Linguists identify 15,000-year-old ‘ultraconserved words’ | Hacker News Reader

Linguists identify 15,000-year-old ‘ultraconserved words’

washingtonpost.com

108 pointstokenadult56 comments

I'm going to reserve judgment on the findings themselves until I read the paper (although, having a background in historical linguistics, I've learned to be very skeptical of any claim that posits historical superfamilies.)

However, I will complain about the pop-science intro, because the thought that a hunter-gatherer has some chance of understanding a sentence like the one given is laughably preposterous. An Old English speaker would only barely understand it (just as we would only barely understand gief þæm menn ealdan þisne fýr[1]), to say nothing of the many thousands of years of change between us and the hunter-gatherers spoken of. There is a vast difference between 'two words have a statistically provable correspondence' and 'two words are understood as the same word.'

[1]: My old English is pretty weak, so I take full responsibility for grammatical errors.

Edit: accidentally used 'th' instead of 'þ' in the Old English.

What do you think about theories like Vilayanur Ramachandran's that some words exist because they are sort of onomatopoeic?

(http://www.bbc.co.uk/radio4/reith2003/lecture4.shtml)

I can certainly believe it for one of the examples given, "to spit". The audio examples are clearly imitative, especially in the terminal syllable, which almost always uses a consonant like [p]. (Just like someone spitting.)

http://www.washingtonpost.com/wp-srv/special/national/words-...

That is someone's theory? That's what 'onomatopoeia' means.

I've really dumbed it down, and horribly mangled it.

Read the link I provided, he does a much better job there. If you don't want to read all of it just search for "Now, remember I said the third thing you have to do in science is show that this is not just some quirk.", and read from the top of that section.

'slush' is onomatopoeic. Is 'axe'? How about 'milk' or 'mother'?

Totally agree with your skepticism. Even the Uralic-Altaic superfamily has been rejected for decades (http://en.wikipedia.org/wiki/Ural%E2%80%93Altaic_languages), let alone a combination of seven families.

Could you convert your example to phonetic english? Spelling is one thing, pronunciation another.

In the IPA: /gyf θæːm menː ˈældan ˈθisne fyɹ/

In rough English approximation: güf thæm men ÆL-dun THEES-nuh für, where ü is as in German and æ is the vowel sound in 'cat' in most dialects (i.e. not in the dialects of Scotland or New Zealand.)

I think the more normal word order would be “gief þæm ealdan menn þisne fyr” (if I am correct in understand the sentence you're trying to do). As for the pronunciation:

1. the <g> in the first word manifests in that particular phonological environment as [j] (palatal approximant, as in English "y"), if I'm not mistaken.

2. The <ie> is pronounced either [i:] or as a diphthong [ie] (but probably not [y]). The diphthong is probably older, and sort of has an Old Saxon twang to it, if I may say.

3. <ea> is not pronounced [æ] but rather as a super cool-sounding diphthong, [æɑ].

4. There's not consensus as far as I know about the pronunciation of <r> in OE. I tend to go with the alveolar trill, but other options are the alveolar approximant (as you have suggested) and the alveolar tap.

So, an amended pronunciation gloss might look like this:

    yeef thæm ÆAHL-dun men THEES-nuh für.

yeef thæm ÆAHL-dun men THEES-nuh für.

Thanks. I'm sure all agree that's much clearer.

I'd be very skeptical of this as well. I am Albanian, and expect mother (mama), and father (baba), the rest of the words just don't sound the same at all. Albanian is a separate branch of Indio-European languages (so is Greek, Latin, etc), so it should have some words that are similar to these ultra conserverd word.

This is the citation to the underlying research study.

