected as lacking adequate evidence.

oh. in that case, it actually makes sense that Talsukubilos would only accept a subset of them.

[Nortaneous]

The likelihood of chance resemblances certainly does goes up as you sample more words and languages; but it also goes down as you tighten restrictions on phonological and semantic similarities. Currently the range of phonogical correspondences accepted by mainstream historical linguists is very narrow: only words corresponding to each other by regular changes are accepted.

This isn't entirely true. Regular changes are preferred, but irregular changes are allowed.

The range of semantic correspondences is also restricted — a Greenbergian comparison of ‘udder’ to ‘suck’ to ‘throat’ (say) would never be acceptable.

Sometimes too restricted - on multiple occasions (which I've since forgotten) I've seen linguists write off as too farfetched semantic developments that happened in English.

(And by the way, I meant that there were ‘hundreds of cognates’ for each sound change for each language. Probably thousands of cognate pairs in total.)

Hundreds of cognates for each sound change? Maybe there are a hundred for, say, the Grimm's Law development of PIE *t. (And maybe this isn't entirely mainstream thought, but IMO "cognates for sound changes" is a type error - there are cognates for correspondences which are later arranged into sound changes. Verner's Law was a set of irregular but accepted cognates until it was formulated as a regular law, wasn't it?)

I mean, let’s go through this step-by-step. You accept that the IE language family is real; that is, that the various IE languages are all related. You accept that the reconstructed ‘PIE’ protoforms are all real; that is, that each set of cognate words from the various IE languages ultimately descent from a single word used thousands of years ago. But you don’t accept that these protoforms all came from the same language‽ How would it even happen, historically, that a bunch of words all ended up in exactly the same set of modern-day languages, but they didn’t originate in a single protolanguage?

Northwest Group words like *h2ebol- and *bhabh- are reconstructed for PIE even though the mainstream position is that they were not inherited from the common ancestor of IE languages.

Indo-Europeanists err on the side of back-projecting words to PIE - the sound correspondences are well enough known that they can do that, and it might be useful for etymological research in some other IE language, so you may as well.

According to Piotr Gąsiorowski, the numeral *kʷetwōr- '4' would be a derivative of a fossilized verb lexeme *kʷet- 'to group into pairs', which IMHO would be also the origin (through reduplication) of Lithuanian kek(e)tà 'detachment, flock' and Uralic *kakta ~ *kæktæ '2'. The latter would be cognate to IE *Hoḱte-h₃(u) '8', a fossilized dual whose original value appears to have been doubled.

Is that reduplication pattern attested elsewhere in either family?

[bradrn]

The likelihood of chance resemblances certainly does goes up as you sample more words and languages; but it also goes down as you tighten restrictions on phonological and semantic similarities. Currently the range of phonogical correspondences accepted by mainstream historical linguists is very narrow: only words corresponding to each other by regular changes are accepted.

This isn't entirely true. Regular changes are preferred, but irregular changes are allowed.

The range of semantic correspondences is also restricted — a Greenbergian comparison of ‘udder’ to ‘suck’ to ‘throat’ (say) would never be acceptable.

Sometimes too restricted - on multiple occasions (which I've since forgotten) I've seen linguists write off as too farfetched semantic developments that happened in English.

I can accept this, though I assume that linguists do try their best to reduce the number of irregular changes. (Something like ‘p → z / only in these five words’ would make me extremely suspicious, for instance.) I was only trying to make the point that stringency in phonological developments, and not just blindly accepting every apparent sound correspondence, helps to reduce the number of chance resemblances.

(And by the way, I meant that there were ‘hundreds of cognates’ for each sound change for each language. Probably thousands of cognate pairs in total.)

Hundreds of cognates for each sound change? Maybe there are a hundred for, say, the Grimm's Law development of PIE *t.

Well, I was specifically thinking of Grimm’s Law when I wrote that. I can accept that other sound changes will be supported by fewer cognates.

(And maybe this isn't entirely mainstream thought, but IMO "cognates for sound changes" is a type error - there are cognates for correspondences which are later arranged into sound changes. Verner's Law was a set of irregular but accepted cognates until it was formulated as a regular law, wasn't it?)

I agree with this.

