Combining historical linguistics with mathematics, glottochronology is the controversial method to determine the elapsed time since the divergence of two languages; it is the lingual equivalent of carbon dating, though, for reasons that will be explained, it has received far less support. It was proposed by Robert Lees and Morris Swadesh in the late 1940s, and though it is accurate given known examples (It is based upon the rough elapsed time since the divergence of the sister Romance languages from Latin.) it relies on several somewhat dubious presumptions and is quite difficult to prove correct without outside, independently verified numbers.
Firstly, though, glottochronology works by comparing basic vocabularies. A basic vocabulary is a list of “universal” words, all colors, body parts, geographical features, and other concepts that are present (hopefully) in all languages. Obviously, cultural and geographical bias must be avoided; a basic vocabulary containing words for “coconut” and “television” would not be of much use in comparing medieval Inuit tongues. In addition, the genetic history of the languages to be compared must be known and verified; they must have a common mother tongue, or any results obtained are bunk. Any two languages are statistically bound to have at least a handful of cognates, and this guaranteed amount of similarity will only return nonsense if the languages aren’t related; in short, just make sure that any cognates discovered are not coincidental.
With this basic vocabulary and verified ancestry, the number of cognates is determined by comparing equivalent terms, allowing for phonetic change: Portuguese “pai” and Italian “padre,” both meaning “father,” are allowed as cognates because of explainable phonetic drift. This process of comparing vocabularies is called lexicostatistics. From the number of cognates the percentage of similarity is easily calculated, and from there it is a simple matter of plugging numbers into the equation t = (log c) / (2 log r), where t is the time of separation, c the percentage of similarity, and r the glottochronological constant. This constant is usually 86%, from the theory that after 1,000 years of separation two languages will still share 86% of a core 100 words from the basic vocabulary. All this results in a simple correlation: the greater difference between two languages, the greater the elapsed time since their divergence.
Detractors point out that it is nearly impossible to construct a universal basic vocabulary without any bias whatsoever. One could easily construct a list for every pair of languages to be analyzed, but glottochronology relies on the glottochronological constant, itself based on a single universal basic vocabulary. These words must be universal; if a language, for whatever reason, loses even one or two words over the years, the margin of error will be in the centuries. Changing the given basic vocabulary makes the calculated 86% figure rather arbitrary, for what guarantees that two different, constructed lists are both comprised of concepts that will persist? Swadesh produced at least two commonly used lists, one of 215 words and the other of 100. The latter is provided below:
I, me you we this that
who what not all many
one two big long small
woman man person fish bird
dog louse tree seed leaf
root bark skin flesh blood
bone grease egg horn tail
feather hair head ear eye
nose mouth tongue tooth claw
foot knee hand belly neck
breast(s) heart liver to eat to drink
to bite to see to hear to know to sleep
to die to kill to swim to fly to walk
to lie to come to sit to stand to say
sun moon star water rain
stone sand earth cloud smoke
fire ash burn path mountain
red green yellow white black
night hot cold full new
good round dry name to give
Of course, it could well be that the glottochronological constant is arbitrary in the first place, as there is no evidence to prove that the rate of language attrition is constant across all languages at all times; in fact, there is ample evidence to prove that the rate of attrition is not even constant in one language over a short time frame, not to mention a number of sister languages over several millennia! Additionally, the current glottochronological constant was derived from studying Indo-European languages. Should Amerindian languages use the same constant? How does glottochronology take into account languages like Korean, which replaced almost the entire basic vocabulary over a course of 500 years?
The unfortunate truth is that there are simply too many variables to consider; glottochronology is doomed to failure because of the wild variations in human language, not to mention ever-fluctuating rates of change, vocabulary replacement, and innumerable other events and processes that will throw off the resulting figure. Still, however, there are linguists that cling to it despite the mountain of contradictory evidence; the truth be told, it is an elegant little theory, and for some people that's evidently enough.