Mark Pagel, Quentin D. Atkinson, Andreea S. Calude, and Andrew Meade. Ultraconserved words point to deep language ancestry across Eurasia. PNAS 2013; published ahead of print May 6, 2013,

doi:10.1073/pnas.1218726110

Abstract link:

http://www.pnas.org/content/early/2013/05/01/1218726110

Link to .PDF full text:

http://www.pnas.org/content/early/2013/05/01/1218726110.full...

languagehat writes, over at metafilter:

Yeah, this is complete bullshit. Actual linguists weigh in in this LH thread [ http://www.languagehat.com/archives/004994.php ];

Marie-Lucie Tarpent writes:

The article announces "Eurasiatic" (a hypothetical supergroup originally suggested by Joseph Greenberg of "Amerind" notoriety) as a major discovery by scientists (not linguists) but seems somewhat confused as to its relation with PIE. It also quotes a few words which the authors interpret as culturally too important to have been lost or replaced, so that they have lasted 15,000 years (one of these words is "bark" (of a tree) for which the authors propose an explanation). Like Renfrew and the Proto-World people, the authors do not seem to differentiate between the survival of a lexical item (although made unrecognizable through millennia of phonological changes) and the survival of the sounds that compose it (which are independent of the meaning of the whole word) (eg Renfrew et al thought that words for 'nephew' had endured almost unchanged for hundreds of years because of the importance of this concept, but actually this longevity is due to the fact that the consonants in the word had been more resistant to change than others, as shown by the behaviour of those consonants in other words totally unrelated semantically to 'nephew'). In any case, the article does not cite any actual forms, only meanings. Read it at your own risk.

And Piotr Gąsiorowski says:

This is exactly the kind of approach which makes wishful thinking look like science and gets it past reviewers. Even if the numerical methods are basically sound, the data are garbage (obtained by the intuitive eyeballing of reconstructions from the Tower of Babel database -- itself a highly questionable source -- without any actual comparative analysis). No different from ordinary "mass comparison", except perhaps for a tighter control of semantic matches. [...] I realise that one has to start somewhere, but mass comparison is at best of some help in formulating preliminary hypotheses, and mass comparison based on unreliable data is no use at all.

Very few reputable linguists believe in a hypothetical supergroup of "Eurasiatic". The reason that the same linguists are skeptical is that related languages diverge fast enough that after a few millenia, any similarities beyond that are indistinguishable from chance similarity.

I don't get this. I know this is a monstrously crude approach, but I put the sentence given in the article to Google translate

"You, hear me! Give this fire to that old man. Pull the black worm off the bark and give it to the mother. And no spitting in the ashes!"

After translating it into a few different languages, almost none of the words sound anywhere close to their English counterparts. French and Danish had a fair bit of similarities but that's to be expected. Am I fundamentally misunderstanding something here?

For reference I tried estonian, albanian, czech, finnish, and georgian

Try setting it up, walking away for 5 minutes or so (clearing your mind of the original English), and then comparing the translations that aren't English.

The idea is that the fact that we understand English gives it a different quality, and you want to subtract that from your analysis.

Don't know if it'll help, but it might.

I believe you are confusing the spoken sound of a particular cognate with the semantic meanings of the word that have remained consistent across different languages.

For example, "fire" has a very direct translation in a large number of languages. Historically, it was believed that the evolution of word-use would preclude such semantic stability, but the words these researchers identified are perhaps suggestive of a common proto-Eurasiatic language. The fact that less used words like "flow" and "worm" remained consistent legitimizes this theory.

>Am I fundamentally misunderstanding something here?

Perhaps.

For one, Google translate is not very accurate, even for simple stuff.

Second, the translations it picks might no be the form that has survived from the old times. Newer words might have put the use of the older form in disfavour, but it can still exist in the language. If something was the same in a language from 15,000 years ago up to 500 years ago, it won't be picked up in Google Translate, which will pick the current vernacular.

Third, they probably give an exact table of the equivalencies they found in the actual paper. It would be easier to check that if they have it available for free.

Four, they speak about how words sound, not how the look. So you have to take pronunciation of each language into account, which often can be quite different from what is written.

Yes, saying that "thou" is a word that has survived for hundreds of centuries is a bit odd to me. As far as I can tell, the only reason thou even exists as a word in modern English (at least in the US) is because it's so heavily used in the Bible (not exactly a modern book). The only times I hear someone using thou is in context of the Bible or ye olde tyme joking.

