Page 8 of 21

Re: Soft Sciences Vs. Hard Sciences

Posted: Fri Mar 08, 2019 6:55 am
by teo123
brimstoneSalad wrote:Perhaps you should not: there's a bias in that people have misplaced confidence in things they think they understand better.
How is that a bias? It's just basic logic, you can't justify a confidence in something if you don't understand how it can be proven or disproved.
brimstoneSalad wrote:You mean unobservable due to redshift or what?
Well, yes. If other galaxies weren't visible with normal telescopes, the Big Bang Theory wouldn't have been postulated.
brimstoneSalad wrote:Not sure what your point is.
I meant to say that saying "Things in languages are right or wrong only because people who speak those languages agree on that." is wrong for the same reason saying "Money has value because people agree it has value." is wrong.
brimstoneSalad wrote:I already talked about that xkcd comic that showed the principle...
But it doesn't matter if they made new observations or have simply done new statistical analyses on previous observations. Had they made 20 new experiments, each with the p-value of 5%, chances are, they would have probably also "found" a type of jellybean that caused acne. Doing the same experiment again and again is also a form of p-hacking.
brimstoneSalad wrote:What are the odds of taking ten red samples from an infinite pool of 50%-50% red and green?
Actually not that low.
How is it not low, it's less than 1/1000?
brimstoneSalad wrote:If there are even just 75% red apples, you get .75^10 which is almost a 6% chance.
So, we can say "At least 75% of all apples are red." with 94% certainty, right?
With the Havlik's Law, the calculation is not that simple. What's the probability that a word in a modern Slavic language appears to follow the Havlik's Law if it doesn't actually do that? Well, almost every word in Proto-Slavic had a yer in it (almost every noun ended with a yer, which disappeared in modern Slavic languages), and we can simplify that every yer was either vocalized or unvocalized in some modern Slavic language (let's ignore the clusters yer+r and yer+l, which obeyed different laws in quite a few languages, and let's ignore that, had the Havlik's Law been invalid, we wouldn't expect all the vocalized yers to turn into one sound in each modern Slavic language). So, we can say that the probability of a word appearing to follow the Havlik's Law when it doesn't actually do that is 0.5^number_of_ProtoSlavic_yers_in_it. Let's say there are 500 words in each Slavic language (there are probably more) to test for the Havlik's Law. I don't know right now what number should be put as the average number of yers in a Proto-Slavic word, but even if we put only 1.0, the p-value is still insanely low. Even if we try to calculate the p-value for the Havlik's Law being apparently accurate for 300/500 words (using the binomial coefficients), it's so low the calculator can't calculate it. The p-value gets to 5% already at 268/500 words, so if the Havlik's law was apparently only 53% accurate, and I guarantee you it's more accurate than that.
brimstoneSalad wrote:That sample is likely not very random at all.
Yes, but that's only because we know a lot about apples. If we didn't, it would be scientific to assume the sample is indeed random, right?

Re: Soft Sciences Vs. Hard Sciences

Posted: Fri Mar 08, 2019 7:12 am
by teo123
Sorry, @brimstoneSalad, I haven't seen your last post until now. OK, what you are trying to say is that, because of the small sample size, there is a big difference between, for example, 9/10 apples being red and 10/10 apples being red, right? Well, we are not really dealing with such small sample sizes.

Re: Soft Sciences Vs. Hard Sciences

Posted: Sat Mar 09, 2019 4:05 am
by brimstoneSalad
teo123 wrote: Fri Mar 08, 2019 6:55 am
brimstoneSalad wrote:Perhaps you should not: there's a bias in that people have misplaced confidence in things they think they understand better.
How is that a bias? It's just basic logic, you can't justify a confidence in something if you don't understand how it can be proven or disproved.
There's a difference between thinking you understand something and actually understanding it.
teo123 wrote: Fri Mar 08, 2019 6:55 am Well, yes. If other galaxies weren't visible with normal telescopes, the Big Bang Theory wouldn't have been postulated.
Well red shift is what gave us the idea, yeah, but there are potentially other ways to arrive at that model.
However, if none of that were available we simply wouldn't have the theory. That's a lot different from speculating on things that you may never be able to prove. That's not speculating on things because you can't prove it... kind of the opposite.
teo123 wrote: Fri Mar 08, 2019 6:55 amI meant to say that saying "Things in languages are right or wrong only because people who speak those languages agree on that." is wrong for the same reason saying "Money has value because people agree it has value." is wrong.
...Contemporary language functions by a broad enough consensus to arrive at understanding.
Now you can look at dead languages and try to decode them, but any attempt to do so is only going to provide one possible model unless you have experimental confirmation (which again, is rare).
teo123 wrote: Fri Mar 08, 2019 6:55 amBut it doesn't matter if they made new observations or have simply done new statistical analyses on previous observations. Had they made 20 new experiments, each with the p-value of 5%, chances are, they would have probably also "found" a type of jellybean that caused acne. Doing the same experiment again and again is also a form of p-hacking.
Yes, that is also bad.
But at least in the latter case you can repeat it with more data and realize it's wrong.
teo123 wrote: Fri Mar 08, 2019 6:55 am With the Havlik's Law, the calculation is not that simple. What's the probability that a word in a modern Slavic language appears to follow the Havlik's Law if it doesn't actually do that?
You wouldn't break it down by words, you'd do it by entire languages. Whether a language obeys or not.
Breaking it down by words is like sampling ONE apple, and breaking the skin up into cells and checking the color of each and finding them all red and so arriving at an amazing p value.

You have ten new languages Havlik didn't know about, how many obey Havlik's law?
That's what you have to do.
teo123 wrote: Fri Mar 08, 2019 7:12 am Sorry, @brimstoneSalad, I haven't seen your last post until now. OK, what you are trying to say is that, because of the small sample size, there is a big difference between, for example, 9/10 apples being red and 10/10 apples being red, right? Well, we are not really dealing with such small sample sizes.
No, I'm saying you can always find some wacky correlation in an established sample when you're working backwards.
You can find something that's true of 10/10 of the item you sampled, and yet in reality is true in only half or less.

You can't take a sample and look at what they have in common (any and all features) and draw conclusions from that.
You have to make a hypothesis, then take a random sample and look at only that feature you hypothesized about.

There's a huge difference between:
1. Collecting ten apples before coming up with a hypothesis, seeing that they're all red, and then saying therefore most apples are red
and
2. Hypothesizing that most or all apples are red, then collecting ten random apples, seeing they're all red (analyzing ONLY redness and drawing *no other* conclusions), and concluding that most or all apples are red.

The second is legitimate.
Yes you *could* do a million experiments and still find some things true that aren't (like the xkcd comic) but that's a lot harder than just forming an ad hoc hypothesis based on an extant sample you analyzed.

