Eurasiatic, Nostratic and Borean

[Rock]

Eurasiatic superfamily includes Indo-European, Uralic–Yukaghir, Altaic, Chukotko-Kamchatkan and Eskimo–Aleut languages. Also, in Eurasiatic superfamily Korean–Japanese–Ainu, Nivkh and Etruscan are sometimes included.
Nostratic superfamily includes even wider range of language families, such as Afro-Asiatic, Kartvelian and Dravidian.
Borean superfamily is the biggest language family I've ever seen. Not only it includes Nostratic and Eurasiatic, it also includes Dene-Daic.
You can even see Borean language tree.https://upload.wikimedia.org/wikipedia/commons/1/14/BoreanLanguageTree.png
Keep in mind, that different version of Eurasiatic, Nostratic and Borean exist. Each version of Eurasiatic, Nostratic and Borean have different reconstructions and different range of languages or language families.
What do you think about this idea? In my opinion, superfamilies are good thing, because it makes a little easier to trace all languages to single source - Proto-Human or Proto-World language.https://en.wikipedia.org/wiki/Proto-Human_language

[Daniel]

Proto-World is the root of a much bigger tree, and, really, not much less plausible than Borean. The absolute best we can do at that time depth is... "maybe".

I'll admit it's fun to look at the trees, and to wonder. And there's also something relatively reasonable about assuming some connections (indeed, there must be some connections), but we shouldn't take any (really, any) details seriously beyond relatively well-established families.

[Forbes]

You are faced with several problems if you want to create superfamilies or trace their history back to a single language.

No one knows when humans first started speaking, but assuming conservatively it was 50,000 years ago that is a very long time when you look at how rapidly languages can change. Irish and Bengali are related but quite different from each other. The furthest we can go back without hypothesising is 5000 years and then only for a tiny number of the world's languages - and even then there are uncertainties about how to interpret ancient texts, especially those not written with alphabetic or syllabic scripts. Families like Indo-European and Afro-Asiatic are the exception rather than the rule.

When comparing two proto-languages you need to bear in mind that you are comparing two hypotheses which puts you on rather shaky ground. If a language family has only been written recently you cannot go as far back with your construction of a proto-language as you can with one which has a millennia-long history. How valid is it to compare a hypothesis that takes you back 7000 years with one which only takes you back a century or two? You also have the problem that languages borrow from each other. If two proto-languages share a lexeme how can you know whether it is because one borrowed from the other or because they had a common ancestor? You have to be careful not to be seduced by apparent resemblances when what is important is correspondences. Semantics are unrealiable - if the word in proto-language A for "rope" corresponds with the word in proto-language B for "snake" it may or may not be significant. If you consider that within Indo-European there is no agreement about how Slavic and Baltic relate to each other you are not going to get agreement about how proto-languages relate to each other.

If you are going back far into pre-history linguistic methods alone are not enough. A multi-disciplinary approach is required with linguistics, anthropology, archaeology and genetics all playing a part with none taking precedence. Whilst interesting theories may emerge, the fact remains that languages leave few fossils and the speed with which languages change rules out going back to when humans first started speaking, whenever that was. The reconstuction of the first language is a chimera.

[Rock]

I'm replying too late, but:

You are faced with several problems if you want to create superfamilies or trace their history back to a single language.

At first, I'm not a linguist,just interested in languages and phonology so I wouldn't create superfamilies. Atleast, very good understanding of grammar would be needed.
Yes, there are some problems for making superfamilies and first language reconstruction. But, I think, common problems for superfamilies or first language is what makes superfamily or first language reconstruction interesting.

No one knows when humans first started speaking, but assuming conservatively it was 50,000 years ago

Reconstructed Proto-Human or Proto-World (not fully recostructed) hypothesized to have been spoken between 100,000 and 200,000 years ago. Atleast this is what Wikipedia are saying.

that is a very long time when you look at how rapidly languages can change.

The Norman Conquest is what makes the English language rapidly changing.

and even then there are uncertainties about how to interpret ancient texts, especially those not written with alphabetic or syllabic scripts.

This is proto-writing. Examples of proto-writing are Vinča symbols and Jiahu symbols.

If two proto-languages share a lexeme how can you know whether it is because one borrowed from the other or because they had a common ancestor?

Yes, this is true. If two proto-languages have something common, we cannot know if this something common are borrowed or not. Indo-Uralic hypothesis is great example of this. In my opinion, Proto-Indo-European and Proto-Uralic are probably descsended from Proto-Indo-Uralic, because this is where is the point of superfamilies.

if the word in proto-language A for "rope" corresponds with the word in proto-language B for "snake" it may or may not be significant.

It may be significant, because similar appearance is what causes rope and snake to be easily confused. With this confusion, snake can be easily interpreted as moving rope. If the words of rope and snake are similar, it can cause even more confusion.

If you consider that within Indo-European there is no agreement about how Slavic and Baltic relate to each other you are not going to get agreement about how proto-languages relate to each other.