Bible wasn't originally written in English of course. The more recent the translation, the less Thou you get. We didn't have much alternative for 2nd person singular pronouns until "You" wiped "Thou" off the map somewhat recently and attempts at turning "Thou" into a formal religious version of "You" seem to have flopped. Then again we did pretty well without a second person plural until y'all invented y'all, plus or minus some "ye" anyway.

Theres some idioms that'll never die even if no one understands the history. "Holier than thou". Much like "Your name is Mudd" sometimes kinda horrifies people who don't know the story of Abe Lincoln and Dr Mudd. I've met people who claim to not know the history of "cotton pickin hands" or even acknowledge its racial background, which seems hard to believe, then again cotton hasn't been hand harvested in a couple generations locally so its possible the kids were telling the truth about not knowing what they're saying...

Well, "thou" also had a use for a period of time as an intimate form of "you." The intention behind the use of "Thou" to address God in the Bible is not to place distance between the speaker and God; on the contrary, it is to indicate closeness and intimacy.

The Wikipedia article on thou explains this briefly in the introductory paragraphs: https://en.wikipedia.org/wiki/Thou

Looking up "cotton pickin" in Google ngram viewer, it appears to be mostly a product of the 1960s-1970s ruralsploitation fad: The Beverly Hillbillies, Hee Haw and Foghorn Leghorn.

It's probably just an intentionally quaint pseudo-folk saying invented by someone in Burbank.

I don't think the Google ngram viewer is an appropriate tool for something that mostly belongs to the spoken language, and even more that of pour illiterate blacks and southerners.

"It's probably just an intentionally quaint pseudo-folk saying invented by someone in Burbank."

LOL no, assuming you young folks pretending not to understand are serious. As a hint take a look at the skin color and local social hierarchy position of the people picking cotton back when that was still done by hand, by property or former property, as they used to say. It was basically used in place of the "N-word" when it was just barely not socially acceptable to use the "N-word" in that situation.

Shakespeare, among others.

>Yes, saying that "thou" is a word that has survived for hundreds of centuries is a bit odd to me.

Well, for linguists "thou" and "you" is the same word.

Not just in that they mean the same, but in the same sense that in evolution X1 is said to be an ancestor species to X2.

"Thou" and "You", is the same word passed through a process of changes in pronunciation and spelling during the years. Linguists keep track of those kinds of things.

"Thou" is recently deceased, and while it hasn't been thoroughly documented, it's as likely as not due to a combination of the over-extension of the polite form of address and the use in early English typography of the "y" from Latinate typefaces for the thorn character. ("Ye" of "ye olde" is an alternate spelling of "the", and not related to the second person plural subject pronoun "ye".)

The polite plural is common enough in European languages; there are French dialects (in Canada, at least) where a good rule of thumb is "don't use tu unless you're sleeping with that person or are their parent". With the second person singular restricted to intimate use, it's not hard to imagine it falling away.

If you take this tree [1] as canon, English and Czech are fifth cousins, once-removed. If you compared me to my fifth cousin, I probably would look nothing like him, but I might carry similar traits or certainly similar ancestry.

Although the article doesn't convey it very well (and I am skeptical of the finding itself anyway), I believe what they are trying to say is that these words all 'have not changed' in that they all have derived from the same eventual ancestor. That is, the word's pronunciation may have morphed, but the words were never abandoned in favor of new ones. An example here would be the English 'four' vs. the Spanish 'cuatro'. If you believe in the reconstruction of the proto-language Proto-Indo-European (and Wiktionary [2]) then these both derive from the word * kʷetwóres[3].

The Spanish one is a bit easier to understand.

   * kʷetwóres (PIE) -> quattuor (Latin) -> cuatro (Spanish)

Nothing too shocking there, I hope. Latin dropped the "es" off of the end and changed the first vowel a bit, then Spanish simplified the "uo" diphthong into "o" (perhaps through Vulgar Latin? [4]), and reversed or->ro, and of course changed spelling because it's a different language.