Re: Soft Sciences Vs. Hard Sciences

Posted: Sun Mar 10, 2019 6:17 am
by teo123
brimstoneSalad wrote: There's a difference between thinking you understand something and actually understanding it.
Yes, but if you don't think you understand something, then you, by definition, don't understand it.
brimstoneSalad wrote:Well red shift is what gave us the idea, yeah, but there are potentially other ways to arrive at that model.
Then that model would predict the redshift which wouldn't be observable. Again, you have information needed to confirm the theory lost to time.
brimstoneSalad wrote:Contemporary language functions by a broad enough consensus to arrive at understanding.
Consensus is formed because something is in a language, it's not that something is in a language because of the consensus. What's in a language is determined by linguistic laws. Words that are "formed" by juxtaposing random sounds together, as well as ungrammatical phrases, don't get accepted. There being words and phrases whose origin is not clear to modern speakers of some language, or which seem ungrammatical, is simply a result of a language being conservative.
The same goes for, for example, prices. If some price is too far from what the economic laws determine, nobody will agree on that price.
brimstoneSalad wrote:You have ten new languages Havlik didn't know about, how many obey Havlik's law?
I agree that would be interesting to know, but, unfortunately, there are only finitely many languages in the world, especially when we are only talking about languages that have descended from Proto-Slavic. When discussing methodology, we need to set realistic goals. And that's what scientific methodology and logic are there for, to be able to get reasonably certain conclusions from limited data, right?
Besides, even if we indeed find an undocumented Slavic language that doesn't obey the Havlik's law, what does that have to do with how good description the Havlik's law is of the sound changes that have happened in Croatian, Polish, Russian, Czech, Slovenian, Bulgarian, Ukrainian..., and all the dialects of those languages except partly the Chakavian dialect of Croatian? If we indeed find an undocumented Slavic language in which, for example, all the yers have remained, the right inference from that would be that it split off from Proto-Slavic before the sound changes described by the Havlik's law started to operate. That would bring into question the validity of the notion that the sound changes described by the Havlik's law are bound to happen whenever the phonotactics of a language is sufficiently restrictive and some vowels occur way more frequently than others (as was the case in Proto-Slavic), but that wouldn't mean the Havlik's law itself is wrong.
The way to show the Havlik's law is incorrect is to show that the data Havlik cited is wrong or that the Havlik's law doesn't indeed explain the data it's supposed to explain.
brimstoneSalad wrote:No, I'm saying you can always find some wacky correlation in an established sample when you're working backwards.
There are patterns you can expect to occur in random data, and there are those you cannot expect to occur in random data. That's the core premise of historical linguistic.
You should expect quite a few words to sound similar in unrelated languages (Japanese "namae" and English "name").
But you shouldn't expect there to appear to be phonetic laws describing the relation between unrelated languages, such as the phonetic law that says that Latin 's' corresponds to Greek 'h' at the beginning of a word: sex-hex (both meaning "six"), septem-hepta (both meaning "seven"), sal-hals (both meaning "salt"), sol-helios (both meaning "sun"), sus-hys (both meaning "pig"), super-hyper (both meaning, among other things, "above"), similis-homoios (both meaning "similar"), sollus-holos (both meaning, among other things, "whole"), silva-hyle (both meaning, among other things, "forest")...
I'd encourage you to read this:
http://www.zompist.com/chance.htm

Re: Soft Sciences Vs. Hard Sciences

Posted: Wed Mar 13, 2019 4:36 am
by brimstoneSalad
teo123 wrote: Sun Mar 10, 2019 6:17 am
brimstoneSalad wrote: There's a difference between thinking you understand something and actually understanding it.
Yes, but if you don't think you understand something, then you, by definition, don't understand it.
My point is that not understanding something or finding something hard to understand is not a good reason to deny consensus on it or be particularly skeptical about it. That's an argument from ignorance.
You should accept particle physics AND the shape of the Earth based on professional consensus. If you personally think you understand the facts on why the Earth is an oblate spheroid better then that's great, but that's kind of immaterial to why you should accept it.
Thinking you understand how something works shouldn't really do much to increase your confidence in the fact of it, at least at this level (again, remember your misplaced confidence in flat earth?).
You shouldn't really be any more confident that your plane won't fall out of the sky once you understand the dynamics of lift and can calculate it on paper vs. not.
I understand of course how in terms of human psychology you may *feel* more confident in something you understand, but my point is that perhaps you should not be.
teo123 wrote: Sun Mar 10, 2019 6:17 am Then that model would predict the redshift which wouldn't be observable. Again, you have information needed to confirm the theory lost to time.
Again, like I said, an area of physics that was speculating without evidence on something unobservable would be soft.
We talked about string theory.

It happens that the Big Bang is hard science because it's observable.
teo123 wrote: Sun Mar 10, 2019 6:17 amConsensus is formed because something is in a language, it's not that something is in a language because of the consensus.
Chicken and egg. As languages form and drift, they follow a changing pattern from consensus. Eventually the laggards give way to popular usage.
teo123 wrote: Sun Mar 10, 2019 6:17 amWhat's in a language is determined by linguistic laws.
What's in a language is determined by human beings, as a group, liking or not liking something, or finding it useful or not.
Words are coined, borrowed, spread, die out, and so do arbitrary grammar rules.
teo123 wrote: Sun Mar 10, 2019 6:17 amWords that are "formed" by juxtaposing random sounds together, as well as ungrammatical phrases, don't get accepted.
They can be, but due to human physiology some things sound better or more appropriate than others, and it has to do with what we're accustomed to hearing.
Get somebody from an atonal language to listen to a tonal language and that person may be incapable of picking up the tones. It's kind of a firmware issue. Same for phonemes they don't use, they may hear it as something else.

Perception of sound is very much reliant on certain physiological filters.

The Laurel or Yanny thing was a great example of the subjectivity of perception of language for a lot of people.

This is a matter of neuroscience, a hard science. And if you're interested in the overlap between that and linguistics that's cool.

But again, we're talking about humans deciding things. The fact that humans may be said not to have free will in what they decide because they're only capable of processing certain things efficiently? Well, that's interesting, but doesn't really change my point. There's nothing stopping a very foreign phoneme from surviving in a language if people wanted it to. We get that a little bit with certain religious words where people find preservation more important. With enough motivation people can overcome the cognitive inconvenience of dealing with weird sounds.
teo123 wrote: Sun Mar 10, 2019 6:17 am
brimstoneSalad wrote:You have ten new languages Havlik didn't know about, how many obey Havlik's law?
I agree that would be interesting to know, but, unfortunately, there are only finitely many languages in the world, especially when we are only talking about languages that have descended from Proto-Slavic.
And THAT is why it's a soft science.
If physics only had that few samples in something, it too would be remarkably soft. This is hard to overcome.
teo123 wrote: Sun Mar 10, 2019 6:17 amWhen discussing methodology, we need to set realistic goals.
:lol:
No, you don't get to say 'lets be fair and let everybody be a hard science with different standards, because it's not realistic for these other fields to attain the same level or rigor'.
That's the like alt-med people trying to get Wikipedia to loosen its standards for them.

