Nostratic, Eurasiatic, Mitian, ...

[WeepingElf]

Indeed, the "Haida and Na-Dene" thread is derailed, and I expect it to be locked by the mods soon, thanks to Vlürch's "Indo-Uralic can't be true because it would mean that our noble Finnish race would be related to the Swedish and Russian barbarians who oppressed us all that time" nonsense. And the discussion of Indo-Uralic of course belongs here.

Just a brief comment: AFAIK, Japanese namae is a compound, written with two kanji, and it wouldn't suirprise me if there were on'yomi involved, i.e. borrowings from Chinese.

[linguoboy]

Just a brief comment: AFAIK, Japanese namae is a compound, written with two kanji, and it wouldn't suirprise me if there were on'yomi involved, i.e. borrowings from Chinese.

Not everything written with multiple kanji is a compound; jukujikun are a thing.

In this case, it is a compound, but both elements are native. (In fact, the second element is itself historically a compound of ma "eye" and pe "direction".) The initial element meaning "name" can also be used independently, e.g. 名ばかりの na bakari no "in name only".

[Tropylium]

Etymology
Compound of 名 (na, “name”) +‎ 前 (mae, used as a suffix in words describing a person to emphasize a quality or feature).[1][2][3]

Appears as a term only in the last couple centuries, becoming more widely used during the Meiji period (starting 1868).

OK fair, this is a pretty good example of how to disprove an etymology.

[KathTheDragon]

an (otherwise unknown) root *h₃neh₃-, perhaps 'to call'

Kloekhorst believes that the root can also be found in Hittite ḫannai "to sue, judge" < *h₃eh₃noh₃- (i.e. "call to court") and Greek ὀνομαι "to scold, blame, insult" < *h₃n̥h₃- (i.e. "call names")

[WeepingElf]

an (otherwise unknown) root *h₃neh₃-, perhaps 'to call'

Kloekhorst believes that the root can also be found in Hittite ḫannai "to sue, judge" < *h₃eh₃noh₃- (i.e. "call to court") and Greek ὀνομαι "to scold, blame, insult" < *h₃n̥h₃- (i.e. "call names")

Fine!

[KathTheDragon]

While we're discussing the "name" word, it's interesting to note that Tocharian has *ñem, pointing to a "post-laryngealistic" *nēmn̥, which is a closer match for Uralic *nimi.

[WeepingElf]

While we're discussing the "name" word, it's interesting to note that Tocharian has *ñem, pointing to a "post-laryngealistic" *nēmn̥, which is a closer match for Uralic *nimi.

It cannot be excluded that Proto-Uralic borrowed the 'name' word from an IE dialect (or a lost sister language of PIE) with a similar form, even though 'name' is a basic vocabulary item (Décsy's claim that the early Uralians did not use names and acquired naming only when they met Indo-Europeans is an anthropological impossibility and - like almost everything Décsy wrote - is to be rejected). Such "luxury borrowings" do happen when a language gets into contact with another, more prestigious one.

[hwhatting]

It cannot be excluded that Proto-Uralic borrowed the 'name' word from an IE dialect (or a lost sister language of PIE) with a similar form, even though 'name' is a basic vocabulary item (Décsy's claim that the early Uralians did not use names and acquired naming only when they met Indo-Europeans is an anthropological impossibility and - like almost everything Décsy wrote - is to be rejected). Such "luxury borrowings" do happen when a language gets into contact with another, more prestigious one.

Case in point - Uzbek has three words for name - inherited Turkic ot, Persian nom, and Arabic ism, of which ism (mostly for personal names) and nom (mostly for non-personal names, e.g. place names, file names, etc.) seem to be much more frequent in use than ot, from the limited amount of Uzbek that I've seen.

[Vlürch]

Sorry for derailing the other thread... I promise I won't do it again. I'll stick to the linguistic stuff and relevant history, avoiding politics and everything else that's irrelevant.

This pdf about Indo-European loanwords in Uralic languages is really interesting. The author makes a bunch of implications that could be used to support any side of the debate, like seems to almost always be the case when it comes to the historical relations between Indo-European, Uralic and Altaic languages. Anyway, she argues that there are no ancient Indo-European loanwords in Uralic languages, or at least that some of them are just chance resemblances, and that there may have never been a single Proto-Uralic and Proto-Indo-European at all, etc.

