Uralic homeland and genetics and their implications for PIE

[Jaska]

People from different cultures don't mix readily. It's a process that takes time and usually shows very specific patterns.

Of course they do. There is no single widespread language family with the uniform ancestry. Not necessarily within one or two generations, but they always mix with locals.

If Seima-Turbino network is connected to the Uralic spread, it must have been male-biased, meaning stronger autosomal admixing with the local populations.

[Jaska]

I think we can cut Jaska a bit of slack here. We all have biases, and I agree his argumentation in parts where he tries to tie linguistics to archaeological cultures and populations is super poor (it makes sense the members of this forum may be more informed about these matters than even professionals--this is a very new field anyway).

Here you have misunderstood something. I never try to tie linguistic results to archaeological cultures, just like I never try to tie linguistic results to genetic results - just the opposite: I encourage everybody not to do it. Every connection between language, culture and DNA are always momentary and have no predictive force in another time or place.

That is the reason why it is not high in my priority list to find archaeological culture matching the linguistic results. But when I do so, I do it with the scientific method:
1. I take the linguistic results.
2. I take the archaeological results.
3. I find a match concerning time and place.

That' why the Garino-Bor culture is a possible example: it matches the time and place, where we know there were Proto-Uralic speakers. Still, we must acknowledge that cultural limits and linguistic limits do not always meet, and there could have been several languages spoken within one culture.

I am also familiar with all the same aDNA studies than other people here. I just know that genetic results can never overrule the linguistic results concerning language, just like I know that linguistic results can never overrule the genetic results concerning DNA.

Many loud members in this forum seem to think that genetics could force the interpretation of archaeological cultures, just like they think that genetics can force the interpretation of linguistic results. Still, that is and always will be an unscientific view. The bubble will break eventually.

Those guys here are not on the front wave of a new science, but of a pseudoscience. Whatever they think they can trace along with genetic lineages and ancestries, it is not language - it is just DNA. They are only repeating mistakes of other erroneously multidisciplinary researchers like Colin Renfrew, Mario Alinei, Kalevi Wiik etc.

[Generalissimo]

It follows, that any lineage and any ancestry present in the Northeastern Europe gets similar correlation with the Uralic languages. Just like any lineage and any ancestry in the Northwestern Europe gets a correlation with the Germanic languages.

But the correlation between N-L1026, Nganasan-related ancestry and Uralic languages isn't just found in Northeastern Europe. It's also found in Siberia and in ancient DNA from the Carpathian Basin (Hungarian conquerors).

Why do you think that N-L1026 and Nganasan-related ancestry correlate better with Uralic speakers and ancient peoples regarded to have been Uralic speakers than with groups from any other language family, especially considering the rapid and relatively late spread of N-L1026 and Nganasan-related ancestry across North Eurasia?

Is there a plausible alternative explanation to Proto-Uralics and early Uralics being rich in N-L1026 and Nganasan-related ancestry?

[Generalissimo]

Here you have misunderstood something. I never try to tie linguistic results to archaeological cultures, just like I never try to tie linguistic results to genetic results - just the opposite: I encourage everybody not to do it. Every connection between language, culture and DNA are always momentary and have no predictive force in another time or place.

That is the reason why it is not high in my priority list to find archaeological culture matching the linguistic results. But when I do so, I do it with the scientific method:
1. I take the linguistic results.
2. I take the archaeological results.
3. I find a match concerning time and place.

That' why the Garino-Bor culture is a possible example: it matches the time and place, where we know there were Proto-Uralic speakers. Still, we must acknowledge that cultural limits and linguistic limits do not always meet, and there could have been several languages spoken within one culture.

I am also familiar with all the same aDNA studies than other people here. I just know that genetic results can never overrule the linguistic results concerning language, just like I know that linguistic results can never overrule the genetic results concerning DNA.

Many loud members in this forum seem to think that genetics could force the interpretation of archaeological cultures, just like they think that genetics can force the interpretation of linguistic results. Still, that is and always will be an unscientific view. The bubble will break eventually.

Those guys here are not on the front wave of a new science, but of a pseudoscience. Whatever they think they can trace along with genetic lineages and ancestries, it is not language - it is just DNA. They are only repeating mistakes of other erroneously multidisciplinary researchers like Colin Renfrew, Mario Alinei, Kalevi Wiik etc.

This is a genetics forum, and most people, including myself, are actually more interested in the spread of genes than languages.