In my opinion, most linguists think that Baltic and Slavic related to each other and descended from Proto-Balto-Slavic language.

The reconstuction of the first language is a chimera.

If the reconstuction of the first language is a chimera, then most proto-languages would be chimera, probably because lack of writing. Even Proto-Indo-European. But there are really some people interested in proto-languages, or even first language. Or even proto-Human (Proto-World.) For these people, it's not even important whether proto-language is chimera or not, I think. For me, the most important part is that the proto-language is even reconstructed.

[Forbes]

"If the reconstuction of the first language is a chimera, then most proto-languages would be chimera, probably because lack of writing. Even Proto-Indo-European. But there are really some people interested in proto-languages, or even first language. Or even proto-Human (Proto-World.) For these people, it's not even important whether proto-language is chimera or not, I think. For me, the most important part is that the proto-language is even reconstructed."

The point is that proto-languages are constructed with reference to attested forms. They have to be regarded as provisional for all sorts of reasons. Suppose language A develops into two distinct varieties A1 and A2 spoken today and there is no record of A. Conceptually we have two different things. One is the language A which was something which existed and is unknown. The other is Proto-A which is a hypothetical reconstruction. Whilst linguists may be confident that  the reconstruction is satisfactory (and they will probably concede that some parts are less satisfactory than others) they will not insist that they have revealed to the world what language A was really like. Proto-A is not a chimera because it is a sustainable hypothesis.

When you propose a proto-language what you are saying is: We definitely have a+b+c+d....+n which may give us x. When you compare two proto-languages and suggest a proto-proto-language what you are saying is: We may have x and y which may give us z. We have to be less confident about z than we are about x and y. Comparing proto-proto-proto-languages takes us further into the realms of uncertainty. There will be a limit to how far you can go back and the point you reach will be nowhere near when humans first started to speak. You also have to remember that your starting point was attested languages; your reconstruction cannot be convincing because an unknown number of unattested languages have not been taken into account.

The reconstruction of the first language is a chimera because we simply do not have and never will have sufficient information.

[Rock]

When you propose a proto-language what you are saying is: We definitely have a+b+c+d....+n which may give us x. When you compare two proto-languages and suggest a proto-proto-language what you are saying is: We may have x and y which may give us z. We have to be less confident about z than we are about x and y. Comparing proto-proto-proto-languages takes us further into the realms of uncertainty. There will be a limit to how far you can go back and the point you reach will be nowhere near when humans first started to speak. You also have to remember that your starting point was attested languages; your reconstruction cannot be convincing because an unknown number of unattested languages have not been taken into account.

Yes, this is true. Most superfamily reconstructions are inaccurate. The main thing of inaccuracy is such things as unknown vowel (V) and unknown consonant (C).  For example, if word of sky in Proto-A is reconstructed as *CVCV, it can be any of /tata/, /hoho/, /nini/, etc.
Also, look at the Proto-Borean reconstruction compiled by Sergei Starostin. The most part that covers this reonstruction is also unknown vowels and unknown consonants.

The reconstruction of the first language is a chimera because we simply do not have and never will have sufficient information.

BUT THERE IS some some sort of information of first language also known as Proto-Human or Proto-World, not matter how sufficient.
At first, how is the Proto-Human or Proto-World reconstructed? Quote from Wikipedia:

Ruhlen tentatively traces a number of words back to the ancestral language, based on the occurrence of similar sound-and-meaning forms in languages across the globe. Bengtson and Ruhlen (1994) identify 27 "global etymologies". The following table, adapted from Ruhlen (1994b), lists a selection of these forms:
Based on these correspondences, Ruhlen (1994b:105) lists these roots for the ancestor language:

Keep in mind that Ruhlen reconstructed only 11 words of Proto-Human or Proto-World. That's all I know about Proto-Human or Proto-World.

[Forbes]

Suppose there was a time when all humans spoke the same language. As the speakers dispersed varieties would have emerged and, assuming a lack of contact, in time have become mutually unintelligible. Suppose ten distinct families arose. If we send a comparative linguist back to that time he can have a good guess at what the first language was like. We fast forward a few millennia and suppose that half of the language families have died out. If we send the linguist back to that time not only will he be lacking half the information he needs, but the remaining five languages will have moved further apart from each other and from the original language. Much less for the linguist to go on.

As we stand, we do not know how many language families have been lost. Apart from that not only do words change but they are also subject to semantic shift. We know from exisiting languages that words that sound similar may have dfferent etymologies. Most signficant of all, we cannot be certain that language arose only at one time in one place. Our ancestors may have reached a point where they had not developed fully-fledged language but it was inevitable they would and then separated into different groups.

The analogy is not perfect, but what we have today is like having a handful of jigsaw pieces and not knowing how many pieces the jigsaw has and then speculating on what the picture is like.

[Rock]

As we stand, we do not know how many language families have been lost.

Some of Paleo-European or Pre-Indo-European language families may have been lost, because of Indo-European migrations.
Some of African, American and Oceanic language families may have been lost, because of colonisation.
But how many? We don't know exact number.