I personally don't buy that they're all just coincidences, but isn't it possible that they're loanwords into both from a third source; maybe the Botai if they're ancient enough and Hunnic if they're more recent? If the word for horse in Ugric languages is a possible Botai loan, couldn't there have been more and in different languages? What if the words that can't be conclusively identified as loanwords or cognates between Uralic, Indo-European and Altaic languages were borrowed from them? I mean, if the Botais domesticated the horse, they could've spread all across Eurasia, disseminating loanwords and other influences into various languages and cultures.

It wouldn't be that far-fetched for them to have been at least partially ancestral to the Huns, who are known to have spread across Eurasia on horses. Assuming some connection between the Xiongnu and all the other groups called "Huns", it'd be plausible for many loanwords to exist and remain undetected until they're compared and a reconstruction is attempted but fails to be explained. Maybe that's too many assumptions, but still, it's better than concluding that everything is a coincidence or time travel.

Indeed, the "Haida and Na-Dene" thread is derailed, and I expect it to be locked by the mods soon, thanks to Vlürch's "Indo-Uralic can't be true because it would mean that our noble Finnish race would be related to the Swedish and Russian barbarians who oppressed us all that time" nonsense.

My only argument on that matter is that unless someone can conclusively prove it by excluding all other relations, I refuse to believe that Indo-Uralic by itself, alone, without any relation to other language families, is true. That's what I was saying, not that Uralic and Indo-European aren't related at all. Yeah, what I said was based on emotion and as such invalid and I shouldn't have said it (but being emotional makes people say stupid things that would've been better left unsaid (like everything from that post onwards)), but the point is that AFAIK there isn't even any evidence to suggest that Indo-Uralic is a separate genetically related family, especially one that has no relation to any other languages whatsoever, or even that it's one united branch within a larger Eurasiatic family, without including at least Altaic as well. All three could be equally distant from/close to each other, or Uralic and Altaic could be more closely related to each other than IE but more extensive and later IE influence in Uralic could have clouded that, or that IE and Uralic are more closely related but that there was some intermediary branch of Eurasiatic between them, etc. It's also possible none of them are related at all to each other except through contact, but that I just can't believe.

For me lumping is fun, but lumping in ways that excludes history and modern similarities is pointless. Uralic and Altaic had contacts and both had contacts with IE, so saying that Uralic and Altaic have no relation at all but that IE and Uralic or IE and Altaic are related is just weird. When lumping, all the lumpables have to be considered because it creates a wider net that also makes it easier to identify loanwords or innovations, etc. That way, every proto-word can be reconstructed individually, and if they match at all but not necessarily by regular sound correspondences, it's possible that the cognates will fall in place and the rest are shared borrowings from an unrelated language.

And even if languages have many regular correspondences but no common ancestor can be reconstructed, the languages could still be related, just not by descent. Instead of sister languages, they could be cousin languages or spouse languages or something. I've been told before that that counts as creolisation, but I don't think it does unless it also involves simplification and other things that aren't necessary in that scenario but do happen with creoles. At least that's how I understand it, creoles being necessarily impossible to categorise into a language family and being in some way simplified, but I could be wrong.

There's also the possibility that there have been language families that can't be reconstructed because all that's left are isolates like Basque, Sumerian, Burushaski, Nivkh and Ainu, etc. but that influenced other languages in ways that can be researched. Maybe the extinct families they belong to were also part of the bigger Eurasiatic family, or at least Pre-Proto-Eurasiatic or something. Then again, maybe they weren't, but it's likely that there's at least some truth behind the lumping even if it's not genetic relationship. There could've been ancient civilisations that have been lost more or less without a trace, so their languages can't be reconstructed and can only be speculated to have existed, too. If enough is learned about ancient languages and enough proto-languages or even just scattered proto-words are reconstructed, it could lead to new progress being made in determining what ancient languages were like even if they're not exact matches and there's a lot of variation that can't be accounted for, like alternative reconstructions and whatnot. It could just mean that there were different dialects or something in the proto-language that formed "para-languages" (like Khitan being para-Mongolic), or that the words were borrowed from different related languages that are (currently) unknown, etc.

Maybe it's just because I'm a "fan" of Starostin, though, that I think that way... but still, it doesn't hurt anyone to "reconstruct" something that never existed, and it could even lead to new discoveries if something else is realised to be true in the process. Most likely, no one will ever know for certain either way. If nothing I wrote makes sense, sorry; I'm trying to make sure that what I'm saying isn't too vague but also avoiding making assumptions about connections between languages that haven't been demonstrated to have any relations or I'd ramble on forever about the same things I've already rambled about before...

It cannot be excluded that Proto-Uralic borrowed the 'name' word from an IE dialect (or a lost sister language of PIE) with a similar form

Realistically, what else could it even be than an IE loan? Also, at least according to Wiktionary, the PU is *nime while *nimi is just Proto-Finnic, and is the word used in Finnish, Estonian and apparently all other Finnic languages except Livonian.

The Mongolian word for "name" also has /n/ in it: нэр. According to Wiktionary it's pronounced with an /i/ rather than /e/ even though it's spelled with <э>, too. Not proof of anything, but interesting anyway.

[KathTheDragon]

I refuse to believe that Indo-Uralic by itself, alone, without any relation to other language families, is true.

Nobody's claiming this. To think that Indo-Uralic just appeared by magic is nonsensical.

without including at least Altaic as well.

You speak as if Altaic is a proven family.

Realistically, what else could it even be than an IE loan?

What a compelling argument.

Also, at least according to Wiktionary, the PU is *nime while *nimi is just Proto-Finnic, and is the word used in Finnish, Estonian and apparently all other Finnic languages except Livonian.

Ah, my bad.

[Tropylium]

This pdf about Indo-European loanwords in Uralic languages is really interesting. The author makes a bunch of implications that could be used to support any side of the debate, like seems to almost always be the case when it comes to the historical relations between Indo-European, Uralic and Altaic languages. Anyway, she argues that there are no ancient Indo-European loanwords in Uralic languages, or at least that some of them are just chance resemblances, and that there may have never been a single Proto-Uralic and Proto-Indo-European at all, etc.

Ah yes, Angela Marcantonio, notorious crackpot. As you might be able to smell from the abstract already: "borrowing among proto-languages is impossible a priori" — note that she does not mean the case of PIE and PU in particular, she really claims that it would be logically incoherent for any proto-language to have ever possibly borrowed words from any other proto-language:

This process of interference and subsequent borrowing takes place within the concrete ‘speech act’ of real speakers, that is, at the level of ‘parole’ (to use Saussurian terminology), with all its social and geographical variants, and not between and/or at the level of, abstract linguistic systems, as the like of ‘reconstructed’ languages / language families (see Gusmani 1986: 137–138). (…) This being the case, it can be argued that borrowing among proto-languages is not only unlikely, but is actually impossible a priori.

…Where she is either ignoring or somehow failing to realize that "abstract linguistic systems" only ever exist in the form of some particular real idiolects, even if we do not claim to reconstruct the exact idiolects behind a reconstructed proto-language.

Ironically though, I still agree with her main thesis: there is no such thing as secure cases of PIE loans into PU. All cases can be interpreted, and several cases should be interpreted, as being loaned in parallel from early IE dialects such as Proto-Indo-Iranian into early Uralic dialects such as Pre-Proto-Finnic.

isn't it possible that they're loanwords into both from a third source; maybe the Botai if they're ancient enough and Hunnic if they're more recent?

Possible in principle for sure, but without any kind of limits on what we might hypothesize to come from Botai or other extinct cultures' languages, it's also basically the "aliens did it" approach to etymology. Or if we do limit ourselves to typical loanwords such as 'horse', well, then we're back to square one when it comes to typically native lexicon such as 'I' or 'die' or 'tall'.

Also, at least according to Wiktionary, the PU is *nime while *nimi is just Proto-Finnic, and is the word used in Finnish, Estonian and apparently all other Finnic languages except Livonian.

Ah, my bad.

That's the usual uncertainty in how to reconstruct the PU unstressed vowel subsystem. The "traditional" reconstruction, proposed since the 1890s, would be *nime; the "updated" reconstruction, proposed since the 1980s, would be *nimi; and the "minority" reconstruction, first proposed in the 1930s but revived less than ten years ago, would be *nimə.

I find the last of these best-argued-for: it is based on the comparative evidence of the vast majority of Uralic, while the first is based solely on Finnic, and the second is based on no direct evidence whatsoever, only on an IMO misunderstood application of the principle of phonological dispersion.