English, however, took a little bit odder of a path. Somewhere before Proto-Germanic, * kʷetwóres->* petwṓr, which was an irregular change (perhaps to mimic five, which was * pénkʷe->* pémpe [5][6]). Then, a very regular rule, called Grimm's Law [7], kicked in in the first millennium BC, which shifted p sounds toward ɸ sounds ("For English-speakers, it is easiest to think of the sound as an f-sound made only with the lips, instead of the upper teeth and lower lip."[8]). So * petwṓr->* fedwōr (Note the t got vocalized into a d as well) in Proto-Germanic. West Germanic then lost the hard d in the middle, to make * fewōr, which became Old English fēower, into Middle English fower, into Modern English four. So:

    * kʷetwóres->* petwṓr->* fedwōr->* fewōr->fēower->fower->four

Both of these are sets of incremental changes from a common ancestor. Individually, the small changes make some sense (I hope), but once you get several thousand years away from the root, words like "cuatro" and "four" seem like they are completely unrelated by sound, even though they're very related by ancestry.

I'm no historical linguist, but hopefully that helps explain why these things sound nothing alike. :)

[1]http://upload.wikimedia.org/wikipedia/commons/4/4f/IndoEurop...

[2]http://en.wiktionary.org/wiki/Appendix:Proto-Indo-European/k...

[3] The * indicates that the word is a reconstruction, not actually something attested to in found writings.

[4] http://en.wikipedia.org/wiki/Vulgar_Latin#Monophthongization

[5] http://en.wiktionary.org/wiki/Appendix:Proto-Germanic/fedw%C...

[6] http://en.wiktionary.org/wiki/Appendix:Proto-Germanic/fimf

[7] http://en.wikipedia.org/wiki/Grimm%27s_law (Yes, one of the Brothers Grimm!)

[8] http://en.wikipedia.org/wiki/Voiceless_bilabial_fricative

Thanks - that was a good read!

But what about cuatro & four qualifies these as "ultraconserved"? Do these stand in contradiction to other words that arose without some gradual evolution?

If a word that evolves in a multitude of branches to arrive in a multitude of various forms is ultraconserved, what is a word that is just kinda conserved, or not conserved at all?

I personally chose 'four' to write about on a whim, because 'cuatro' sounded different enough, but was relatively explainable in a paragraph. But this only goes back to languages in the Indo-European tree.

The paper addressed here says that these 'ultraconserved' words share an ancestor before Indo-European. Via page two:

    In addition to Indo-European, the language families included Altaic (whose modern members include Turkish, Uzbek and Mongolian); Chukchi-Kamchatkan (languages of far northeastern Siberia); Dravidian (languages of south India); Inuit-Yupik (Arctic languages); Kartvelian (Georgian and three related languages) and Uralic (Finnish, Hungarian and a few others).

So, if you can imagine doing that process we did, but backwards, trying to take all these hundreds of languages, living and extinct, in the Indo-European family [1] and making one giant proto language that they all descended from, with a bunch of rules that determine how they morphed to become the child languages, then you get Proto-Indo-European. Now do that again for the other language families above. Then compare the resultant hypothetical languages, and see if they share enough between them to say that they are all children of the same language. This is a lot more controversial, see [2] (what this paper seems to subscribe to) and [3].

What the hypothesis seems to be here (this is speculatory; I haven't read the paper) is "If we look at these proto languages, there seem to be words that are similar enough that they probably are the last remnants of some ultra-proto language that all of these language families descended from".

And all words are coined and undergo evolution at some point, but often the popularity of words for the same thing can change, and not all words come from antiquity.

Inventions like 'piano' and 'internet' obviously haven't been around very long in the grand scheme of things. But we have other words like 'cater' and 'embrace' that are attributed to Shakespeare [4]. Some words that haven't been conserved: leafworm for caterpillar, wig for war [5], clepe for 'to call', except in the nearly-dead yclept [6], which also contains the ge/y prefix that makes past participles [7].

