Slashdot Mirror
Grammar Traces Language Roots
mlewan writes "Researchers use grammar to trace relations between Papuan languages. What is interesting is not that much that they use grammar features to do this, but that they seem to have given up using vocabulary as a help."
Use the same techniques to decipher Slashot headlines
Researchers use grammar to trace relations between Papuan languages. What is interesting is not that much that they use grammar features to do this, but that they seem to have given up using vocabulary as a help.
Help me take back Slashdot. When did 'News for Nerds' become 'FUD and Conspiracy Theories for Extremist Nutjobs'?
Even in England, different regions use different words and pronunciations (which could count as different words). But we all use the same grammar. It's easy to change the sounds of a sentence, but to change the structure requires hefty evolution, and hence a separation of culture.
Argh.
Why is there a story about grammar on a site whose editors can't understand the difference between "its" and "it's"?
The preceding message was based on actual events. Only the names, locations and events have been changed.
Actually, a linguist would tell you this is just the natural way language evolves. What you call a grammatical error in the second case, a linguist would call the creation and adoption of a new word to fill an important gap in the language. Previously, there was no simple way to communicate the concept of informal or inexact quotation - in other words, when you aren't quoting a person but paraphrasing their responses or words. It is quite awkward to say "He said something along the lines of..." or "He said something like..." repeatedly, so the phrase "He was like..." developed in response to a clear need in the language, which I think in part explains its rapid adoption around the US (though still stronger in certain regions, perhaps) and its stickiness as an informal usage.
The former example, take vs. bring, is a case in which a distinction between two similar words is so obscure as to be effectively meaningless for most communications. That one may be considered "proper" by a grammarian is irrelevant to a linguist if both easily communicate the same concept to a speaker of the language. This is one way in which language regularly evolves.
I appreciate that certain usages may sound grating on the ears to somebody who had that particular point beaten into their head as a schoolchild (i.e. take vs. bring). But this is part of the continuous process of linguistic evolution, NOT some sudden degradation in American English indicative of the downfall of our society.
The article reports that "the researchers made a database of 125 grammatical features in 15 Papuan languages. This included how word types, such as nouns and verbs, are ordered in a sentence, and whether nouns have a gender, as they do in languages such as German and French."
Unfortunately, these aren't reliable characteristics for determining language relatedness. For example, English and German are both undisputably West Germanic languages and are very closely related, having branched less than 2000 years ago. Nevertheless, German nouns have grammatical gender, while English nouns don't. German verbs come at the end of the clause (except in the main clause), while in English the placement of the verb is much more flexible but rarely at the end of the clause. Other examples could readily be given.
There is one, and only one, method for determining relatedness between languages which is generally accepted by specialists in the field: namely, by identifying a core of lexical and morphological items which show systematic correspondences in their sounds between languages (e.g. English father, fish, Latin pater, piscis), and which can't reasonably be attributed to borrowing or to chance.
Of course it would be nice if we could show relatedness between languages which branched further back than 10,000 years or so. Because of the way in which languages change, it's very unlikely that we'll ever be able to do so, at least if we are observing accepted standards of scientific rigor. Approaches roughly similar to the one described here have been attempted repeatedly in recent years, and have been repeatedly answered in the literature. You don't earn brownie points for sexing up an unreliable methodology by involving computers.
IAAPHCL (I am a professor of historical and comparative linguistics).
That Nature article is badly misleading in claiming that traditional historical linguistic methods are based on vocabulary and that it is an innovation to use grammar. It is true that amateurs' ideas about linguistic relationship are based almost entirely on vocabulary, but that isn't true of what professional historical linguists do.
To begin with, there are two different problems to be addressed. The first is, given a bunch of languages, are they related at all, where by "related", we mean "descended from a common ancestor". The second problem is, given that a bunch of languages are related, HOW are they related, that is, what is the family tree, in what order did they separate?
To determine whether languages are related, we look at "similarities". I put this in scare quotes because the relevant sorts of "similarities" are more accurately described as congruences, that is, systematic relationships between languages that may not necessarily be "similar" in the usual sense. For example, English and Armenian are distantly related members of the Indo-European language family. Proto-Indo-European *dw appears in English as /t/, as in "two", while in Armenian it appears as /erk/
as in /erku/ "two". Proto-Indo-European *dw -> Armenian erk is a regular sound change in that it happens in all of the attested cases in which that sequence of sounds is found. It is almost certainly the result of a series of less peculiar changes of which the intermediate stages happen not to be attested. The point is that this kind of systematic relationship is evidence of historical relationship between languages but is not a similarity in the usual sense.
Given some similarities or congruences between languages, the first question that arises is whether they might be due to chance. It is easy to find examples. For example, the Korean word for "language" is /mal/, as is the Icelandic word. There is no other reason to think that Korean and Icelandic are related, so this is written off as a coincidence. Amateurs tend not to realize how high the probability is of chance resemblences - there is a large crank literature in which people list words that they consider similar in sound and meaning in two languages and offer this as evidence of relationship.
One reason that historical linguists look for regular sound changes like Proto-Indo-European *dw -> Armenian erk, or less exotic, Proto-Indo-European *p -> English f (e.g. English "father", Latin "pater", Sanskrit "pitar") is that regular sound changes, which are reflected in regular sound correspondences among the daughter languages, greatly reduce the number of degrees of freedom and therefore provide evidence that the similarities observed are not merely coincidences.
A first point, then, is that even to the extent that historical linguists rely on vocabulary for establishing relationships, what they rely on are the regular sound correspondances, not raw similarities in words.
Now, given that we have reason to believe that there are similarities between two languages that are unlikely to be due to chance, we still have to determine their origin.One possibility is that they are due to common descent,in which case we have evidence of a genetic relationship. The alternative is that the similarities are due to diffusion. Diffusion can consist of outright borrowing, e.g. English acquiring karate from Japanese, or it can be less direct, e.g. Amharic and Tigrinya shifting away from the old Semitic verb-initial word order to verb-final word order under the influence of the neighboring languages in Ethiopia and Eriterea. The problem is, how can we tell whether a given similarity is due to genetic relationship or to diffusion?
The answer is, sometimes we can, but often it is hard, maybe even impossible. If you have multiple sets of regular sound correspondances, at most one of them can be genetic. The others must reflect borrowing. If the vocabulary that show
It's very interesting and much easier than English. When you take a closer look at it, it seems as someone really thought about it and removed every trace of difficult grammar. The one thing you have to learn is the vocabulary.
When I travelled to Jakarta (capital of Indonesia) the first time, I found out that that noone really speaks Indonesian there.
Bahasa Indonesia is a derivative of what we used to call Dienstmaleis ('service malay') in the Netherlands. This is the standardized language taught in the Netherlands to civil servants who were sent to the Netherlands Indies, and it is based on similarities between Malay dialects of Islamic merchants who travelled between the islands. It became the national language of Indonesia because it was the only language, besides Dutch, that the native civil service class on the islands shared with eachother. This precursor language has never been a living language; It was designed at the universities of the colonial oppressor. Indonesia doesn't like to acknowledge this, because these mythological Indonesian-speaking merchants who existed before the Dutch are central to the claim of being a historical 'nation'.