[kanejam]

For me lumping is fun, but lumping in ways that excludes history and modern similarities is pointless. Uralic and Altaic had contacts and both had contacts with IE, so saying that Uralic and Altaic have no relation at all but that IE and Uralic or IE and Altaic are related is just weird. When lumping, all the lumpables have to be considered because it creates a wider net that also makes it easier to identify loanwords or innovations, etc. That way, every proto-word can be reconstructed individually, and if they match at all but not necessarily by regular sound correspondences, it's possible that the cognates will fall in place and the rest are shared borrowings from an unrelated language.

Proving a relation doesn't mean disproving all other relations, and saying that investigating Indo-Uralic without also investigating Altaic is weird is just bizarre - if we somehow prove Indo-Uralic and then somehow discover that Altaic forms a subbranch with Uralic, what have we lost? Albanian and Anatolian certainly didn't factor into the original PIE reconstruction but that didn't make it invalid. We should rather look at likely relation based on similar morphology and similar lexical items, as is the basis for Altaic, but that certainly doesn't guarantee relation.

And no we shouldn't just grab words with similar meanings and make proto-forms for them because that is not the comparative method, but rather what Kath's linked article refers to aptly as 'wordlist linguistics'. I don't disagree that doing so can sometimes lead to valid discoveries of cognates, but claiming that it is scientific or valid is wrong.

And even if languages have many regular correspondences but no common ancestor can be reconstructed, the languages could still be related, just not by descent. Instead of sister languages, they could be cousin languages or spouse languages or something. I've been told before that that counts as creolisation, but I don't think it does unless it also involves simplification and other things that aren't necessary in that scenario but do happen with creoles. At least that's how I understand it, creoles being necessarily impossible to categorise into a language family and being in some way simplified, but I could be wrong.

I don't know what you're trying to say here. Being related in a historical linguistics sense implies a shared ancestor - contact is the other option, with intense contact leading to mixed languages, which is probably the term you want rather than creole.

[Howl]

I do think that there is probably a core-Eurasiatic family with at least Indo-Uralic and micro-Altaic. But it will be very difficult to prove something like that. We do have a good sense of what the proto-languages of Indo-European and Uralic were like 5000 years ago. We don't have that for Turkic, Mongolic and Tungusic. And there are many clues that IE and Uralic are more closely related to each other than to the rest of Eurasiatic. So to me it makes sense to compare the two and see what we can get from that.

[WeepingElf]

I concur with you that IE and Uralic are probably more closely related to each other than to Turkic, Mongolic and Tungusic. It is uncertain, though, whether the latter three form a valid node or not. Turkic looks as if it was closer to Indo-Uralic than the other two, but that may be deceptive (the main point is Turkic's personal endings, which Mongolic and Tungusic lack, but that's probably just a shared retention which doesn't prove anything).

The Yukaghir, Chukotko-Kamchatkan and Eskimo-Aleut languages, which I also consider parts of the Mitian macrofamily, are another matter. Eskimo-Aleut looks especially close to Uralic, but that's IMHO probably just due to the conservativism of these two families. Yukaghir is often held to be the closest living kin of Uralic, but the lexical resemblances between the two look more like borrowings from a Samoyedic or Para-Samoyedic language into Yukaghir, and the morphologies are quite different (though Yukaghir looks undeniably Mitian). Chukotko-Kamchatkan is weird, and Fortescue has dropped it from the Uralo-Siberian family (U, Y, ChK, EA) he had proposed in his 1998 book Language Relations across Bering Strait, in an article published in the Lingua journal in 2011 (I think), proposing a connection to Nivkh instead.

Fortescue's reconstruction of Proto-Uralo-Siberian may actually give a good idea of what Proto-Mitian could have looked like, a better one IMHO than either Dolgopolsky's or Bomhard's Proto-Nostratic, which both include data from Afroasiatic and Dravidian, entities which IMHO probably are nowhere close to Mitian, resulting, among other problems, in a bloated phoneme inventory (Bomhard reconstructs 50 consonants, while most Mitian languages are somewhere around 20, with the PIE system probably beefed up by rather recent changes).

