Friday, April 11, 2008

The Descent of Languages (Pseudolinguistics and Ideal Languages, Part 2)

This is part two in a series that began here with a general account of two common types of argumentation one encounters when dealing with pseudoscience. Before I begin debunking some specific historical linguistic accounts, I thought I might need to explain in some form what historical linguistics is all about (or at least, explain a little bit about some of what it's about...).

Since at least the days of Classical Greek grammarians, people have been aware of similarities between languages. Although others had attempted to systematically demonstrate the resemblance between two or more languages, Sir William Jones is most often credited with inventing the idea of using such comparisons to reconstruct a proto-language, a theoretical ancestor to modern languages known only through the evidence it left in its descendants.

All spoken languages change. One has only to listen to Beowulf or The Lord's Prayer translated into Old English to realize that the English spoken today hardly resembles the English spoken in the 10th or 11th Centuries. While change in language is often brushed off as corruption or degeneration, what is truly surprising about language change is how regular it can be, particularly when it is below the level of salience. We all note the introduction of words like truthiness or email, but few of us notice how the vowels in our speech subtly change decade to decade, generation to generation. The differences between languages that are related by a common ancestor are not entirely accidental: with careful study, one can rules that underly the relationship between words in each language that represent the changes those languages underwent since they split.

When the Roman empire fell, Latin speakers in France, Spain, Portugal and elsewhere were left disconnected from one another to varying degrees. Over time, the Latin spoken in these places diverged from the Latin that had been spoken in 5th Century Rome and earlier. While pure-blood Romans may have considered these divergences a result of the Celts and other inferior races attempting to speak the glorious Latin tongue, linguists can note a great deal of regularity in the changes that took place. That is, far from being chaotic or purely random, the changes follow rules. Consider the following sets of words.
Latin Italian Spanish Portuguese French gloss
capra capra cabra cabra chèvre goat'
/kapra/ /kabra/ /kabra/ /shevr/
caru caro caro caro cher dear'
/karo/ /karo/ /karu/ /sher/
caput capo cabo cabo chef head, top'
/kapo/ /kabo/ /kabu/ /shef/
caro/carn carne carne carne chair meat, flesh'
/karne/ /karne/ /karne/ /sher/
canis cane can cao chien dog'
/kane/ /kan/ /kaw/ /shien/
Forgetting the spellings and concentrating on the pronunciations (given in italics below each word), one can create a set of rules that produce the modern form from the Latin ancestor. For example, whereas all of the other languages maintained the /k/ sound at the start of each word, modern French has a sound like English /sh/. Likewise, where other languages have mostly maintained at least some vowel in the second syllable (even if changing /o/ to /e/), French has lost all of these vowels entirely. While the internal consonants of Italian are mostly faithful to the Latin original, Spanish and Portuguese have regularly changed the /p/ sounds to /b/. If you've never studied phonetics, try saying "puh buh puh buh..." aloud to yourself to see how similar these sounds are. (You might want to wait until no one's looking. Don't worry, all linguists have been there.) The difference lies in whether the vocal cords vibrate. Today, linguists know that in a context between two vowels or other voiced sounds, there is a tendency for sounds like /p/ to shift to their voiced counterpart over time.

In the same cases, the French has replaced /p/ with /v/ and /f/. If you take the time, you might try a similar experiment with alternating these sounds with /p/ and /b/ and noting the contexts where French uses one or the other.

Given access to modern Italian, Spanish, and French, a well-trained linguist with no knowledge of Latin could in time construct a large number of Latin words by identifying the similarities and differences between the modern words and using their knowledge of universal tendencies in human language. There would be gaps in their knowledge (the intricacies of the noun class and case system, unpreserved verbal declensions), but the amount that could be reconstructed would still be impressive.

This process, called the comparative method, is what we attribute in part to William Jones. European linguists of the 19th century were able to reconstruct many languages that were never written, including much of proto-Indo-European, the language believed to be ancestor of most European languages, Persian, and Sanskrit. In some ways, it is very similar to what geneticists can now do with genomes when analyzing the relationship between species (indeed, biological metaphors underly many discussions of language families). Changes can be analyzed in part because, like genes as a unit of biology, phonemes and other linguistic units behave in ways that we can approximate with models that assume a great deal of regularity or rule-like behavior between generations.

Like pure descent-with-modification in biology, there are many well-known complications to the comparative method. As their equivalent of viral or bacterial DNA in the human genome, languages are hosts to borrowings from other languages that can, over time, completely obscure that language's relationship to its genetic sisters. (If you've never noticed how many English words are derived from French, Greek, and Latin roots, you may want to open a dictionary and see for yourself how great a change borrowing can impose on a language.) Like biological species, when a language is spoken over a large area, changes can be slow to spread from one population to another. The great deal of modern sociolinguistics is focused on directly observing how changes spread across geographic and class boundaries in real time or by reconstructing past trends. More often than not, these exceptions prove the rule in that they affect languages in fairly predictable ways, even if the rules that govern them are more statistical and less definite than those identified in comparative reconstruction.

Although the degree of evidential and theoretical strength behind any reconstructed proto-language is always changing, as is true of all scientific statements, the trust linguists place in any model depends upon the methodology used to reach it. As I'll attempt to demonstrate in future installments, pseudolinguistic claims often attempt to borrow the look and feel of historical linguistic argumentation and the comparative method without actually borrowing its methodology.

No comments: