17 Oct 2010

The time-limit meme

Memiyawanzi recently quotes Jerzy Kuryłowicz from page 58 of The inflectional categories of Indo-European (1964): "One cannot reconstruct ad infinitum". No further commentary was added and so this hanging quote lured me to comment. Perhaps I read more into this than most but I notice a universe of issues in this deceptively simple quote. I'm just itching to elaborate.

Putting the quote in context

This quote showcases a certain irrational belief that has developed in modern comparative linguistics. From this belief arose a hypocritical attitude by many IEists towards more long-range theorists like Nostraticists. The reconstructions of long-range theorists deserve close criticism but it must be for the right reasons, otherwise we should dismount from our high-horse failing good methodology.

To avoid misinterpretation, we have to first understand that Kuryłowicz's above quote is followed by "We must be satisfied with stages bordering historical reality."[1] and we can see that he really is making a strong assertion about the existence of a definable time limit (ie. "historical reality") beyond which comparative reconstruction magically becomes useless. Neither the word "bordering" nor "historical reality" is of any informative value to the reader and nonetheless can't be rationally proven even if accurately defined. Heeding this advice, arbitrary limits would shift depending on the different possible meanings each individual might give to this vague notion.

Why this principle is irrational

Potential knowledge is infinite and it can be safely said that Kuryłowicz's mind, brilliant as it was, was not. So it should be clear here why it's invalid to make bold claims about what we may or may not know in the infinite span of future time. One might call it the Oracle Fallacy. We may say that living to 400 years old will always be impossible, for example, but such predictions are drawn out of thin air. Kuryłowicz's statement simply has no noteworthy informative value, no matter how we choose to interpret it.

If we can't interpret it in a logical way, it's doomed to be misinterpreted and this is precisely what has happened.

Everybody plays the meme game

Alexis Manaster Ramer sees past these fallacies in his article "Some uses and abuses of Mathematics in Linguistics" (1999) where he made a tragic list of otherwise educated linguists making arbitrary assertions about a temporal bound to valid reconstruction. Kleiner, Lewin and Nichols - all guilty of the same statements devoid of logical proof. He notes how Kleiner insists 10 000 years before present is the upper limit of valid reconstruction while Nichols waffles between 7000, 8000 and 10 000 BP. As we can see, not only are these empty opinions but they might even be said to stray from Kuryłowicz' "border of historical reality". Many would not agree that 10 000 BP is anywhere near the border of "historical reality" since the first attested writing goes back merely to approximately 3200 BCE (ie. circa 5200 BP).

The fear that holds us all back

A common fear seems to be that by dismissing this arbitrary time limit for comparative reconstruction, it now gives carte blanche to any loon with multiple pocket dictionaries and a dream to reconstruct what they like. However, Occam's Razor is already limiting such far-flung reconstructions, so this bogus time limit rule of Kuryłowicz is not just baseless but completely unnecessary.

A much-maligned theory like Proto-World fails not because it's a theory on an exceedingly old stage of human language well beyond "historical reality" (however that may be defined) but because the theory is a ridiculous lump of unverified and currently unverifiable assumptions that pales in comparison to less extravagant theories available. Occam's Razor alone beats the pulp out of these extravagant claims so such concerns are unwarranted.

The hypocrisy

I've read many a famed Indo-Europeanist talk a good game about methodology and will often attack more long-range theories like Proto-Nostratic because they are less secure in their argumentation. This is of course inevitably true because the further back in time one investigates, the more difficult it tends to be to verify etymological claims.

However, as my previous blog entries like Sowing wild oats and plowing the fields reveal, many dubious roots have been reconstructed in the name of Proto-Indo-European. So it's conversely not necessarily true that the validity of a reconstruction increases the closer it lies to the present. Adams and Mallory fill their books with a large amount of shaky roots based on unacceptably sparse or geographically biased cognate sets. If these authors need to especially label a root "Western IE" in distinction to PIE proper, one must wonder what methodological standards, if any, guide their decisions. How are these standards nobler than those of the more self-indulgent Nostraticists who have published?

A final plea to reason

The healthier way to reason is simply to evaluate each claim based directly on the application of Logic itself, not secondhand through unverified mantras. It takes more effort to painstakingly reason things through for oneself instead of somnulently following thoughtless bureaucratic policies set up by others.

[1] Schmalstieg, Indo-European linguistics: A new synthesis (1980), p.127 (see link).


  1. One claim I often run into, using the Australian languages as evidence, is that in hunter-gatherer societies historical linguistic evidence is wiped out by a turnover caused by word taboos and that in such societies areal borrowing is dominant (even with regards to morphology) and the notion of coherent language families is meaningless.

    I consider this notion to be complete nonsense.For one thing, there is no evidence for such extensive word tabooism among other hunter-gatherer societies.

    Secondly, typical patterns of language family expansion are quite visible in North America, even where the speakers are hunter-gatherers. In North America there is a distinct tendency for language families to originate in the west and expand eastwards.

  2. Yes, the sloppy "taboo" excuse is used a lot in Indo-European linguistics too.

    In Italy, PIE *wĺ̥kʷos 'wolf' undergoes irregular transformation to Latin lupus and... you guessed it... some linguists leap to a taboo solution whose details they invariably leave unexplained and unproved.

    Let's all agree that if you're going to claim that some word or name was altered by taboo, back it up with a decent explanation on the nature of the taboo and any relevant evidence supporting such a taboo. If no evidence is present, just call it an unexplained irregularity.

    "In North America there is a distinct tendency for language families to originate in the west and expand eastwards.

    Yes, this is seen by the unusually high density of cultural and linguistic diversity to the west coast of the continent compared with the more uniform east.

    However note too that Australia was first colonized by humans 60,000 years ago, much further back than North American colonization.

  3. I have read that the Proto-Germanic word for "bear" originated from a term meaning "the brown one", and that PIE *hrktos was lost because it was taboo. But I don't know if that is valid or BS.