[Howl]

I concur with you that IE and Uralic are probably more closely related to each other than to Turkic, Mongolic and Tungusic. It is uncertain, though, whether the latter three form a valid node or not. Turkic looks as if it was closer to Indo-Uralic than the other two, but that may be deceptive (the main point is Turkic's personal endings, which Mongolic and Tungusic lack, but that's probably just a shared retention which doesn't prove anything).

Evenki (Tungusic) also has personal endings.

The Yukaghir, Chukotko-Kamchatkan and Eskimo-Aleut languages, which I also consider parts of the Mitian macrofamily, are another matter. Eskimo-Aleut looks especially close to Uralic, but that's IMHO probably just due to the conservativism of these two families. Yukaghir is often held to be the closest living kin of Uralic, but the lexical resemblances between the two look more like borrowings from a Samoyedic or Para-Samoyedic language into Yukaghir, and the morphologies are quite different (though Yukaghir looks undeniably Mitian). Chukotko-Kamchatkan is weird, and Fortescue has dropped it from the Uralo-Siberian family (U, Y, ChK, EA) he had proposed in his 1998 book Language Relations across Bering Strait, in an article published in the Lingua journal in 2011 (I think), proposing a connection to Nivkh instead.

If Eskimo-Aleut is so Mitian, where are the M-T pronouns? When I look at the pronouns of various EA languages, they're not M-T. And that puts it in the same category as Afroasiatic, Dravidian and Basque for me.

[WeepingElf]

Evenki (Tungusic) also has personal endings.

I didn't know that.

If Eskimo-Aleut is so Mitian, where are the M-T pronouns? When I look at the pronouns of various EA languages, they're not M-T. And that puts it in the same category as Afroasiatic, Dravidian and Basque for me.

Sure, the independent pronouns cleverly hide their Mitianness, but the personal paradigms of Eskimo-Aleut are remarkably similar to those of some Uralic languages such as Northern Samoyedic, Khanty and Mansi. Also the dual marker -k and the plural marker -t.

[Tropylium]

If Eskimo-Aleut is so Mitian, where are the M-T pronouns? When I look at the pronouns of various EA languages, they're not M-T. And that puts it in the same category as Afroasiatic, Dravidian and Basque for me.

What counts as "Mitian" on the surface kind of depends on what specific person markers we are looking at: the independent pronouns? the verbal endings? the possessive suffixes? They're all pretty closely linked in many of the languages (in some languages like Sami or Mari all three are equally "Mitian"), plus they form a kind of a three-member chicken-and-egg puzzle: which came first?

I'm kind of warming to the idea that the possessive suffixes might be the oldest "Mitian" markers. In Uralic there are precedents both for (1) forming pronouns' oblique cases by attaching possessive suffixes to a demonstrative stem, and (2) for more than one set of verbal endings (the "determinativity" contrast), with one set identical to the poss. suff. and the other distinct; IE does "suppletive nominatives" which might similarly point to the oblique cases having been re-built in some fashion. Still, the "typical" pronouns would in this case seem to involve prefixation, not suffixation.

I don't have an opinion on if Dravidian is especially closely related to the other Nostratic candidats, but I think comparison with it is likely going to be more useful than many other fringe contenders. It's been hanging in fairly central and well-traffic'd parts of Eurasia for a while in any case, so I wonder if we could not use it as an aid for internal reconstruction of e.g. IE or Uralic (or Elamite, or Sumerian, etc. take your pick), regardless of if the lexical parallels are cognates or not.

By contrast Eskaleut sure has a few promising morphological data points, but lexically it is so far out there — both due to the polysynthesis and, probably, due to absorbing several substrata on its way across Siberia and Alaska — that you cannot really use it to reconstruct much of anything. Yukaghir and Mongolic might have similar problems (too young, too mobile). To a lesser extent this problem also comes up with the "family-internal isolates" like Armenian and Hungarian (and maybe some AA examples like Beja).

[Howl]

My own hyper-speculative theory is that mV/tV were originally demonstratives. And a path to 1st/2nd person via possession (this thing (near me) -> my thing; that thing(near you) -> your thing) would fit in nicely with that. The "determinativity" contrast in Uralic also has parallels with the IE stative. But it is hard to reconstruct a paradigm from these second sets of verbal endings.