I mean, let’s go through this step-by-step. You accept that the IE language family is real; that is, that the various IE languages are all related. You accept that the reconstructed ‘PIE’ protoforms are all real; that is, that each set of cognate words from the various IE languages ultimately descent from a single word used thousands of years ago. But you don’t accept that these protoforms all came from the same language‽ How would it even happen, historically, that a bunch of words all ended up in exactly the same set of modern-day languages, but they didn’t originate in a single protolanguage?

Northwest Group words like *h2ebol- and *bhabh- are reconstructed for PIE even though the mainstream position is that they were not inherited from the common ancestor of IE languages.

Yes, but the whole point about those words is that they are only supported by evidence from a limited subset of branches. My point was that it’s unreasonable to take words which are attested from the vast majority of IE subgroups, reconstruct a protoform, but then say that protoform was not from PIE even though that’s the only way it could have been inherited into so many languages.

Indo-Europeanists err on the side of back-projecting words to PIE - the sound correspondences are well enough known that they can do that, and it might be useful for etymological research in some other IE language, so you may as well.

Then they should at least bother to say that they’re doing so! I’m uncomfortable about the idea of back-projecting without any further evidence and without any notice that there is no evidence.

[Pabappa]

i think thats just how it's done ... since, by definition, all PIE roots are unattested, why not project every known proto-Germanic root back into PIE since we have a pretty strong impression of what they would have been in PIE if they were in fact inherited?

for whatever reason, though, it seems that, at least on Wiktionary, Greek is dominated by doubters who assign word after word to an unknown pre-Greek substrate, even words that have a perfectly sound Greek-like consonant pattern. Perhaps this is just Wiktionary, or perhaps it's the Greek scholar community in general, or perhaps it's known that Greek really did take all those words from the substrate.

[Talskubilos]

It might exist, but its lexicon would only be only a subset of the +2000 items commonly reconstructed for PIE.

Who reconstruct 2000+ items? That's the count for Pokorny, and many of his entries are rejected as lacking adequate evidence.

Mallory & Adams is a more recent source.

[Raholeun]

is this one of those times where "primitive peoples don't need as many words" as some critics say?

It should be remembered that oftentimes "primitive people" use far more words about things in the natural environment than are typically part of the active vocabularies of people from developed societies. When you're a hunter-gatherer, it matters to know that plant A is good to eat whereas plant B is deadly poisonous.

"In Samoa, Cox discovered that Polynesian herbal doctors had an extensive nomenclature for endemic diseases and a separate one for those introduced by Europeans. Their sophistication is not unique. The taxonomies of endangered languages often distinguish hundreds more types of flora and fauna than are known to Western science. The Haunóo, a tribe of swidden farmers on Mindoro, an island in the Philippines, have forty expressions for types of soil. In Southeast Asia, forest-dwelling healers have identified the medicinal properties of some sixty-five hundred species."
- by Thurman, A Loss for Words.

[WeepingElf]

It might exist, but its lexicon would only be only a subset of the +2000 items commonly reconstructed for PIE.

Who reconstruct 2000+ items? That's the count for Pokorny, and many of his entries are rejected as lacking adequate evidence.

Yes, Pokorny heavily over-reconstructs: there are many items with limited distributions which are more likely loanwords from substratum languages into intermediate nodes or loanwords from one PIE dialect into other nearby dialects; and there are so many homonyms and synonyms that they can hardly all have existed in the same language at the same time. So it has to be used with care (and of course, it uses an obsolete phonology, but rewriting the items according to the currently accepted model is in most cases not difficult and in many cases trivial).

But well, the dictionary should best not be conceived of as the vocabulary of the proto-language, but as a collection of data about cognates within the family, including areally skewed ones. As such, Pokorny's tendency to err to the side of inclusion is not a bug but a feature - it gives access to etymologies which a dictionary that only includes what can be safely ascribed to "PIE proper" would miss because they are too limited in scope. So the over-inclusiveness increases the usefulness of Pokorny's dictionary rather than compromising it, as long as you are aware of this.

This doesn't mean, though, that PIE did not exist as a language in space and time, which of course was divided into dialects and changed over time (and of course, the dissolution of PIE was a gradual process - the speech community repidly expanded, and the dialects diverged more and more until they became different, mutually unintelligible languages - the same process we can observe with written history in Romance), only that not all items in Pokorny's dictionary can safely be ascribed to it.

