LINGUIST List 3.853

Sat 31 Oct 1992

Disc: Probabilistic Reasoning in Linguistics

Editor for this issue: <>

Directory

1. , Scientific American

Am I the only one who is appalled to see the abuse of probabilistic
reasoning which Greenberg and Ruhlen resort to in their article
in November Sc. Am.? They show that two languages, Halkomelem
and Tfalkit (or something like that) have similar words with
a similar meaning (just one word per language), calculate the
probability of a pair like occurring in a random pair of languages,
and say, My God, it's got to be common origin. This utter nonsense,
since all the calculation shows is that you should not expect to
find the same in every (or even in many) pairs of languages, but
there is nothing wrong with it happening in a particular pair.
It's like saying that Columbus and Leif Eriksson must have been
working together, because it is just too unlikely that they could
have discovered the same continent by chance.

Incidentally, this kind of reasoning seems to recur fairly often
in various sciences (I recall some papers by Guiora on the so-
called Sapir-Whorf hypothesis a few years ago that did something
similar). Is there any literature on such fallacies?

(Linguistically, of course, the best way to see this is a fallacy
is to see that similar forms with similar meanings are not all
uncommon in situations where we KNOW the origin is not the same,
e.g., English 'ear (of corn)' vs. 'ear', German 'Braten' (noun)
vs. 'braten' (verb), and so on. If we can have coincidences like
this in a single language, we can certainly have them across
a pair of languages.)