[WeepingElf]

Yes, I consider it plausible that the "Southeast Uralic" (i.e., Hungarian and Selkup) indefinite conjugation was cognate to the IE perfect/middle conjugation, though only the 1st person singular forms match well (the 2nd person is different in Hungarian and Selkup, and neither form looks like the IE one). The underlying system may have been an active-stative one. The "Northeast Uralic" (Northern Samoyedic and Khanty-Mansi) pattern, where object number but not object person is marked on the transitive verb, and which Seefloth reconstructs for Uralo-Siberian on the ground of a very similar paradigm in Eskimo-Aleut, may be a residue of an earlier Indo-Uralic/Proto-Mitian bipersonal transitive conjugation of the kind usually found in active-stative languages.

[mèþru]

I think that Pre-PIE was an active-stative language.

[WeepingElf]

I think that Pre-PIE was an active-stative language.

So do I.

[hwhatting]

Highly likely.

[Tropylium]

by default true soundlaws are much less common than borrowings and coincidences

I'm pretty sure that there are way more regular sound correspondences than loanwords between e.g. Tocharian and Latin.

But that's a biased sample - because we know Tocharian and Latin are related!

That's however not prior information: it's something that follows from that they show several regular resemblances while being way too far away for this to be possibly explainable through loaning.

Chance resemblances are indeed always a risk, but once we have enough data to rule this out, relatedness probably requires fewer additional points of evidence than loaning: it is possible regardless of distance, while loaning only works between neighboring languages.

(Semantics alone doesn't make one or the other a "more default" hypothesis; they both stand or fall depending on how well the similarities cover core lexicon and grammatical morphemes. Going just by this, Indo-Uralic would be a shoo-in, but it's of course not sufficient.)

But it's not either/or. There are also compromise solutions: loaning thru a related language that was spoken closer by. There's a whole bunch of similar vocabulary between English and Karelian, even though the two have never been in substantial contact… yet it's not because of relatedness, it is because they've come in through the more close-by Proto-Scandinavian to Swedish lineage (related to English) and Finnish (related to Karelian). Your scenario of Para-IE hanging out near the Urals and coming in contact with a more east-derived pre-Uralic would be one application of this to Indo-Uralic. I wonder if the opposite might also work: there have been many cultural drifts coming from central Eurasia to eastern Europe around the Neolithic, and one or more of them could have put some old Para-Uralic group in touch with pre-Indo-European speakers.

But the odds of two lanugages being (relevantly closely) related or unrelated is not 50/50. It's hundreds to one

"Relevantly closely" is doing a lot of work here.

The odds of two languages being related at all at any level of history is something like 95% (we can leave some room for polygenesis, but given the general low amount of isolates in Africa or even the Near East, this is unlikely to make a big difference). The odds of them being obviously in a common language family is far less, and strongly depends on distance. But there's middle ground, too: languages can be demonstrably related while not being obviously related.

Most language families so far have been put together by "transitive obviousness". English is obviously related to Old English is obviously related to Gothic is obviously related to Latin is obviously related to Greek is obviously related to Sanskrit is obviously related to Middle Indo-Aryan is obviously related to Hindi, though English is not obviously related to Hindi; even Old English probably not to MIA. But there are major counterexamples as well. Within IE that's cases like Armenian and Albanian on one hand (requires unraveling layers of loanwords and extensive sound changes); Tocharian and Anatolian on the other (only really clear by comparison primarily with the proto-language).

And there is no consensus yet on what's sufficient evidence for establishing non-obvious relatedness. If you ask Africanists, Niger-Congo "is obvious" already based on just a single typological feature, i.e. class prefixes; if you ask people working with languages in Eurasia 150 years ago, they would also think that a grouping like Uralo-Altaic "is obvious" per multiple similar kinds of evidence (vowel harmony, front rounded vowels, possessive suffixes, local cases…); and yet people working with these languages today would think it's not, and that either they're mostly still not demonstrated related. (The one exception that's been settled since then is that Samoyedic belongs in Uralic, which took similar work as the cases of Tocharian and Anatolian.) And if you ask anthropologists, they may think that genetics + the rarity of primogenesis is a slam-dunk argument that already proves that Amerind exists, Proto-Australian exists etc.