Linguistics doesn't get a handicap. It gets to be soft, that's it, and it has to deal with it or find new ways to innovate and harden itself. Like crossing itself with neuroscience to gain more data about low level linguistic cognition and the hardware involved.
teo123 wrote: Sun Mar 10, 2019 6:17 amAnd that's what scientific methodology and logic are there for, to be able to get reasonably certain conclusions from limited data, right?
Not to cheat, no. Sometimes there's just not enough data to have much certainty at all. Sometimes there's so little that we can't even quantify our certainty, and that's unfortunate.

But again, like I said, I err on the side of expert opinion. I will *assume* Havlik's law is correct, because I have no reason to assume otherwise and no alternative.
However, that assumption needs to be tempered with the degree of certainty that it deserves based on the rigor of the field it's coming from... which is a lower degree of certainty.

Regardless of how well or poorly I understand particle physics or linguistics, I have a high degree of certainty in the former and a lower in the latter by virtue of the nature of those fields.
teo123 wrote: Sun Mar 10, 2019 6:17 am Besides, even if we indeed find an undocumented Slavic language that doesn't obey the Havlik's law, what does that have to do with how good description the Havlik's law is of the sound changes that have happened in Croatian, Polish, Russian, Czech, Slovenian, Bulgarian, Ukrainian..., and all the dialects of those languages except partly the Chakavian dialect of Croatian? If we indeed find an undocumented Slavic language in which, for example, all the yers have remained, the right inference from that would be that it split off from Proto-Slavic before the sound changes described by the Havlik's law started to operate.
The *right* inference would be to retain the original hypothesis at all cost?
*cough* ad hoc hypothesis *cough*

This is what I mean. What discovery *could* convince you it's wrong? And that the apparent adherence of some languages is just a coincidence?
teo123 wrote: Sun Mar 10, 2019 6:17 am The way to show the Havlik's law is incorrect is to show that the data Havlik cited is wrong or that the Havlik's law doesn't indeed explain the data it's supposed to explain.
Then it's not a law in any scientific sense, but a simple descriptive shorthand for how a small collection of languages happen to currently sound or happened to sound.

Like if you made teo's law to describe how 9/10 of *these* ten apples are red. You'd never really be able to falsify it, because it doesn't have any implications beyond those ten apples. It also wouldn't be very useful.

teo123 wrote: Sun Mar 10, 2019 6:17 amThere are patterns you can expect to occur in random data, and there are those you cannot expect to occur in random data. That's the core premise of historical linguistic.
There are patterns that are more or less likely in random data. Very large correlations are something to look at and suggest relatedness (the same is true in DNA analysis), but if you can not test things like *laws* that you say languages obey then those claims are ad hoc.

There's a difference between looking at correlations and getting a probability of some kind of common origin (which is not a bold claim at all), and looking at how sounds happened to evolve in a few cases and presuming that to be a general law.

Re: Soft Sciences Vs. Hard Sciences

Posted: Wed Mar 13, 2019 3:11 pm
by teo123
brimstoneSalad wrote:My point is that not understanding something or finding something hard to understand is not a good reason to deny consensus on it or be particularly skeptical about it. That's an argument from ignorance.
Do you think that saying "I am not confident that's true because I don't understand how it can be falsified." is also an argument from ignorance? For instance, I am currently taking a physics course on the university (studying computer science), and the things we are told about particle physics still sound like baseless assertions to me, just like they did in primary school and in high school. I only have a very dim idea how those things could be tested. So, how can I be confident there is much truth to particle physics?
brimstoneSalad wrote:As languages form and drift, they follow a changing pattern from consensus.
Languages change primarily because of the linguistic laws such as the Grimm's law, the Verner's law and the Havlik's law, and not because people want them to change. Sound changes happen no matter how badly people want to stop them, the speakers are not even aware of them as they happen, as long as the language is living.
brimstoneSalad wrote:The Laurel or Yanny thing was a great example of the subjectivity of perception of language for a lot of people.
Yeah, but I don't think that says much about how languages actually work. It's more of an interesting coincidence, that people hear different words in some noise depending on which frequencies of that noise is better heard. Whenever I listen to it, I hear the Croatian word "jedi" (to eat).
brimstoneSalad wrote:This is a matter of neuroscience, a hard science. And if you're interested in the overlap between that and linguistics that's cool.
Well, not really. I've always assumed that what neuroscience has to say today has very little relevance to how languages actually work. The data from the actual languages has much more value than speculations based on neuroscience. For instance, the supporters of the phonosemantic hypotheses often cite the studies from psycholinguistics that seemingly support their "theories", and ignore the fact that studies on the actual languages don't confirm their "findings" (and they also refuse to answer the obvious question of how can, for example, the Grimm's law be true if the phonosemantic hypotheses are true).
Perhaps computer science has a lot more to say about how languages work, but that can also be brought into question. What if the parallels between programming languages and natural languages are only superficial? What if the syntaxes of natural languages don't actually function by forming AST-es? There are some good reasons to think they don't, AST-es appear to be unable to deal with the phenomena such as pseudocoordination. The speech recognition software is useful, but does it actually model the way humans process sounds in a language? There are some good reasons to think otherwise, the errors the speech recognition software make don't at all resemble the types of errors human beings make, and this is especially true for vowel sounds.
Regardless, once you look at the statements from historical linguistics such as that English 't' corresponds to German 'z' at the beginning of a word, all of that sounds like a baseless speculation. The claim that English 't' corresponds to German 'z' in the beginning of a word is so precise that it sounds like it can't be true, yet it's so demonstrably true, it's obvious how to test it, and it proves itself again and again.
brimstoneSalad wrote:I will *assume* Havlik's law is correct, because I have no reason to assume otherwise and no alternative.
What do you actually think that the Havlik's law says? You shouldn't assume your understanding of it is right, since it doesn't really appear to be.
brimstoneSalad wrote:And that the apparent adherence of some languages is just a coincidence?
Is that even a reasonable question to ask? Like I've explained before, the probability that a Slavic language appears to follow the Havlik's law if it doesn't actually do that is so low no mathematical program I've tried to calculate that probability with (using the binomial coefficients) gives me a result. The probability of only 53% of the words in a Slavic language to which the Havlik's law potentially applies appearing to follow the Havlik's law if they don't actually do that is 5%.
Finding some 300 words in older texts in a language to which the Havlik's law potentially applies but doesn't actually do that would perhaps make it reasonable to doubt that the Havlik's law is even approximately valid, but that's very unlikely to happen.
brimstoneSalad wrote:You'd never really be able to falsify it, because it doesn't have any implications beyond those ten apples.
Then we must have very different definitions of falsifiable. Of course that a claim such as "All the apples that are currently in the Eiffel's Tower in Paris are red." is falsifiable. The claims such as "The English 't' at the beginning of a word corresponds to German 'z'." or "The Latin 's' corresponds to Greek 'h' at the beginning of a word." are even more so, you can test them here and now using the data from Wiktionary or something like that.
brimstoneSalad wrote:It also wouldn't be very useful.
Well, the claim that Latin 's' corresponds to Greek 'h' at the beginning of a word certainly is useful, like when you are trying to remember Latin and Greek words, or when you are trying to guess the meaning of an unknown word.
brimstoneSalad wrote:There's a difference between looking at correlations and getting a probability of some kind of common origin (which is not a bold claim at all), and looking at how sounds happened to evolve in a few cases and presuming that to be a general law.
And yet, as the web-page I linked you to in my last post explains, if you try to propose a common origin of a some languages without proposing the phonetic laws that operated in them since they split (such as one that the English 't' corresponds to German 'z' at the beginning of a word), you are probably practicing pseudoscience. Mere phonetic similarity should play no role in historical linguistics (because that's also not how languages work), what should play a role are phonetic laws that are easy to test.