The reason that the Uralic language expansion is of such an interest to us, is because it's the best explanation for the rapid spread of N-L1026 and the very specific Nganasan-related autosomal ancestry across North Eurasia during the metal ages.

Genes don't just spread via osmosis for no reason, especially when they do it this fast in very specific directions. The expansions of N-L1026, Nganasan-related ancestry, and Uralic languages are obviously intimately connected.

You can't accuse us of pseudoscience for noticing this obvious fact.

[DMXX]

There's an element of talking-past-each-other in this discussion that's difficult to ignore.

Jaska's fundamental points across several threads (linguistic problems cannot be readily solved by genetic, archaeological etc. data; a robust multidisciplinarian approach is necessary to either solve, approximate or construct parsimonious theories pertaining to human migration patterns or ethnogenesis; the typical approach had on lay discussion forums cannot conceivably replace any robust multidisciplinarian approach)... As far as I'm concerned, these statements are completely true.

However - And in defense of the members arguing in the opposite direction - There are several points that either work against a practical application of these points, or underscore the fragility of the paradigm in which these points are expected to function within:

• As Ryu's stated, the Uralic linguistic data's (apparently*) unsettled. If a consensus hasn't been reached (yet?) with the linguistic question concerning the pU urheimat, it may be considered reasonable (on the basis of alternate practices within linguistics**) to appraise urheimat scenarios based on other, ostensibly secure, non-linguistic evidence. This may not be "best practice" (or "ideal practice"), but it's certainly not unheard of within linguistics (further below).
• Jaska's general position in these discussions is that lay individuals have a strong tendency to employ a quasi-Bayesian, non-methodological approach to linguistic problems (or attempt a multidisciplinarian take without a clear methodology), potentially guided by some form of bias (clearly systemic if there isn't a predefined methodology). While I agree completely, and in accordance with the first point, it's somewhat unfair to criticise the well-versed lay participants of our community for such an approach, when plenty of published linguists had also operated similarly, historically-speaking. The intentionality of that bias is a separate matter (based on my experience with the members here, assuming better should be the default).
• I come from a different academic background to Jaska (STEM). There may be a difference in application with respect to conceptualising, implementing and validating scientific questions in linguistics. My impression (based on these papers**) is that the requirement for specific and discrete research question design and hypothesis formation that's seen in STEM isn't universal in linguistics (perhaps historically; from his other posts, Jaska would probably appropriately refer to these as "junk linguistics" papers, which I'd agree with, from a research design perspective).
  Most users on any forum aren't trained in the scientific method, so the conception->background->research Q->hypothesis conveyor belt (with all the bells and whistles per step) isn't adopted. Instead, these discussions (by virtue of the environment and the background of those participating) typically involve a hive-mind, quasi-Bayesian maelstrom (with a silent hope that individual biases are modulated through the confluence of differing backgrounds). One cannot reasonably criticise a lay community for not undertaking a scientific discovery process when the majority are either ignorant of said process, or aren't trained to competently execute it***. I suspect this is (understandably) why a professional linguist who's attempting to elevate the state of linguistics with a rigorous STEM-based approach, may become frustrated with the proceedings in these discussions (on a personal note, as a STEM guy, the use of the word "correlation" without any statistical data is puzzling at best).
• Derivative hypotheses, in some circumstances, may be appropriately forwarded with a positive proposition. These may be axiomatic (i.e. based on preceding perspectives, even if they're anecdotes from qualified experts, which some of the conclusions in linguistic literature appear to be - Placing them at the bottom of the hierarchy of evidence).
  This is precisely what many members are doing in this thread - It may not be the "ideal" approach (what's "ideal" is a moot point given none of these hypotheses can be assessed for stats significance anyway), but it's certainly a valid approach. Expecting a well-read, brighter-than-average lay community to adhere to "best practices" with research question formation, hypothesis formation and hypothesis testing is (in accordance with the above) also somewhat unfair.
• Formal, statistical assessment of the genetic data to determine relationships between genetic and non-genetic variables isn't a regular occurrence in these population genetics papers. As far as I'm aware, regression analysis, correlation matrices etc. aren't a staple of these aDNA papers (which, since Lazaridis, basically follow the usual formula of uniparentals, f3/f4/qpWave/qpAdm, a PCA or two - Maybe an ADMIXTURE chart to keep those stuck in 2008 happy). I've never seen, for instance, an official correlation matrix between the frequency of Y-DNA N1c, Nganasan-like admixture and Uralic languages among any cohort of samples belonging to one well-defined time interval. Ergo, an argument of correlation is conjectural either way.