To an extent this is more of a disagreement about what "relatedness" means thank about the actual strength of evidence. The whole typological area of noun-class prefixes in Africa, or vowel harmony in northern Eurasia, or dental-alveolar-retroflex-palatal consonat systems in Australia, does have to have a single main source, it cannot have happened just by accident. But we know by now that typological realignment or independent geographically adjacent innovation, while rare, can happen. Therefore this type of argument doesn't show that the entire area belongs in a single language family, only that some substantial parts probably do, and we need more fine-grained research to figure out which exactly.

Linguistic reconstruction deeper than the level of currently established families would probably benefit from more application of Fortescue's concept of "meshes": areas where we can assume a common typological profile and common lexical parallels to have existed even if we cannot yet say what parts of it are due to contact and what parts due to common descent. Altaic is obviously one mesh, so is Indo-European / Caucasus / Semitic, so is the Pacific Northwest, so are probably e.g. parts of Nilo-Saharan. But we don't seem to have a whole lot of reliably reconstructed meshes around. Even Fortescue's own Uralo-Siberian mesh doesn't really stand up to scrutiny: he has since then relinquished his attempts at including Chukotko-Kamchatkan by extensive internal reconstruction, and his treatment of Yukaghir has some similar issues (partly originating from attempts at proving Uralo-Yukaghir).

[WeepingElf]

Most language families so far have been put together by "transitive obviousness". English is obviously related to Old English is obviously related to Gothic is obviously related to Latin is obviously related to Greek is obviously related to Sanskrit is obviously related to Middle Indo-Aryan is obviously related to Hindi, though English is not obviously related to Hindi; even Old English probably not to MIA. But there are major counterexamples as well. Within IE that's cases like Armenian and Albanian on one hand (requires unraveling layers of loanwords and extensive sound changes); Tocharian and Anatolian on the other (only really clear by comparison primarily with the proto-language).

Yes. It is hard to see that Hindi and English are related; we know because there are the intermediate "stepping stones" you mentioned. Without such "stepping stones", a relationship is much harder to demonstrate! For instance, Lydian and Lycian were known before Hittite was discovered, but until Hrozný showed that Hittite was IE, these languages were not understood, and nobody had an idea where to put them - some scholars suspected an IE connection already, but the matter was controversial, and others suspected a relationship to Etruscan, Semitic or whatever. Only after Hittite provided the "missing link", it could be shown that Lycian and Lydian were IE languages, belonging to the same branch as Hittite. (Neither is a direct descendant of Hittite, though; while it is generally agreed that Lycian descended from Luwian, a sister language of Hittite, the exact position of Lydian within Anatolian is still controversial: some scholars consider it another descendant of Luwian, while others opine that it descends from an unattested fourth (besides Hittite, Luwian and Palaic) Bronze Age Anatolian language.)

And there is no consensus yet on what's sufficient evidence for establishing non-obvious relatedness. If you ask Africanists, Niger-Congo "is obvious" already based on just a single typological feature, i.e. class prefixes; if you ask people working with languages in Eurasia 150 years ago, they would also think that a grouping like Uralo-Altaic "is obvious" per multiple similar kinds of evidence (vowel harmony, front rounded vowels, possessive suffixes, local cases…); and yet people working with these languages today would think it's not, and that either they're mostly still not demonstrated related. (The one exception that's been settled since then is that Samoyedic belongs in Uralic, which took similar work as the cases of Tocharian and Anatolian.) And if you ask anthropologists, they may think that genetics + the rarity of primogenesis is a slam-dunk argument that already proves that Amerind exists, Proto-Australian exists etc.

Africanists are much more inclined towards macro-relationship than today's Eurasianists, it seems, though the support for Nilo-Saharan seems to be waning lately, and Khoisan has pretty much been abandoned. I feel that the relationship between, say, Semitic and Cushitic is probably almost as deep as that between IE and Uralic, making Afroasiatic a macrogroup like Mitian (note that there is no consensus on Proto-Afroasiatic reconstruction; I have heard of two reconstruction attempts, one by Ehret and one by Orel and Stolbova, which use different sound correspondences and reconstruct different proto-systems and proto-vocabularies; they cannot both be right, so they are probably both wrong), and Niger-Congo probably is a similar thing. Yet, I am no Africanist and cannot judge these matters properly.