Re: Soft Sciences Vs. Hard Sciences

Posted: Thu Mar 14, 2019 4:24 am
by brimstoneSalad
teo123 wrote: Wed Mar 13, 2019 3:11 pm
brimstoneSalad wrote:My point is that not understanding something or finding something hard to understand is not a good reason to deny consensus on it or be particularly skeptical about it. That's an argument from ignorance.
Do you think that saying "I am not confident that's true because I don't understand how it can be falsified." is also an argument from ignorance?
Its getting there.
teo123 wrote: Wed Mar 13, 2019 3:11 pmI only have a very dim idea how those things could be tested. So, how can I be confident there is much truth to particle physics?
Based on consensus behind them in the field, and the hardness of the field generally.
teo123 wrote: Wed Mar 13, 2019 3:11 pmSound changes happen no matter how badly people want to stop them, the speakers are not even aware of them as they happen, as long as the language is living.
That's quite fatalistic, and it's a very bold claim about the power of these forces on language.
Where do these forces come from? Essentially human behavior. So right off the bat you're probably going to be talking about something softer than psychology.

And like gravity, or the electromagnetic force, can you actually measure them experimentally? I don't think so.

I can think of a few ways to test them, but none are simple and it would require long durations of controlled observation.
teo123 wrote: Wed Mar 13, 2019 3:11 pmWell, not really. I've always assumed that what neuroscience has to say today has very little relevance to how languages actually work. The data from the actual languages has much more value than speculations based on neuroscience.
No, you're not understanding. Neuroscience could prove these laws by pinpointing the neuroanatomical structures that cause people to favor certain sounds. Much in the way we can study optical illusions by better understanding how the eye works and how the brain receives and interprets those signals.

Proving even a small pressure would do a lot; how "micro-evolution" essentially proves "macro-evolution".
A small force applied over a long time can be inferred to do a large amount of work.
teo123 wrote: Wed Mar 13, 2019 3:11 pmFor instance, the supporters of the phonosemantic hypotheses often cite the studies from psycholinguistics that seemingly support their "theories", and ignore the fact that studies on the actual languages don't confirm their "findings" (and they also refuse to answer the obvious question of how can, for example, the Grimm's law be true if the phonosemantic hypotheses are true).
What would be ideal is a concordance between multiple lines of evidence.

The laws you're talking about are ultimately much like the phonosemantic hypotheses. It's the idea that there's something in the brains of human beings that creates a selective pressure for sound changes; one that's consistent across different languages, regions, etc. and operates broadly on whole languages.

It's not as dramatic as saying sounds have innate meanings, but there is a strong claim there and it needs experimental evidence, not just ad hoc support from correlations.
teo123 wrote: Wed Mar 13, 2019 3:11 pmThe claim that English 't' corresponds to German 'z' in the beginning of a word is so precise that it sounds like it can't be true, yet it's so demonstrably true, it's obvious how to test it, and it proves itself again and again.
A mere descriptive claim like that -- which could even just be describing a coincidence -- is trivial. That's not like a law that is presumed to convey something fundamental and prescriptive on how languages evolve.
teo123 wrote: Wed Mar 13, 2019 3:11 pmWhat do you actually think that the Havlik's law says?
It doesn't matter. I will assume linguists are correct about it if it is broadly believed by professionals.

However, as should be, since it's a soft science I will have less confidence in that than the claims from hard sciences.
teo123 wrote: Wed Mar 13, 2019 3:11 pmIs that even a reasonable question to ask? Like I've explained before, the probability that a Slavic language appears to follow the Havlik's law if it doesn't actually do that is so low no mathematical program I've tried to calculate that probability with (using the binomial coefficients) gives me a result. The probability of only 53% of the words in a Slavic language to which the Havlik's law potentially applies appearing to follow the Havlik's law if they don't actually do that is 5%.
Finding some 300 words in older texts in a language to which the Havlik's law potentially applies but doesn't actually do that would perhaps make it reasonable to doubt that the Havlik's law is even approximately valid, but that's very unlikely to happen.
Again, we're talking about an ad hoc observation. The probability of any ad hoc hypothesis fitting known data is very high. Doesn't matter if that data set was random, the hypothesis is made to fit.
teo123 wrote: Wed Mar 13, 2019 3:11 pm
brimstoneSalad wrote:You'd never really be able to falsify it, because it doesn't have any implications beyond those ten apples.
Then we must have very different definitions of falsifiable.
It's either true or false based on the accuracy of the original observation alone. There's not really any additional experiment which can be done to falsify it because it's not making any claims that go beyond the original observation.

What we want in a theory in science is that it's making meaningful claims about the world that can be falsified by new experiments. Things we do not yet know.