Summarised, the position(s) and means of reaching this/these position(s) by our members aren't unreasonable, nor are they substantively less methodological than a lot of what I've (we've?) seen in historical linguistics papers. Expecting an ideal standard strikes me as an unfair position, given the above.

There are, however, some productive ways to address this interesting dilemma:

• Jaska - If you happen to be aware of any current "best practice" guidelines regarding a methodological assessment of linguistic issues, we'd be very grateful to receive this from you. It'd certainly elevate the level of discourse, in addition to perhaps granting some in the community with a clearer understanding of your perspective and why the scattershot, here's-a-paper-I'll-hinge-my-views-on collective approach to complex problems isn't the best one.
• At the very least, if more members of the forum understand research questions and hypotheses from a strictly-academic perspective, we'll have a much more productive set of discussions in the future. There are plenty of authoritative books on the matter. For a quick read, see this.
• An understanding that everyone's here to learn, is well-meaning and willing to engage in productive discourse, and that actions derivative of the alternative won't benefit anyone (which includes borderline-ad hominem's about wishing to "remain ignorant" or "grandstanding"). General point.
• More quantification, less qualification (f.ex. describing a correlation without a formal, statistical assessment of correlation means little).

A fruitful engagement in the above may result in this apparent impasse becoming a watershed moment of sorts in the lay-professional interaction zone (which would be hugely beneficial for all).

* I am not especially well-read on this matter and rely on what I've read incidentally concerning the state of Uralic research in the context of IE interactions. Hands raised - I am an Uralic ubernewb.

** Over the years, I've encountered quite a few linguistic papers - Chiefly those pertaining to pIIr or various IIr branches - That appeared to arbitrarily cite non-linguistic evidence to scaffold a particular assertion (I'll link specific examples if I revisit this thread later on).

*** No gatekeeping - Informally learning the scientific method is clearly feasible thanks to the Internet, and at least half of this forum (easily) is competent-enough to adopt it.

[Generalissimo]

There's an element of talking-past-each-other in this discussion that's difficult to ignore.

Jaska's fundamental points across several threads (linguistic problems cannot be readily solved by genetic, archaeological etc. data; a robust multidisciplinarian approach is necessary to either solve, approximate or construct parsimonious theories pertaining to human migration patterns or ethnogenesis; the typical approach had on lay discussion forums cannot conceivably replace any robust multidisciplinarian approach)... As far as I'm concerned, these statements are completely true.

However - And in defense of the members arguing in the opposite direction - There are several points that either work against a practical application of these points, or underscore the fragility of the paradigm in which these points are expected to function within:

• As Ryu's stated, the Uralic linguistic data's (apparently*) unsettled. If a consensus hasn't been reached (yet?) with the linguistic question concerning the pU urheimat, it may be considered reasonable (on the basis of alternate practices within linguistics**) to appraise urheimat scenarios based on other, ostensibly secure, non-linguistic evidence. This may not be "best practice" (or "ideal practice"), but it's certainly not unheard of within linguistics (further below).
• Jaska's general position in these discussions is that lay individuals have a strong tendency to employ a quasi-Bayesian, non-methodological approach to linguistic problems (or attempt a multidisciplinarian take without a clear methodology), potentially guided by some form of bias (clearly systemic if there isn't a predefined methodology). While I agree completely, and in accordance with the first point, it's somewhat unfair to criticise the well-versed lay participants of our community for such an approach, when plenty of published linguists had also operated similarly, historically-speaking. The intentionality of that bias is a separate matter (based on my experience with the members here, assuming better should be the default).
• I come from a different academic background to Jaska (STEM). There may be a difference in application with respect to conceptualising, implementing and validating scientific questions in linguistics. My impression (based on these papers**) is that the requirement for specific and discrete research question design and hypothesis formation that's seen in STEM isn't universal in linguistics (perhaps historically; from his other posts, Jaska would probably appropriately refer to these as "junk linguistics" papers, which I'd agree with, from a research design perspective).
  Most users on any forum aren't trained in the scientific method, so the conception->background->research Q->hypothesis conveyor belt (with all the bells and whistles per step) isn't adopted. Instead, these discussions (by virtue of the environment and the background of those participating) typically involve a hive-mind, quasi-Bayesian maelstrom (with a silent hope that individual biases are modulated through the confluence of differing backgrounds). One cannot reasonably criticise a lay community for undertaking a scientific discovery process when the majority are either ignorant of said process, or aren't trained to competently execute it. I suspect this is (understandably) why a professional linguist who's attempting to elevate the state of linguistics with a rigorous STEM-based approach, may become frustrated with the proceedings in these discussions (on a personal note, as a STEM guy, the use of the word "correlation" without any statistical data is puzzling at best).
• Derivative hypotheses, in some circumstances, may be appropriately forwarded with a positive proposition. These may be axiomatic (i.e. based on preceding perspectives, even if they're anecdotes from qualified experts, which some of the conclusions in linguistic literature appear to be - Placing them at the bottom of the hierarchy of evidence).
  This is precisely what many members are doing in this thread - It may not be the "ideal" approach (what's "ideal" is a moot point given none of these hypotheses can be assessed for stats significance anyway), but it's certainly a valid approach. Expecting well-read, brighter-than-average lay community to adhere to "best practices" with research question formation, hypothesis formation and hypothesis testing is (in accordance with the above) also somewhat unfair.
• Formal, statistical assessment of the genetic data to determine relationships between different variables isn't a regular occurrence in these population genetics papers. As far as I'm aware, regression analysis, correlation matrices etc. aren't a staple of these aDNA papers (which, since Lazaridis, basically follow the usual formula of uniparentals, f3/f4/qpWave/qpAdm, a PCA or two - Maybe an ADMIXTURE chart to keep those stuck in 2008 happy). I've never seen, for instance, an official correlation matrix between the frequency of Y-DNA N1c, Nganasan-like admixture and Uralic languages among any cohort of samples belonging to one well-defined time interval. Ergo, an argument of correlation is conjectural either way.