If you check out [8], you'll spot a bunch of words you probably know, but with senses they don't have anymore.

So words are always coming and going, and it seems that this paper is trying to say that certain prevalent words today statistically seem to be from a time before PIE. This reeks a bit of glottochronology [9] to me, though, which is fairly controversial.

[1] http://upload.wikimedia.org/wikipedia/commons/4/4f/IndoEurop...

[2] http://en.wikipedia.org/wiki/Eurasiatic_languages

[3] http://en.wikipedia.org/wiki/Proto-World

[4] http://www.pathguy.com/shakeswo.htm

[5] http://en.wikipedia.org/wiki/Changes_to_Old_English_vocabula...

[6] http://en.wiktionary.org/wiki/yclept

[7] http://en.wiktionary.org/wiki/y-

[8] http://en.wiktionary.org/wiki/Category:English_archaic_terms

[9] http://en.wikipedia.org/wiki/Glottochronology

The original publication is linked to from here: https://news.ycombinator.com/item?id=5667610

In fact, they calculated that words uttered at least 16 times per day by an average speaker had the greatest chance of being cognates in at least three language families. If chance had been the explanation, some rarely used words would have ended up on the list. But they didn’t.

I don't think that this logic is valid. It assumes that if the languages had had no common ancestry that any random word would be similarly likely to match between the languages. I'm no linguist but in my experience the most fundamental and commonly used words tend to be shorter, often composed of a single syllable. It also makes sense that the more fundamental words would be created early on and be given simple sounds while later words would be more complicated by necessity. If this is the case then you would expect more random coincidences within the cores of different languages than outside of them. This is also consistent with the fact that most of the examples in the article are very short: thou, give, hand, bark, spit, worm. I think that this would need to be taken into account in the statistical analysis before attempting to draw any conclusion.

I wonder what kind of metrics could be achieved simply by measuring how many mAhs it takes from google translate to translate from one language to another?

EDIT: Put simply, converting understandable sentences from one language to another is a lot more than just counting word frequences. Sorry having been mentioned an actual company name there!

I just wrote a lenghty text in Finnish and had it translated by an on-line service to English. Suprisingly, it wasn't the vocabulary but is was the structure that mattered. The machince got 99% of the words right, but it lost the meaning in 1/10 of the cases.

EDIT2: Not really sure who I am arguing here, on a stale thread, but I just have to finish the thought.

So it takes a measureable amount of work to analyze a sentence in a language into an abstract syntax tree.

An another amount of work to synthetize that back into an understandable sentence in another language.

The average sum of that work for all given sentences is the lingual distance of the two languages. No?

Of course all language analyzers and synthetizers are not equally good. What caught my eye, however, was how poor English a network translator was able to produce out of my Finnish. I'm pretty sure (althought the jury is still out on that) that the Russian it produced out of the same origin was remarkably more understandable. The main reason being the liberal word order.

I somehow assumed the english synthetizer would have been the top of the line. Turned out if wasn't. Or else the language model was not deep enough, which actually is a more plausible explanation.

Ok, thanks!

Naive review of the interactive infographic

thou

Altaic, Dravidian, Indo-European and Uralic sound similar enough

I can see how Chukchi-Kamchatkan and Kartvelian might end up where they are

And then we end up with Inuit-Yupik which sounds so dissimilar, the only thing I hear in common with the others is that it was apparently made by a human

to give

okay, I can kinda see maybe where Altaic, Inuit-Yupik and Uralic ended up where they are, but they're definitely different sounds

Dravidian and Indo-European are a bit more similar, but don't sound like they have anything in common with the others at all

hand okay, these are sorta similar, there's a clear m-vowel-n element in most of them but Indo-European gets all "weird"

bark

Altaic and Uralic may as well be slight accents

Inuit-Yupik diverges quite a bit

and Indo-European even more.

