The Indo-Austronesian Hypothesis
ATTENTION: Some of the opinions stated in the following text are contrary to the mainstream science. I will not advise you to read it if you don't have a substantial background in linguistics. I am not a conspiracy theorist who wants to bombard people with controversial statements they don't know how to evaluate, and I am not denying it is possible my work is to historical linguistics what Anatoly Fomenko's work is to history. If you are ready to read it, click here.
Here are the words from the Swadesh list which I consider to be
potential cognates between Proto-Indo-European and
Proto-Austronesian:
*treys (three)-*telu (three)
*ronk
(hand)-*lima (hand/five)
*ser (to flow)-*qalur (to flow)
*skend (skin)-*qanic (skin)
*stembh (to walk)-*qaqay (foot)
*smew (smoke)-*qabu (ash)
*serw
(to watch)-*qalayaw (day)
*sal (salt)-*qasira (salt)
*bhas (to talk)-*baqbaq (mouth)
*bhewgh (bow)-*busur (bow)
*bher
(to give birth)-*bay (woman)
*dwoh1 (two)-*dusa (two)
*dyews (sky)-*daya (upwards/height/sky)
*danu (river)-*danaw (lake)
*dngjhuh2 (tongue)-*dilaq (tongue)
*delh1 (wide)-*dempad (wide)
*egjoh2 (I)-*aku (I)
*men (to think)-*nemnen (to think)
*h1nomn (name) could perhaps be a cognate to
Proto-Polynesian *hingoa (also meaning "name", but that root isn't
attested outside of Polynesian)
The regular sound correspondences are somewhat hard to establish. Ones
that appear obvious are s:q, r:l, bh:b and d:d.
People might think that there are not any actual correspondences and
that what I find are just coincidences. But think of it this way: I
have found a simple sound law (that PIE *s corresponds to PAN *q), and
there are seven examples of that on the Swadesh list. What’s the
probability of that if those languages are not actually related? We
are mostly dealing with 2-consonantal roots here. Let's be generous
and say I allowed myself the semantic drift of 3 words. Both
proto-languages have about 20 consonants. So, if the word in PIE has
an *s, the probability that I find a word in PAN in which PIE *s
corresponds to PAN *q is 1-(1-(1-(1/20))^3)^2=26%. Swadesh list has
100 words, so we can expect that 100/20=5 words where the first
consonant in PAN is *q, and 5 words where the second consonant is *q,
so that there are 10 words where PIE *s can potentially correspond to
PAN *q. The probability that the rule coincidentally works in 7/10
words is 1-((1-0.26^(7-1))^10)=0.31%. The probability of finding such
a pattern in two truly unrelated languages is 1-(1-0.0031)^20=6%.
That's pretty low.
UPDATE on 05/04/2025: You know what, I withdraw the statements I made above. I think that it is committing the fallacy of doing calculations making wildly unrealistic linguistic assumptions. For example, it is assuming the collision entropy of a consonant in an average language is around log2(20) bits per symbol. It is not, it is far from it. The collision entropy of a single consonant in the English language, as can be measured using the methods I described elsewhere, around log2(11) bits per symbol. Just by taking that into account, the probability of finding a single word in which Proto-Indo-European *s corresponds to Proto-Austronesian *q increases from 26% to 93.82%. And the probability of finding 7 such words on the Swadesh List increases from 0.31% to 99%. The p-value of the pattern that PIE *s corresponds to PAN *q is almost 100%. However, I still think that my alternative interpretation of the Croatian toponyms might be right, because the p-value of the k-r pattern in the Croatian river names seems to indeed be between 1/300 and 1/17, as the computer simulations adjusted for real-world data suggest that.