[keenir]

is this one of those times where "primitive peoples don't need as many words" as some critics say?

I was referring to the stereotype, not to the truth of the matter

[hwhatting]

Quoting from memory, these would include some numerals, namely '2' and '7', kinship terms such as 'daughter' or 'sister' (matrilineal system?) and cultural innovations such as 'horse' and 'wheel'.

Well, we did discuss "horse" already and didn't agree. "7" looks like a good candidate for a wanderwort, what with it looking similar to words in Semitic and Caucasian, and keeping in mind that higher numbers are usually easier loaned than lower numbers. What are the sources for "two" and "wheel", in your opinion (keeping in mind that for "wheel", there is also a convincing internal etymology)?

[bradrn]

Quoting from memory, these would include some numerals, namely '2' and '7', kinship terms such as 'daughter' or 'sister' (matrilineal system?) and cultural innovations such as 'horse' and 'wheel'.

Well, we did discuss "horse" already and didn't agree. "7" looks like a good candidate for a wanderwort, what with it looking similar to words in Semitic and Caucasian, and keeping in mind that higher numbers are usually easier loaned than lower numbers. What are the sources for "two" and "wheel", in your opinion (keeping in mind that for "wheel", there is also a convincing internal etymology)?

He already said:

You must be joking. Taking just one of your examples, how are twa, duo, dúō, dou, dvì, dvà~dvé, dy, wu (from resp. OE, Latin, Greek, Old Welsh, Lithuanian, Sanskrit, Albanian, Tocharian A) not obviously cognate‽

Of course, they're cognate, but the numeral '2' was ultimately borrowed from NW Caucasian, and '5' from NE Caucasian 'fist'.

[hwhatting]

Quoting from memory, these would include some numerals, namely '2' and '7', kinship terms such as 'daughter' or 'sister' (matrilineal system?) and cultural innovations such as 'horse' and 'wheel'.

Well, we did discuss "horse" already and didn't agree. "7" looks like a good candidate for a wanderwort, what with it looking similar to words in Semitic and Caucasian, and keeping in mind that higher numbers are usually easier loaned than lower numbers. What are the sources for "two" and "wheel", in your opinion (keeping in mind that for "wheel", there is also a convincing internal etymology)?

He already said:

You must be joking. Taking just one of your examples, how are twa, duo, dúō, dou, dvì, dvà~dvé, dy, wu (from resp. OE, Latin, Greek, Old Welsh, Lithuanian, Sanskrit, Albanian, Tocharian A) not obviously cognate‽

Of course, they're cognate, but the numeral '2' was ultimately borrowed from NW Caucasian, and '5' from NE Caucasian 'fist'.

Well, he didn't quote any forms there, that's what I'm interested in.

[Talskubilos]

Well, he didn't quote any forms there, that's what I'm interested in.

PNWC (Abkhaz-Adyge) *tˀqˀ^wə '2' -> PIE *dwe-h₃(u) '2' (*-h₃(u) is a dual marker suffix).
PNEC (Nakh-Daghestanian) *fimkˀwV 'fist' -> PIE *penk^we '5'.

[Skookum]

Well, he didn't quote any forms there, that's what I'm interested in.

PNWC (Abkhaz-Adyge) *tˀqˀ^wə '2' -> PIE *dwe-h₃(u) '2' (*-h₃(u) is a dual marker suffix).
PNEC (Nakh-Daghestanian) *fimkˀwV 'fist' -> PIE *penk^we '5'.

The NEC connection is interesting, but what is the argument that PIE borrowed it from NEC and not the other way around, since I think it's been suggested that the PIE word for "five" is related to/derived from the root for "fist"?

On the other hand, the NWC connection for "two" is pretty unconvincing, since I doubt *qˀʷ would be borrowed as *w.

[Rounin Ryuuji]

Well, he didn't quote any forms there, that's what I'm interested in.

PNWC (Abkhaz-Adyge) *tˀqˀ^wə '2' -> PIE *dwe-h₃(u) '2' (*-h₃(u) is a dual marker suffix).
PNEC (Nakh-Daghestanian) *fimkˀwV 'fist' -> PIE *penk^we '5'.