"I observed this apple I'm holding and found it to be red, my hypothesis is that the apple I'm holding now is red" is useless nonsense.
A meaningful hypothesis would be something like "I can only hold red apples" which could be falsified by additional data despite the original observation still being correct.
teo123 wrote: Wed Mar 13, 2019 3:11 pmWell, the claim that Latin 's' corresponds to Greek 'h' at the beginning of a word certainly is useful, like when you are trying to remember Latin and Greek words, or when you are trying to guess the meaning of an unknown word.
The utility is limited to the fact itself, and that's fine, but that's not so much science as it is a mnemonic device.
It's as scientific as the observation "righty tighty lefty loosey" when dealing with threaded connections. That is, not.
teo123 wrote: Wed Mar 13, 2019 3:11 pm And yet, as the web-page I linked you to in my last post explains, if you try to propose a common origin of a some languages without proposing the phonetic laws that operated in them since they split (such as one that the English 't' corresponds to German 'z' at the beginning of a word), you are probably practicing pseudoscience. Mere phonetic similarity should play no role in historical linguistics (because that's also not how languages work), what should play a role are phonetic laws that are easy to test.
And you could always do that if you made the "laws" complicated enough
We can try to employ Occam's razor, but that doesn't really help you out with a p value.
Unless you have experimental means to show those laws in operation, it's not going to be a particularly hard science.

Re: Soft Sciences Vs. Hard Sciences

Posted: Fri Mar 15, 2019 5:25 pm
by teo123
brimstoneSalad wrote:Its getting there.
And yet saying "I don't understand how the supernatural claims can be tested, therefore I don't take theology seriously." is somehow not personal incredulity?
brimstoneSalad wrote:Based on consensus behind them in the field
So, how do I estimate, for example, how many physicists actually agree that quantum mechanics is true? Is it 85%, 90%, 95%, 97% or 99%? Quite a big difference there, isn't it?
brimstoneSalad wrote:and the hardness of the field generally.
If anything is obvious from this conversation, it's that you can study some field for years without having an idea of how hard (reliable) it is. I know more about some parts of linguistics than most linguists do, yet I am still not sure how reliable they are.
And I am taking a physics course at the university right now, and it doesn't appear to be a very reliable science either. In our laboratory excercises, we sometimes get "measurement errors" of as much as 50% without being explained why. And the explanations for the formulas usually sound like meaningless word salad to me (How the hell can a sum of many inverse square roots of one minus the square of some number be equal to an arcus sine of that number in radians?!). Linguistics makes so much more sense! Too bad studying it doesn't pay as much!
brimstoneSalad wrote:That's quite fatalistic, and it's a very bold claim about the power of these forces on language.
How is that a bold claim? Look at the history of English. Its speakers have always had a relatively high literacy rate and have valued preserving their language, yet it changed drastically in the last few centuries. The same goes, although not so drastically, for Greek.
brimstoneSalad wrote:I can think of a few ways to test them, but none are simple and it would require long durations of controlled observation.
I don't think there is much need to test it again, you can see it happen by looking at the history of various languages, and how strong that force is can especially be seen in English.
brimstoneSalad wrote:Neuroscience could prove these laws by pinpointing the neuroanatomical structures that cause people to favor certain sounds
That wouldn't be a proof of any kind. Like I've said, somebody who knows neuroscience but doesn't know linguistics would probably conclude phonosemantic hypotheses are true.
brimstoneSalad wrote:Much in the way we can study optical illusions by better understanding how the eye works and how the brain receives and interprets those signals.
And can it also predict optical illusions, rather than just "explain" them?
brimstoneSalad wrote:which could even just be describing a coincidence
And how come, as far as I am aware, no linguist has never suggested that to be a coincidence? There are some fringe theorists who say it's a result of isoglosses that date back to paleolithic and not of Indo-European languages actually being related (of course, that's contradictory to what we see how languages work today), but nobody has ever suggested that to be a coincidence. If you suggested that in a room full of linguists, they'll be laughing at you.
I am not saying scientific consensus is always right about which languages are related. It has made mistakes. For the large part of the 20th-century, it was commonly accepted that Tai-Kadai languages were distantly related to other Sino-Tibetan languages (in reality, they are probably distantly related to the Austronesian languages), and it was commonly accepted that the Altaic languages share a common origin at very distant past (we are talking about connections much deeper in time than Indo-European languages are connected). But when the scientific consensus is wrong, there is usually an interesting reason behind that.
And that reason is usually comparing modern languages instead of reconstructed languages. Comparing reconstructed languages instead of modern languages is a way to control for early loan-words in basic vocabulary, which make, for example, modern Tai-Kadai languages more similar to modern Sino-Tibetan languages than Proto-Tai-Kadai is to Proto-Sino-Tibetan. And relying on computational approaches would make things even worse.
brimstoneSalad wrote:That's not like a law that is presumed to convey something fundamental and prescriptive on how languages evolve.
Well, it's at least something that can be known with reasonable certainty.
brimstoneSalad wrote:However, as should be, since it's a soft science I will have less confidence in that than the claims from hard sciences.
OK, maybe. But that effect is certainly offset by the fact that physics is a lot harder to understand than linguistics is. You should be less confident in your understanding of a particular thing in physics than in your understanding of a particular thing in linguistics. This is especially true for me, since I published a few papers in linguistics which almost guarantee I understand linguistics correctly or close to correct.
brimstoneSalad wrote:"I observed this apple I'm holding and found it to be red, my hypothesis is that the apple I'm holding now is red" is useless nonsense.
I don't think that's the right analogy here. Havlik was looking at a few words from different modern Slavic languages and formulated the Havlik's law. Then he looked at more words from modern Slavic languages and he found that they also fit the Havlik's law (I don't know if he tried to make predictions based on ancient Slavic languages before looking at more words, and I don't think that's really relevant.). Again, what do you think the Havlik's law says? You could be misunderstanding how linguistics works because of that.
brimstoneSalad wrote:The utility is limited to the fact itself, and that's fine, but that's not so much science as it is a mnemonic device.
It also helps explaining many etymologies, let's keep that in mind.
brimstoneSalad wrote:And you could always do that if you made the "laws" complicated enough
Which is why there is emphasis on using simple sound laws to prove that languages have common origin, such as those that English 't' corresponds to German 'z' at the beginning of a word. More complicated laws, such as the Havlik's law, aren't used to prove that languages have common origin, they are used to prove particular etymologies once it's accepted languages have common origin. The probability of there appearing to be a simple sound law describing a relation between unrelated languages is easily calculated, see here.

Re: Soft Sciences Vs. Hard Sciences

Posted: Sat Mar 16, 2019 4:59 am
by brimstoneSalad
teo123 wrote: Fri Mar 15, 2019 5:25 pm
brimstoneSalad wrote:Its getting there.
And yet saying "I don't understand how the supernatural claims can be tested, therefore I don't take theology seriously." is somehow not personal incredulity?
It would be! And that would be wrong.

It's when a theologian says not to put God to the test, and when they admit it's all faith, that you can establish the genuine lack of credibility. Philosophy is distinct from empirical sciences, once it treads there you can dismantle it based on logic alone.