Summarised, the position(s) and means of reaching this/these position(s) by our members aren't unreasonable, nor are they substantively less methodological than a lot of what I've (we've?) seen in historical linguistics papers. Expecting an ideal standard strikes me as an unfair position, given the above.

There are, however, some productive ways to address this interesting dilemma:

• Jaska - If you happen to be aware of any current "best practice" guidelines regarding a methodological assessment of linguistic issues, we'd be very grateful to receive this from you. It'd certainly elevate the level of discourse, in addition to perhaps granting some in the community with a clearer understanding of your perspective and why the scattershot, here's-a-paper-I'll-hinge-my-views-on collective approach to complex problems isn't the best one.
• At the very least, if more members of the forum understand research questions and hypotheses from a strictly-academic perspective, we'll have a much more productive set of discussions in the future. There are plenty of authoritative books on the matter. For a quick read, see this.
• An understanding that everyone's here to learn, is well-meaning and willing to engage in productive discourse, and that actions derivative of the alternative won't benefit anyone (which includes borderline-ad hominem's about wishing to "remain ignorant" or "grandstanding"). General point.
• More quantification, less qualification (f.ex. describing a correlation without a formal, statistical assessment of correlation means little).

A fruitful engagement in the above may result in this apparent impasse becoming a watershed moment of sorts in the lay-professional interaction zone (which would be hugely beneficial for all).

* I am not especially well-read on this matter and rely on what I've read incidentally concerning the state of Uralic research in the context of IE interactions. Hands raised - I am an Uralic ubernewb.

** Over the years, I've encountered quite a few linguistic papers - Chiefly those pertaining to pIIr or various IIr branches - That appeared to arbitrarily cite non-linguistic evidence to scaffold a particular assertion (I'll link specific examples if I revisit this thread later on).

I don't think this is a fair assessment, for two reasons.

One, the correlations between N-L1026, Nganasan-related ancestry and Uralic expansions have been discussed at length in very recent papers using the latest methods, and Jaska claims to be well aware of these papers.

Two, Jaska just pinned Garino-Bor, seemingly off the cuff, as the candidate for the Proto-Uralic culture.

Indeed, when I asked him to tell us what the geographic anchors were for a Proto-Uralic homeland in Europe (apart from the contested Indo-Aryan contacts), he refused to answer, citing a bad attitude by CopperAxe as the reason, or some such.

Also, keep in mind that Jaska isn't the only linguist working on the Proto-Uralic homeland problem, he's just the only one posting at this forum, so his approach shouldn't be assumed to be the academic standard.

In fact, there were at least two linguistic papers published recently putting the Proto-Uralic homeland in Siberia. Why should we ignore these papers, which included very coherent arguments, just because Jaska is here?

[Generalissimo]

So Jaska, are you going to answer my questions?