If I squint I can kind of see some similarity with a k-vowel then some other stuff so...okay

to spit these are the most similar to my ear, it probably doesn't hurt hat these appear to be onomatopoeia as well. Indo-European just seems to have reversed the bits somewhere along the way. If you squint "spit" in English would fit right in with these.

I always come away from these kinds of linguistic analyses scratching my head. Especially once you hear the words in the different languages.

It says Chinese wasn't included in the seven family of languages.

Yet, I, a english and cantonese speaker, couldn't help but notice these two words which sound similar in both languages.

Mother is "ma" in English, and also "ma" in Cantonese.

Father is "pa" or "dad" in English, and "ba" or "deh" in Cantonese.

Remarkably similar, considering both languages are considered by linguists to be practically completely unrelated.

The words are similar because these are some of the first sounds babies learn to make.

Interesting. Is there some objective way of determining if the words are cognates or not? In particular, in the audio section, the Inuit-Yupik example for "thou" sounds very out of place.

I would like someone to answer as well. See I only know English, Polish, and German pretty well. So take father, ojciec, and vater. I don't see it. Thinking about it the languages I know, I would expect something like ta da pa (or backward) at, ad, ap. Would they say the ojci it sort of like a at sound and close enough?

Edit: Wait a minute, though matka and ojciec are the words for mother and father in Polish, little kids might say more mama and tata. I might have answered my own question here, but I too noticed some in the linked web page did not sound particularly similar to me.

Words are cognate if they stem from a common root. How can we tell? If we have two sets of words S and T, then we need to postulate a third set R of 'proto-words' as well as two mappings R → S and R → T where the mappings can (within reasonable margins of error) predict the form of the words given the roots.

For example, take German and Latin: I assert the words Vater and pater (both 'father') are cognate, but to do so, I need to examine a whole set of corresponding words in both languages, so I take the sets {Vater, Fisch, Fuß} and {pater, piscis, pes}. From these, I can see clear correspondences (e.g. a German f sound always corresponds to a Latin p sound[1], the Latin sc seems to correspond to the German sch) and some things that are unchanged.

Next, I come up with a set of roots like {pater, piš, pes} and a set of rules to apply like {p → f in German, š → sc in Latin, š → sch in German}, and I can state with some confidence that these words descend from a common root.

In practice, you'd do this with way* more than three words—preferably you'd use most of the language—and there is some wiggle room for related meanings or exceptional changes, e.g. the German word alt and English word old seem unrelated to the Latin word senere 'to be old', because they actually stem from words for 'to grow up', and are therefore cognate to the Latin altus 'high, tall'. Additionally, some rules are more plausible than others based on research and phonological considerations, so the change of s → h is well-attested in actual languages and is consistent with similarities between the two sounds, while the change of r → b is highly unlikely.

And while there is some room for error, there is a limit e.g. one might argue that English and Japanese are cognate because in Japanese, 'to eat' is taberu and 'to see' is miru, and in English we eat off a table and see in a mirror. This doesn't get borne out with more thorough phonological rules and doesn't seem to reflect regular semantic processes, either.

tl;dr: Two words are cognate if they come from a common root; without direct knowledge of the parent language (e.g. Latin for Italian and French), we postulate one and show that regular processes can derive the child languages' vocabulary.

[1]: If you don't speak German, know that the German v is pronounced like an f.

That was an excellent explanation, thank you. Also regarding ojciec, I found this: http://en.wiktionary.org/wiki/ojciec#Etymology The changing over time following rules was the missing part for me.

That's because ojciec is not a cognate of father (neither is tata).

Take a look at:

http://starling.rinet.ru/cgi-bin/response.cgi?root=%2Fusr%2F...

Google translate does an ok job of the Russian if it ends up looking too confusing.

That is hard for me to make sense of, it is saying that ojciec comes from PIE átta, meaning something more like male guardian and becoming thus a synonym of father? And that father and vater come from PIE ph₂tḗr which means more directly father?

Yeah, the extra bit of detail theorized there is that atta came from children's vocabulary and the derived word, for whatever reason, replaced the PIE pater-oids. This change is carried forward in all Slavic languages.