Maybe I'm splitting hairs, but from those forms, I would've expected the Proto-Indo-European ones to be something like *tkwe- (I believe we see dual initial clusters in PIE, at least) and pimku~pimkwe; where, exactly, do we lose the */q/, and get the */i/ > */e/ (even if it patterned as an allophone of */j/, I'm pretty sure Proto-Indo-European had *[i])? The first one looks like a coincidence, and the second like the borrowing (if it is one) happened the other way around.

that the PIE word for "five" is related to/derived from the root for "fist"?

I believe it's conventionally (?) connected with English finger, fist, Latin pugnus; I incidentally find it unusual that such a basic word would be borrowed.

[hwhatting]

Well, he didn't quote any forms there, that's what I'm interested in.

PNWC (Abkhaz-Adyge) *tˀqˀ^wə '2' -> PIE *dwe-h₃(u) '2' (*-h₃(u) is a dual marker suffix).
PNEC (Nakh-Daghestanian) *fimkˀwV 'fist' -> PIE *penk^we '5'.

Maybe I'm splitting hairs, but from those forms, I would've expected the Proto-Indo-European ones to be something like *tkwe- (I believe we see dual initial clusters in PIE, at least) and pimku~pimkwe; where, exactly, do we lose the */q/, and get the */i/ > */e/ (even if it patterned as an allophone of */j/, I'm pretty sure Proto-Indo-European had *[i])? The first one looks like a coincidence, and the second like the borrowing (if it is one) happened the other way around.

@Talskubilos: Thanks!
My main issue here is that in order to confirm that hypothesis, we would need some idea on when these loans would have happened. Are PIE and PNWC / PNEC assumed to be comtemporaneous? If not, what would be the respective stages of the language families and the concrete forms that underly the loaning? Languages normally loan numbers above the set they have (or, in case of a strong influence of a trade language, higher numbers get replaced first), so a loaned "2" would imply loaned "3" and "4" as well; are there candidates for that in PNEC or PNWC? Do we have more loans, so that we can establish sound correspondences? Without answers to these questions, this doesn't go beyond the "well, maybe" stage.

that the PIE word for "five" is related to/derived from the root for "fist"?

I believe it's conventionally (?) connected with English finger, fist, Latin pugnus; I incidentally find it unusual that such a basic word would be borrowed.

With basic word you mean "fist / finger" or "five"? For the first, I would agree, for loaning of "5" you'll find many examples in Zompist's lists of numbers.

[Talskubilos]

On the other hand, the NWC connection for "two" is pretty unconvincing, since I doubt *qˀʷ would be borrowed as *w.

Maybe I'm splitting hairs, but from those forms, I would've expected the Proto-Indo-European ones to be something like *tkwe- (I believe we see dual initial clusters in PIE, at least)

Good point. PNWC *tˀ- looks like a fossilized prefix, so *qˀʷə would be the lexeme, corresponding to Daghestanian '2' (but Nakh '3', because in this group '2' was apparently borrowed from Semitic). So the correspondence with IE could be the dual suffix *-h₃(u) found in '2' and '8'.

and pimku~pimkwe; where, exactly, do we lose the */q/, and get the */i/ > */e/ (even if it patterned as an allophone of */j/, I'm pretty sure Proto-Indo-European had *[i])?

As *e is one of the cardinal Ablaut vowels, it "destroys" the previous one.

The NEC connection is interesting, but what is the argument that PIE borrowed it from NEC and not the other way around, since I think it's been suggested that the PIE word for "five" is related to/derived from the root for "fist"?

From my own research, words for '5' are usually derived from 'hand', 'fist' and the like.

[Rounin Ryuuji]

I meant a word like "fist" or "finger".

[Talskubilos]

@Talskubilos: Thanks!
My main issue here is that in order to confirm that hypothesis, we would need some idea on when these loans would have happened. Are PIE and PNWC / PNEC assumed to be comtemporaneous? If not, what would be the respective stages of the language families and the concrete forms that underly the loaning?

In an old Russian article (1988), the late Sergei Starostin proposed a number of loanwords from NEC (he regarded NWC a satellite of it) into IE: Indo-European/North Caucasian Isoglosses

Unfortunately, this article has been largely ignored by IE-ists and never translated into English.

Languages normally loan numbers above the set they have (or, in case of a strong influence of a trade language, higher numbers get replaced first), so a loaned "2" would imply loaned "3" and "4" as well;

That's right. In earlier posts I already mentioned some oddities about IE '4' and '8'.