Why do you think that N-L1026 and Nganasan-related ancestry correlate better with Uralic speakers and ancient peoples regarded to have been Uralic speakers than with groups from any other language family, especially considering the rapid and relatively late spread of N-L1026 and Nganasan-related ancestry across North Eurasia?

This means, of course, that Indo-Iranian influences in Proto-Uralic can't be used as a geographic constraint to claim that the Proto-Uralic homeland was in Europe.

Is there any other reliable evidence/constraint putting the Proto-Uralic homeland in Europe?

[DMXX]

One, the correlations between N-L1026, Nganasan-related ancestry and Uralic expansions have been discussed at length in very recent papers using the latest methods, and Jaska claims to be well aware of these papers.

Were these correlations formally measured through formal statistical means, or merely described qualitatively?

If the former, and if these two papers covered the appropriate timeframes and indirectly validated one another, and if there wasn't a methodological weakness to any underlying assumptions regarding their transformation of qualitative data to quantitative, then there isn't much more to discuss.

A rhetorical question, as I don't intend to get bogged-down in this thread.

Two, Jaska just pinned Garino-Bor, seemingly off the cuff, as the candidate for the Proto-Uralic culture.

Indeed, when I asked him to tell us what the geographic anchors were for a Proto-Uralic homeland in Europe (apart from the contested Indo-Aryan contacts), he refused to answer, citing a bad attitude by CopperAxe as the reason, or some such.

This specific description of the recent events in this thread does not invalidate the general issue(s) that Jaska has rightly raised regarding the process of problem-solving in forums such as this from a "best practices" perspective.

Also, keep in mind that Jaska isn't the only linguist working on the Proto-Uralic homeland problem, he's just the only one posting at this forum, so his approach shouldn't be assumed to be the academic standard.

To the underlined - Of course not. You'll find plenty of seemingly well-crafted STEM papers that, on closer inspection, contained some form of human error, methodological flaw or bias that does affect the results.

I've been involved in STEM peer review. Take it from me - Some highly questionable stuff from otherwise decent departments enter the process and get through.

In fact, there were at least two linguistic papers published recently putting the Proto-Uralic homeland in Siberia. Why should we ignore these papers, which included very coherent arguments, just because Jaska is here?

Are you implicitly suggesting that I'm actively advocating an appeal to authority on Jaska's behalf?

If that's the conclusion you'd reached from my post, I recommend a second read, and perhaps some reading around the topic of bias within research (if there aren't any validated measures to address inter-observer bias in a study, the resultant work is susceptible to bias).

Which leads us to an interesting question - Jaska, does linguistics routinely employ measures that attempt to control for inter-observer bias, or risk of bias between/across cited materials? Are your works concerning the pU urheimat effectively narrative reviews written by yourself, or is there a team with an inter-observer bias panel in play? Linguistics is an alien ecosystem to me, so the conventions aren't clear (there's some validation in exploratory in vitro studies, unsure what the analogue is in your world).

Most things in research are open to questioning and few lines of evidence (especially recent and consistent) should be ignored.

Also - This attempt to shift the trajectory away from the conceptual and towards one person (Jaska) isn't productive.

Best to focus on the evidence, ideally with the above framework in play.

[Generalissimo]

Were these correlations formally measured through formal statistical means, or merely described qualitatively?

This paper went to town with the correlation analyses. The authors tested all sorts of correlations between Y-DNA, autosomal components and languages.

https://genomebiology.biomedcentral....059-018-1522-1

But this paper, with its casual observations, is much better IMO, simply because it uses ancient DNA, archeology and linguistics to prove its point.

https://www.cell.com/current-biology...822(19)30424-5

In any case, there's now practically an academic consensus that the expansions of N-L1026, Nganasan-related ancestry and Uralic languages are linked, if not statistically correlated.

So Jaska doesn't represent the mainstream here.

This specific description of the recent events in this thread does not invalidate the general issue(s) that Jaska has rightly raised regarding the process of problem-solving in forums such as this from a "best practices" perspective.

I think the issues that Jaska has raised here aren't valid, and the reason he's raised them is because he's out of touch with the latest developments in this area. That is, he's wrong for the wrong reasons, not the right reasons.

Let's be honest, no one in this thread is advocating anything surprising that hasn't already been suggested in peer-reviewed literature, except of course Jaska. In regards to the problem-solving in forums such as this, well, that's a different issue and it can't be projected onto this thread and us.

Are you implicitly suggesting that I'm actively advocating an appeal to authority on Jaska's behalf?

Yes I am, although I don't know whether you did this subconsciously or not. Others have done the same here, and it's annoying.