But whatever the details - the reason ojciec and vater don't sound related to you is because they aren't, they don't share a root.

This does pop up even for very basic words in related languages with well-established common origins. Water, Wasser and woda sound related - the Latin aqua, not so much. It's not hard to imagine a relationship between fire, Feuer and, say, pyros (like in pyromaniac) but not so much with ogień. Then again, ogień, 'ignite' and 'agni yoga'. The reason for these seems to be that PIE had two different roots each for fire and water, respectively.

Sometimes the common root lives on in fairly different words - Nebel, nebula, nebo share a root. More unexpectedly, so do guest and hostile. Or German Gift and English gift.

And then you have the real winners like 'kobieta' the origin of which is essentially an unsolved mystery.

Thanks again, I just ran a across this, none is very convincing: http://tech.groups.yahoo.com/group/cybalist/message/17417

Another negative response from a linguist http://languagelog.ldc.upenn.edu/nll/?p=4612

Why couldn't the NYT have transitioned as eloquently as the Washington Post onto the web?

So, suppose we want to know what 'bark' sounded like in the common language 15,000 years ago. Call that common language X.

Well, take languages A and B now where 'bark' is shared as a cognate.

Now, start with language A now and realize that as we go back in time 15,000 years ago to language X how 'bark' was pronounced changed year by year. Then realize that as we go forward in time from language X 15,000 years ago to language B now, how 'bark' was pronounced has changed year by year.

Then the change in how 'bark' is pronounced in languages now is smaller than either the change from language A now to language X 15,000 years ago or from language B now to language X 15,000 years ago.

So, how close is how 'bark' is pronounced now to how it was pronounced 15,000 years ago? Closer than to how 'bark' is pronounced between A and B now.

So, if 'bark' is pronounced nearly the same in languages A and B now, then it was pronounced nearly the same 15,000 years ago. Thus we have an approximation to how 'bark' sounded 15,000 years ago.

It may be that languages A and B have a common ancestor more recently than 15,000 years ago. So, for a better approximation want to pick for languages A and B languages for which we have some hope that their most recent common ancestor was X 15,000 years ago.

How to do this? For candidate pairs A and B, look at geographical and genetic distances and pick the most distant pair.

Individual words often migrate between languages even if they don't share a common ancestor.

Cute. And there are other issues with what I wrote. For a more careful analysis, would have to do, say, a 'probabilistic' analysis, e.g., where in principle can walk back to the beginning and start over but actually, e.g., with some common assumptions, in a space of dimension 3 or greater, probabilistically keep getting farther from the starting point.

Your point should come out in the wash in estimating where there is a common language and how far back if use more than one word, etc.

My guess, especially from many of the other comments on this thread, is that the basic data is so bad, noisy, etc. and pronunciation changes so large that the math I outlined, while basically correct, would mostly give such large distances that we wouldn't learn much.

Maybe a better solution would be just to assume that the tree of languages would follow the tree of genetic inheritance and, then, via genetic analysis trace that tree.

But the article wanted to ask what one of those 'common' words would 'sound' like 15,000 years ago, so I gave a way to get an approximation. The cute part is while 'father' in English sounds a bit different from 'vater' in German, the common ancestor should sound closer to either 'father' or 'vater' than those to two each other, that is, closer rather than farther away. That is, our intuition would be that the sound from 15,000 years ago should be much different from either the current 'father' or 'vater', but with the math model I gave that should be wrong.

“I was really delighted to see ‘to give’ there,” Pagel said. “Human society is characterized by a degree of cooperation and reciprocity that you simply don’t see in any other animal. Verbs tend to change fairly quickly, but that one hasn’t.”

This is the sort of unscientific wishfulness that undermines the analysis.

And is directly contradicted by known animal behavior.

Your way of living is not a happy one.

It's unfortunate that you believe happiness requires self-delusion.

And you just know more than the experts. Armchair expert...

Reply on news.ycombinator.com