Amerind, as you all know, is rejected by almost every Americanist, partly because of the poor method and poor evidence, but partly certainly because of its uselessness in language classification. What use is a "classification" which includes, within the domain of indigenous American languages, everything except two families? It is pointless. Australianists, I have heard, have recently grown sceptical about a "Proto-Australian", and distinguish 28 (or so) language families. Again, I know too little about these matters to arrive at a valid opinion.

To an extent this is more of a disagreement about what "relatedness" means thank about the actual strength of evidence. The whole typological area of noun-class prefixes in Africa, or vowel harmony in northern Eurasia, or dental-alveolar-retroflex-palatal consonat systems in Australia, does have to have a single main source, it cannot have happened just by accident. But we know by now that typological realignment or independent geographically adjacent innovation, while rare, can happen. Therefore this type of argument doesn't show that the entire area belongs in a single language family, only that some substantial parts probably do, and we need more fine-grained research to figure out which exactly.

If a bunch of neighbouring languages are typologically similar but lack cognate lexicon or morphology (or the lexical correspondences are phonologically so trivial that they reveal borrowing), one is probably dealing with convergence. South Asia is a case in point. We know that the substantial typological resemblances between Indo-Aryan, Dravidian and other languages of the region are not due to a common ancestor showing these features because we know that Indo-Aryan descends from something (PIE) that doesn't show them. But that doesn't mean that Indo-Aryan borrowed these features from Dravidian, either - we do not know for how long Dravidian, which unlike Indo-Aryan has no known external relatives, had them. Some are also indeed from Indo-Aryan, e.g., the breathy-voiced stops, a feature already in place in Vedic, and according to the current majority opinion - which has been questioned, see the discussion in the Great PIE Thread - inherited from PIE, which have found their way into various non-Indo-Aryan languages of India through learned borrowings from Sanskrit.

Linguistic reconstruction deeper than the level of currently established families would probably benefit from more application of Fortescue's concept of "meshes": areas where we can assume a common typological profile and common lexical parallels to have existed even if we cannot yet say what parts of it are due to contact and what parts due to common descent. Altaic is obviously one mesh, so is Indo-European / Caucasus / Semitic, so is the Pacific Northwest, so are probably e.g. parts of Nilo-Saharan. But we don't seem to have a whole lot of reliably reconstructed meshes around. Even Fortescue's own Uralo-Siberian mesh doesn't really stand up to scrutiny: he has since then relinquished his attempts at including Chukotko-Kamchatkan by extensive internal reconstruction, and his treatment of Yukaghir has some similar issues (partly originating from attempts at proving Uralo-Yukaghir).

The "mesh" is a useful notion. Altaic is perhaps the paragon example of a mesh - a set of three (some say, five or even six) language families which show resemblances which are hard to decide whether they are due to inheritance from a Proto-Altaic or to convergence. The lexical resemblances between Turkic, Mongolic and Tungusic look more like borrowings; the typological similarities (all three families consist of agglutinating SOV languages with converbs) are too general to ascribe them to a common ancestor; the resemblances in some morphology, especially the pronouns, are perhaps best explained by a common ancestor, but in this case, some "clearly" non-Altaic languages, most relevantly IE and Uralic, show similar items - hence the notion of the macrofamily variously called "Nostratic", "Eurasiatic", "Mitian", "Steppe" or some other names, which this thread is all about.

Fortescue's "Uralo-Siberian" is a difficult case. Fortescue himself has grown sceptical about the inclusion of Chuktoko-Kamchatkan, which he now prefers to connect with Nivkh, and the inclusion of Yukaghir seems to hinge on the tacit assumption that it was related to Uralic, but the lexical similarites between Yukaghir and Uralic rather look like borrowings from a Samoyedic or "Para-Samoyedic" language, and the morphologies, apart from the "Mitian" pronouns, do not resemble each other much. This would reduce "Uralo-Siberian" to a binary relationship between Uralic and Eskimo-Aleut, which has been proposed much earlier (Rask already commented on morphological similarities between these two families in his Undersøgelse in 1814!). And I feel that the seeming closeness of Uralic and Eskimo-Aleut within Mitian is merely due to conservativism of these two families.

And what regards PIE, to me it looks like a Mitian language drawn into a prehistoric West Asian Sprachbund with the three Causasian families, Semitic and perhaps a few lost ones.

[mèþru]

I think the macro-families of Africa should be split until there is conclusive evidence of relations - that goes to Atlantic–Congo as well.