So, is this an example of a pseudoscientific language comparison:


In short, the webpage on the link claims that there are a few regular sound correspondences between Proto-Indo-European and Proto-Austronesian, such as that PIE *s corresponds to PAN *q and that PIE *r corresponds to PAN *l, and that this alone is enough to tell that those languages are related with reasonable certainty.

  • 4
    The basic operation of compiling word lists and showing sound correspondences is not an invalid one. But the probability calculations standing in for analysis, the lack of depth (e.g. discussion of background phonological theory or proposed derivations), and of course its being on a personal website all raise flags. It seems to me that it's not so much "pseudoscientific" as it is a passing speculation that would require vastly more work to substantiate and make interesting. P.S. It would be a stronger question if you added your thought process: What sort of criteria do you have in mind? Dec 23, 2017 at 6:20
  • 5
    Please sketch the Indo-Austronesian hypothesis in your question, links to extermal websites have a tendency to go stale over time, rendering the question in a less than useless state. Dec 23, 2017 at 10:47
  • And that particular one is "sleeping".
    – jlawler
    Dec 23, 2017 at 22:05

1 Answer 1


I wouldn't say it's pseudoscience, no. I have great confidence that it's wrong, but wrong science is still science.

The Indo-Austronesian Hypothesis is exactly what it says on the tin: a hypothesis. The author has proposed the idea that the Indo-European and Austronesian languages are related, which isn't inherently absurd: weirder connections have happened before.

Then, the author's gone through and compiled a parallel word list of maybe-cognates, and come up with a set of regular sound changes that would connect them. This is definitely the right next step for someone who wants to show that language families are related: it's the basis of the Comparative Method, basically the gold standard in linguistic reconstruction.

And then…that's where the author leaves off. They do some sketchy probability calculations but don't claim those calculations are undeniable proof or anything. And math mistakes don't inherently make something pseudoscience.

So I'd say that it's perfectly valid science—it's just not finished (and has some math mistakes). It's a hypothesis without enough evidence behind it to make it a theory. To become an actual theory, it would have to be developed further, with more cognates found, grammatical comparisons, and so on. And the author admits that on the first page: "Maybe some expert will read this and pursue my ideas." So far, as far as I know, no expert has.

TL;DR: It's a perfectly valid (scientific, falsifiable) hypothesis, with a small amount of evidence behind it. It's just not complete. I also believe that it's downright wrong, given my own experience with Indo-European and Austronesian languages, but I'd be happy to be proven wrong on that if the author comes up with a stronger argument.

  • No, lexical/phonetic correspondance alone is not deemed particularly strong, rather typological and morpho-semantic similarities are preferred to establish bona fide "genetic relation", because these are less likely to be loaned. On top of that, corelated geo-historic evidence should seal the deal. Other kinds of relation are of course no less difficult to substantiate. In terms of programming, there's an entry function, a jump table with gotos, a hand rolled permutation generator that produced gibberish for most if not all but the tested inputs and it doesn't link cause ... :/
    – vectory
    Jun 17, 2019 at 16:38
  • Why would anyone necro this 2y old thread that @flatassembler obviously asked about their own research?
    – vectory
    Jun 17, 2019 at 16:39
  • 1
    @vectory Mostly because I go through the "Unanswered" list periodically to pick off the easiest ones to answer :P and because it might be useful for someone in the future. I agree that you'd need grammatical similarities to actually make a strong case for the relation, but coming up with a set of sound changes is a better start than most of the "Proto-World" people make.
    – Draconis
    Jun 17, 2019 at 16:51
  • well, with greater time depths the expectable conjecture is going to be weak, and the evidence required to be particularlarly strong. Having looked at Ruhlens statiscal selection from mostly Nostratic dictionaries, the assorted lists, e.g. for Co(n)gV "knee, bend" are overwhelming, but not secure nor free from criticism.
    – vectory
    Jun 18, 2019 at 3:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.