Otherwise it could have been that empirical theological claims are scientific and tested as such, but you just didn't understand how. Many atheists are too quick to dismiss theology. It's right to do so, but not always right to do so with their limited understandings or poor justifications. You don't get points for guessing and getting lucky about dismissing theology.

I have discussed elsewhere at some length some bad atheist arguments.
teo123 wrote: Fri Mar 15, 2019 5:25 pmSo, how do I estimate, for example, how many physicists actually agree that quantum mechanics is true? Is it 85%, 90%, 95%, 97% or 99%? Quite a big difference there, isn't it?
It depends on the claim you're looking at.
Rejecting wave/particle duality would be like flat-Earthism, though. You'd be very hard pressed to find any disagreement on it now that it's been empirically proved. The same was not the case in Einstein's day, where that rejection was a reasonable approach.
teo123 wrote: Fri Mar 15, 2019 5:25 pmIf anything is obvious from this conversation, it's that you can study some field for years without having an idea of how hard (reliable) it is. I know more about some parts of linguistics than most linguists do, yet I am still not sure how reliable they are.
You should be able to compare pretty easily, we already discussed a few ways to do it, but...
teo123 wrote: Fri Mar 15, 2019 5:25 pmAnd I am taking a physics course at the university right now, and it doesn't appear to be a very reliable science either. In our laboratory excercises, we sometimes get "measurement errors" of as much as 50% without being explained why.
Clumsy physics lab error isn't a good metric. :)
You can look at professional work to give you a better understanding.
teo123 wrote: Fri Mar 15, 2019 5:25 pmAnd the explanations for the formulas usually sound like meaningless word salad to me (How the hell can a sum of many inverse square roots of one minus the square of some number be equal to an arcus sine of that number in radians?!).
That's a mathematics issue. You'll have to take more advanced math classes. Are you not in calculus?
teo123 wrote: Fri Mar 15, 2019 5:25 pmLinguistics makes so much more sense! Too bad studying it doesn't pay as much!
Well, linguistics is more intuitive because we have a lot of wetware for it, and we've also spent our lives passively learning it.
teo123 wrote: Fri Mar 15, 2019 5:25 pm How is that a bold claim? Look at the history of English. Its speakers have always had a relatively high literacy rate and have valued preserving their language, yet it changed drastically in the last few centuries.
English is one of the worst examples, since it has become an international language and the main language of some "melting pot" countries thus many pidgins and diffusion of vocabulary and grammar. As well, it very vigorously adopts words from other languages. I don't think there's much reason to believe the average English speaker is interested in preserving its purity.

Regardless, the "tried and failed" argument isn't a very good one without any hard metrics and control of other variables.
teo123 wrote: Fri Mar 15, 2019 5:25 pmI don't think there is much need to test it again, you can see it happen by looking at the history of various languages, and how strong that force is can especially be seen in English.
You still do not understand. It has not been tested a first time.
teo123 wrote: Fri Mar 15, 2019 5:25 pmLike I've said, somebody who knows neuroscience but doesn't know linguistics would probably conclude phonosemantic hypotheses are true.
That's unlikely. You seem to think neuroscientists are idiots who don't understand that random correlations will occur with a limited number of available sounds.
To even study the issue and arrive at a p value, you'd need to have a good grasp on linguistics too.
teo123 wrote: Fri Mar 15, 2019 5:25 pmAnd can it also predict optical illusions, rather than just "explain" them?
YES! Actually, some fun ones have been invented based on our knowledge of the brain, or using what we know about small illusions and creating stronger ones
Like yellowish-blue.
teo123 wrote: Fri Mar 15, 2019 5:25 pmIf you suggested that in a room full of linguists, they'll be laughing at you.
Wouldn't a room full of theologians do the same of the suggestion that a god doesn't exist?
Nobody really wants to hear their discoveries aren't validated and may just be description of a normal coincidence in random data.
teo123 wrote: Fri Mar 15, 2019 5:25 pmBut when the scientific consensus is wrong, there is usually an interesting reason behind that.
"Interesting" meaning ad hoc explanation to vindicate the mistake?
teo123 wrote: Fri Mar 15, 2019 5:25 pm
brimstoneSalad wrote:That's not like a law that is presumed to convey something fundamental and prescriptive on how languages evolve.
Well, it's at least something that can be known with reasonable certainty.
That's fine, and soft sciences are often good at descriptive analysis.
Look at sociology and opinion surveys.

That's not really applying experimental methodology though. I explained the difference, and it's an important one to understand to grasp the comparative limitations.
teo123 wrote: Fri Mar 15, 2019 5:25 pm
brimstoneSalad wrote:However, as should be, since it's a soft science I will have less confidence in that than the claims from hard sciences.
OK, maybe. But that effect is certainly offset by the fact that physics is a lot harder to understand than linguistics is.
IF understanding had anything to do with it, sure. But again, it really shouldn't.

I could point to two facts in Physics and Chemistry that I have an equally poor understanding of, but there's a reason I'm more confident in the consensus on physics.

You can have confidence in consensus without having confidence in your understanding of the why etc.
teo123 wrote: Fri Mar 15, 2019 5:25 pmI don't think that's the right analogy here. Havlik was looking at a few words from different modern Slavic languages and formulated the Havlik's law. Then he looked at more words from modern Slavic languages and he found that they also fit the Havlik's law (I don't know if he tried to make predictions based on ancient Slavic languages before looking at more words, and I don't think that's really relevant.).
You look at a few apples from a box and find them red, you say "I think this box contains red apples" then you look at more and find those red too. Even if you look at the rest of them and find them all red.

So, what does that say about the content of *other* boxes exactly?
teo123 wrote: Fri Mar 15, 2019 5:25 pmAgain, what do you think the Havlik's law says? You could be misunderstanding how linguistics works because of that.
It's not the precise content of the claims, it's the methodology of the field that I'm talking about.
teo123 wrote: Fri Mar 15, 2019 5:25 pmIt also helps explaining many etymologies, let's keep that in mind.
Explaining is not always predicting, is it?
Sometimes you get the chance to really predict, and that's cool. Very rarely can you get those predictions confirmed, and that's cooler when it happens. The main point is that's rarer in linguistics than it is in other sciences.

It's a softer science, I didn't say it's not a science at all. It's on the soft side (pretty far on it), I don't know why you find that controversial.
teo123 wrote: Fri Mar 15, 2019 5:25 pmThe probability of there appearing to be a simple sound law describing a relation between unrelated languages is easily calculated, see here.
I don't think it is when you have so few samples. It's like trying to calculate the probability of life evolving on a suitable planet with a sample of one (Earth).