Jaska has to be judged against the same standards as the rest of us. If he fails to explain his views properly and actively avoids answering difficult questions, then his posts aren't worth much no matter who he is.

Which leads us to an interesting question - Jaska, does linguistics routinely employ measures that attempt to control for inter-observer bias, or risk of bias between/across cited materials? Are your works concerning the pU urheimat effectively narrative reviews written by yourself, or is there a team with an inter-observer bias panel in play? Linguistics is an alien ecosystem to me, so the conventions aren't clear (there's some validation in exploratory in vitro studies, unsure what the analogue is in your world).

I don't really know how thorough Jaska's workflow is. If you say it's thorough then I'll take your word for it.

But I've got a hunch (actually, it's more than a hunch) that the N-L1026/kra001 = Proto-Uralic theory that I've been promoting in my world is heading for a major paper in a big journal, so let's wait and see.

Also - This attempt to shift the trajectory away from the conceptual and towards one person (Jaska) isn't productive.

Best to focus on the evidence, ideally with the above framework in play.

That's what I'm trying to do, but you put the focus back on Jaska.

I'm still waiting for Jaska to clear up some pertinent points, but apparently he's busy questioning our problem-solving workflow and being wrong, repeatedly, for the right reasons???

[Ryukendo]

Were these correlations formally measured through formal statistical means, or merely described qualitatively?

If the former, and if these two papers covered the appropriate timeframes and indirectly validated one another, and if there wasn't a methodological weakness to any underlying assumptions regarding their transformation of qualitative data to quantitative, then there isn't much more to discuss.

A rhetorical question, as I don't intend to get bogged-down in this thread.



This specific description of the recent events in this thread does not invalidate the general issue(s) that Jaska has rightly raised regarding the process of problem-solving in forums such as this from a "best practices" perspective.



To the underlined - Of course not. You'll find plenty of seemingly well-crafted STEM papers that, on closer inspection, contained some form of human error, methodological flaw or bias that does affect the results.

I've been involved in STEM peer review. Take it from me - Some highly questionable stuff from otherwise decent departments enter the process and get through.



Are you implicitly suggesting that I'm actively advocating an appeal to authority on Jaska's behalf?

If that's the conclusion you'd reached from my post, I recommend a second read, and perhaps some reading around the topic of bias within research (if there aren't any validated measures to address inter-observer bias in a study, the resultant work is susceptible to bias).

Which leads us to an interesting question - Jaska, does linguistics routinely employ measures that attempt to control for inter-observer bias, or risk of bias between/across cited materials? Are your works concerning the pU urheimat effectively narrative reviews written by yourself, or is there a team with an inter-observer bias panel in play? Linguistics is an alien ecosystem to me, so the conventions aren't clear (there's some validation in exploratory in vitro studies, unsure what the analogue is in your world).

Most things in research are open to questioning and few lines of evidence (especially recent and consistent) should be ignored.

Also - This attempt to shift the trajectory away from the conceptual and towards one person (Jaska) isn't productive.

Best to focus on the evidence, ideally with the above framework in play.

Very erudite discussion, just two quick points off the top of my head though:

Tambets et al 2018 actually did go into correlation analyses (I see Generalissimo has already posted it).

I really don't agree that science works in a hypothesis testing scenario, rather than a "Bayesian storm". In fact I would argue scientific discussion moves along lines very similar to how discussion on this forum works!--if much more professionally and less emotionally. If IE linguists had to piece together what happened in the past from fragmented linguistic materials and archaeological/ethnological correlates with only a hypothesis testing approach, they would begin by enumerating the hypothesis space fairly precisely (i.e. all possible hypotheses) and one need not go into the details of how this can happen to see how impractical that would be. The process has to be inductive, not deductive; David Anthony did not write his book by relying on hypothesis testing... The whole point of the Kuhnian understanding of what science is (one paradigm supported by evidence, then countervailing evidence builds up, until no extensions of the paradigm can explain things well anymore and we have to switch to another paradigm) is built out of Bayesian and not hypothesis-testing principles, and a computational cognitive scientist has come up with just such a Bayesian model e.g. here. In fact the entire basis for how computational cog sci explains the human capacity for adaptive theorization and changing inferences with evidence was built out of a Bayesian framework.

Suuper super OT... but I just can't agree that the crowning glories of science emerged out not from induction to parsimonious explanations through Bayesian principles, but from hypothesis testing(!) Could Darwin have come up with the theory of evolution using hypothesis testing?