Re: Soft Sciences Vs. Hard Sciences

Posted: Sun Mar 17, 2019 1:24 pm
by teo123
brimstoneSalad wrote:Philosophy is distinct from empirical sciences, once it treads there you can dismantle it based on logic alone.
Well, I don't think philosophical arguments about the empirical world are to be dismissed a-priori. Perhaps something indeed logically follows (or is very probable) from what we observe, but it's very hard to test experimentally. The ontological argument for the existence of God is wrong because it commits the equivocation fallacy (something being conceivable in the sense of being easy to imagine, with something being conceivable in the sense of existing in some possible world), not because it doesn't rely on any empirical data, right?
brimstoneSalad wrote:Otherwise it could have been that empirical theological claims are scientific and tested as such, but you just didn't understand how.
So, when somebody makes claims that seem not to be falsifiable, it's not the right thing to do to be skeptical of those claims?
What about, for example, Christian Science? They claim their beliefs are based on proofs, rather than opinion and dogmas. Is it therefore somehow more reasonable to believe them than it is to believe theology?
brimstoneSalad wrote:Rejecting wave/particle duality would be like flat-Earthism, though.
How exactly? There is nothing like the "Ships appear to sink when going over the horizon."-proof of the Earth being round to prove wave-particle-duality, right? I mean, yeah, there are everyday "proofs" of quantum entanglement such as the one with three polarizing filters, and there are everyday proofs of the Heisenberg's uncertainty principle such as one that the light appears to disperse when going through a small enough slit, but none of that screams that the quantum mechanics is true nearly as much as the sinking-ship illusion screams that the Earth is round. At best, you can put quantum mechanics denial in the same category as denying that Armenian is distantly related to other Indo-European languages: it can be proved, but you need to have some specialized knowledge to understand the proof.
brimstoneSalad wrote:You can look at professional work to give you a better understanding.
Don't you think you are asking a little too much of me?
brimstoneSalad wrote: That's a mathematics issue. You'll have to take more advanced math classes. Are you not in calculus?
Well, yes, I am taking calculus classes. We were also taught some calculus in high-school (I think the last thing we were taught was integration by parts), and I got a C in it. In the first semester, we were taught (among other things) derivatives in our math classes, and I barely got a passing score. I mean, the basics of derivation make some sense to me, but I don't even understand what's the intuition behind the chain rule, yet alone something more. Now, in the second semester, we are being taught integrals, and I must admit I don't really understand anything. Why on Earth would calculating the area below some curve be the opposite of calculating the rate at which that curve goes up or down? What the hell would be the intuition behind the integration by parts?
And think of it this way: I scored 93% on the mathematics part of the maturity test and was among the first 100 in Croatia in my generation. If I cannot evaluate the claims the physics makes, chances are, nobody really can. They are effectively unfalsifiable partly because it takes a very deep understanding of mathematics to actually test them.
brimstoneSalad wrote: Well, linguistics is more intuitive because we have a lot of wetware for it, and we've also spent our lives passively learning it.
I don't think we've spent our lives passively learning about sound laws. What makes you think some parts of our brain evolved specifically for language learning?
brimstoneSalad wrote:English is one of the worst examples, since it has become an international language and the main language of some "melting pot" countries thus many pidgins and diffusion of vocabulary and grammar. As well, it very vigorously adopts words from other languages.
I was actually referring to the Great Vowel Shift, which happened before English became a lingua franca. It's a drastic and a relatively rapid (making the language virtually incomprehensible to a speaker familiar with the language before the vowel shift within a time period of about two centuries) change in English phonology. Even in medieval times, English phonology was changing rather fast despite a relatively high literacy rate (medieval writers often complained about English spelling not matching pronunciation).
brimstoneSalad wrote: I don't think there's much reason to believe the average English speaker is interested in preserving its purity.
OK, let's then ignore English. You realize that the traditional pronunciation of Latin in various countries also changed quite a bit over the centuries? Not as much as the phonology of Romance languages has changed, but still enough that, for example, Croatians can't recognize Latin words in English, and sometimes even in Italian or German, until they see them spelled. Do you have any doubt in your head that those who used Latin were interested in preserving the language?
What forces accelerate the rate at which sound changes happen is an interesting question, and, from what I've learned by now, it seems to me it has more to do with intralinguistic than with extralinguistic factors. Some languages simply happen to have very unstable phonologies and sound changes are a way for speakers to deal with that. You know, the Late Latin vowels changed drastically but in a relatively similar way in many Romance languages, probably because the vowel system in Late Latin was particularly unstable (after the Sardinian language split off, the vowel length distinction disappeared and there were many words which sounded the same except for a slight difference in vowel quality). Similarly, the sound changes described by the Havlik's law affected all Slavic languages and all their dialects, presumably because the phonology of the ancient Slavic languages begged for something like that to happen (the phonotactics was even more restrictive than modern Japanese phonotactics is, and the short 'i' and short 'u' were disproportionately more common than other vowels were). Also, the Grimm's law appears to describe both the sound changes that happened in Proto-Germanic and the sound changes that happened in Armenian quite accurately, presumably because the sound changes described by the Grimm's law are a natural way to deal with the instabilities of the Proto-Indo-European phonology (the existence of voiced aspirated stops without the corresponding voiceless aspirated stops, not resembling a phonological system of any known language today).
Regardless, that is irrelevant here. Sound changes happen no matter what. Denying that makes you linguistically illiterate, much like denying the law of supply and demand makes you economically illiterate.
brimstoneSalad wrote:Regardless, the "tried and failed" argument isn't a very good one without any hard metrics and control of other variables.
OK, so, what are those other factors? And why do you consider "Before FDA, all you had was snake oil." or "Before the government, the genocides were incredibly common." to be valid arguments, when THOSE arguments scream "There are other factors.".
brimstoneSalad wrote: You still do not understand. It has not been tested a first time.
Just like socialists keep insisting socialism hasn't been tested, when it clearly has been and it failed every time.
brimstoneSalad wrote:You seem to think neuroscientists are idiots who don't understand that random correlations will occur with a limited number of available sounds.
Neuroscientists who talk about linguistics are not to be trusted, they are one of those intellectuals that step outside of their field of expertise, and usually do more harm than good. Natural scientists are not immune to that. Just like Brian Josephson, a Nobel-prize winning theoretical physicists, makes claims about neuroscience, that is about how quantum mechanics makes psychic powers possible. And social sciences are even less related to natural sciences than various natural sciences are related to each other.
You need to learn how to recognize pseudoscience in linguistics. Much like if somebody uses quantum mechanics to explain something about neuroscience, he is most likely wrong, so too is somebody who uses neuroscience to explain something about linguistics probably wrong. You want to tell me that the way synapses work has something to do with quantum mechanics? Fine, that's a plausible assertion. You want to tell me quantum mechanics explains memories and thought? Eh, you immediately lose all your credibility. You want to tell me neuroscience explains why some sound that seems easy to pronounce doesn't occur in languages? Fine, that's a plausible assertion, although it's hard to test. You want to tell me neuroscience has something to say about morphology or syntax? I don't trust you. You want to tell me Chinese language makes people think more with the right hemisphere of their brains, and English makes people think more with the left hemispheres of their brains? Shut the hell up!
brimstoneSalad wrote:To even study the issue and arrive at a p value, you'd need to have a good grasp on linguistics too.
The real question is how it can even be tested. Phonosemantic hypotheses appear to make two implications:
1) You can guess the meanings of the words in unrelated languages by examining the sounds alone, at a rate significantly greater than chance.
This is practically unfalsifiable because it's very hard to quantify how close a guessed meaning is to the real meaning of that word.
2) Word borrowings don't happen between related languages because of the regular sound correspondences.
And this is clearly false. If 'h' meant "color" in English (as in "hue"), how come did English borrow the Latin cognate "color"? And there appears to be no ad-hoc hypothesis around that. If you claim that phonosemantics only applies cross-linguistically, that's basically the same as saying regular sound changes don't happen, yet they clearly do. If you will claim that 't' means something related to finger in all languages, as in English "toe", this is clearly false since English 't' regularly corresponds to German 'z' (Zeh) and to Latin 'd' (digitus).
brimstoneSalad wrote:Like yellowish-blue.
What about it actually? And, by the way, I tried and failed to see that. If I cross my eyes so that the plus sign in the blue square is on the top of the one in the yellow square, I just see a square randomly turning blue and yellow.
brimstoneSalad wrote:Wouldn't a room full of theologians do the same of the suggestion that a god doesn't exist?
Perhaps some would. But theologians studying Buddhism would agree with you. There is no consensus among theologians that a god exists.
brimstoneSalad wrote: "Interesting" meaning ad hoc explanation to vindicate the mistake?
What's the ad-hoc explanation there? It's the self-correcting nature of science. The linguists proposed distant relations between understudied languages, not exactly following the scientific methods (which were very difficult to apply to such under-studied languages). Few decades later, when those languages had been studied more, those proposals were proven wrong. They were over-confident because there appeared to be regular sound laws operating between those language families, and they didn't take into account the possibility that those sound laws would disappear once you reconstruct the ancestors of those languages (which was shown to be true decades later).
brimstoneSalad wrote:Look at sociology and opinion surveys.
I thought those surveys almost always give the results that are indistinguishable from guessing, just like animal testing and a few other things that the public views as scientific.
https://www.youtube.com/watch?v=7Y8cfhQ ... ex=22&t=0s
brimstoneSalad wrote:IF understanding had anything to do with it, sure. But again, it really shouldn't.
Of course it should. If you accept something that you misunderstood as true, your model of reality will be worse because of that, and not better.
brimstoneSalad wrote:So, what does that say about the content of *other* boxes exactly?
I don't understand your analogy here. Again, what do you think the Havlik's law actually says? If you say some claim is unfalsifiable, you should understand, at least superficially, what that claim is. It's at your fingertips (and you can probably find better explanations in English if you Google for a few minutes). It seems like you are being willfully ignorant about it.
brimstoneSalad wrote:Explaining is not always predicting, is it?
Well, yes. But if all you value are predictions, you can often predict what a word would sound like in some Slavic language by taking a Lithuanian word and applying the sound laws to it.
brimstoneSalad wrote: It's on the soft side (pretty far on it), I don't know why you find that controversial.
How is linguistics pretty far on it? I don't see how you can say something like that. When has economics ever predicted anything? And when has sociology ever predicted anything? It's a field incredibly polluted by political agendas such as feminism. And social psychology is incredibly polluted by "experiments" and "observations" that are nothing but p-hacking.
brimstoneSalad wrote:I don't think it is when you have so few samples.
You didn't even read what I linked you to, did you? OK, I can copy that here:
Daniel wrote:
FlatAssembler wrote:I have found a simple sound law (that PIE *s corresponds to PAN *q), and there are six examples of that on the Swadesh list. What's the probability of that if they aren't closely related?
.... The probability of finding such a pattern in two truly unrelated languages is 1-(1-0.0031)ˆ20=6%. That's pretty low.
You have calculated the probability of that particular correspondence, which indeed is very unlikely. But you weren't just looking for one particular correspondence. You were looking for any correspondence-- q/s, q/r, q/q, etc., etc. So for 20 consonants, there are 20*20 = 400 possible correspondences. So you must consider what the odds are that ANY of the 400 correspondences would appear.

Let's look at an analogous problem:
What are the odds that you share the same birthday with your friend? 1/365, right? (Or 1/366 or 365.25 to be precise.)
This means it would be surprising if any friends had the same birthday, right? No. Surprising things happen all the time. That's how statistics work.
Imagine a classroom: what are the odds that someone else in the room shares your birthday? If there are 30 other people, then that's 30/365-- low odds, but still possible. But what about any two people sharing that same birthday?
(...)
By just 24 people in the room, the odds are that some pair shares a birthday! So finding any correspondence between languages is not unlikely at all.

So assuming .31% is right (it's more complicated than that*), that's roughly 1/322 (.31/100). 322 is close to 365, so we can use the birthday problem by analogy. Let's assume 20 sounds in each language: there's a 41.1% chance of a 'shared birthday' (see Wikipedia), or in our case a chance sound correspondence. Actually, it's a little higher than that (or substantially, because 365-322 is a big difference when multiplied out a lot as fractions), so let's say about 50%. That's very different from the 0.31% you calculated!

(The main problem is starting with the Swadesh list because it's too small: you need systematic correspondences across a wide variety of words. Ideally, the correspondences would be exceptionless, setting aside borrowings, and of course any contextually conditioned words. The Swadesh list gives you a good starting point, but it's far from enough to fully determine linguistic relationship.)

Regardless, that math is irrelevant because it's a more complex problem. Finding a recurring correspondence in different words is actually less likely than the math given by either of us. But I'm certain that if you looked at more vocabulary you would start to find exceptions, so it's again irrelevant anyway. The issue is that you would need to find genuine cognates (not just the same number in the Swadesh list) to start to establish real patterns. It's an interesting coincidence, but coincidences happen all the time. Just think about how the odds are stacked against an American PhD student 'randomly' talking to a Croatian 17 year old online! (But the odds are very high that any two 'random' people would be talking online-- a very different question.)
(...)
Having no chance correspondences would actually be surprising. The burden of proof is more than that: it's finding multiple, systematic, widespread correspondences.