The ‘Myth’ of Language History.

This Phys.org story reports on a finding that’s surprising if true:

The ‘myth’ of language history: languages do not share a single history but different components evolve along different trajectories and at different rates. A large-scale study of Pacific languages reveals that forces driving grammatical change are different to those driving lexical change. Grammar changes more rapidly and is especially influenced by contact with unrelated languages, while words are more resistant to change.

An international team of researchers, led by scientists at the Max Planck Institute for the Science of Human History, have discovered that a language’s grammatical structures change more quickly over time than vocabulary, overturning a long-held assumption in the field. The study, published October 2 in PNAS, analyzed 81 Austronesian languages based on a detailed database of grammatical structures and lexicon. By analyzing these languages, all from a single family and geographic region, using sophisticated modelling the researchers were able to determine how quickly different aspects of the languages had changed. Strikingly different processes seemed to be shaping the lexicon and the grammar – the lexicon changed more when new languages were created, while the grammatical structures were more affected by contact with other languages.

The paper is Simon J. Greenhill el al., “Evolutionary dynamics of language systems,” PNAS (2017). As always, I welcome all thoughts on the topic. (Thanks, Trevor!)

Comments

  1. Off-topic: “forces driving grammatical change are different to those driving lexical change” — is “different to”, rather than “different from”, newly popular in standard written English?

  2. Eli Nelson says:

    No; I think “different to” has been common for at least a century in British English

  3. Different to never spread to AmE, which suggests that it is a post-1800 innovation.

  4. Still, it’s been 200 years. PNAS says it’s time to pick up the pace, AmE. This is why everyone thinks you still speak Elizabethan in Virginia.

  5. Marja Erwin says:

    As we all know, dystopian and post-apocalyptic societies have the fastest language change. (Or fiction set in dystopian and post-apocalyptic societies is most likely to have language modifications to emphasize its setting.) So we may see faster language change.

  6. As to the paper itself, it calculates rates of change in vocabulary and grammar among Austronesian languages. Some of the results are expected, but it’s nice to see them confirmed: there are more stable and less stable components of grammar and vocabulary, and languages which are conservative in one may be innovative in the other. One of their points is kind of meaningless, though: they claim that grammar changes faster than vocabulary, but you can’t really compare the two. The “rate of change in grammar” is based on a particular set of features, selected to be comprehensive. The rate of change in vocabulary is based on a set of words selected because they are stable. In any case, you can’t really select a set of grammatical features which is equivalent to the Swadesh list or whatever, in the sense of evolving equally rapidly under some idealized conditions. It’s really a case of apples and oranges.

  7. “long-held assumption”? Only if you equivocate on the meaning of “grammatical”. The kind of morphological similarities that are traditionally weighted highest combine structural and phonological material, and don’t look much like the purely structural typological properties they’re classing as “grammatical” at all. The idea that those are uniquely stable hardly goes back further than Johanna Nichols, and emerged partly in reaction against Greenberg’s over-ambitious attempts at lexical mass comparison.

  8. In my research (on Japanese) our assumption is that grammatical patterns spread easily and pitch accent is preserved for longer, which is the motivation for us to study pitch accent in the first place… (there’s a bunch of geographical patterns and historical documents that fit this assumption, like contact zones where people seem to have substituted their inflections and lexicon towards a prestige standard, but kept the pitch patterns unchanged).

    According to Oxford “different from” is the most common, “different than” is American and “different to” British. I’m always confused by this so I’ll try to remember to pick “from”.

  9. Isn’t the idea of Sprachbünde exactly that the “soul” (phonology, syntax, semantics and a bunch of legitimately Whorfian aspects) of a language is much more easily areally spread than the “body” (forms of words and morphology), which tends to go by the genetic route? (with reference to Dawkins on Cappadocian Greek)

  10. David Eddyshaw says:

    One of Johanna Nichols’ supposedly stable grammatical features is Absolutive/Ergative alignment versus Nominative/Accusative, which is spectacularly not stable in Polynesian; but then I believe Georgian and Mingrelian also differ in alignment, and you’ve got lots of cases (e.g. in India) of past tenses going ergative in honour of their origins.

    In fact, looking at her maps of “stable” features provokes one immediately into coming up with counterexamples.

    That most Greenbergian of syntactic features, SVO versus SOV word order, is highly unstable diachronically and seems very prone to Sprachbund effects, as with Mandinka/Bambara/Dyula, Songhay and Senoufo, which not even the most enthusiastic of lumpers thinks are closely related to each other. Very closely related Songhay languages, come to that, have different “basic word orders” from each other.

    The paper does, to be fair, say that they found word order features to be among the most labile; but then, what else is new?

    It occurs to me too that Proto-Austronesian presumably had something like the weird Philippines-style alignment system, which is surely going to be particularly liable to change in unusual ways into one or another more common system in different daughter languages. I mean, if the ancestral language was typologically unusual syntactically, wouldn’t you expect more than the usual rate of changes in the descendants?

  11. David Eddyshaw says:

    Another factor which comes to mind as potentially making Austronesian an unreliable test-bed for these matters is that a good many Austronesian languages have been clearly strongly influenced, as substrates or by contact, by “Papuan” languages, which are all over the place syntactically, and themselves belong to a great many different unrelated families.

  12. Indeed, I often see Philippine alignment referred to as “Austronesian alignment”, implying that it is a primitive feature.

    Different than is often disparaged as an Americanism, but as the OED says, it has a long history in BrE. It probably arose from sentences like “We make use of them in a quite different manner than we did in the beginning” (1644), a straightforward comparative construction of inequality which was then generalized to the use of than after predicative different. The OED’s citations also show different with and different against, which are now obsolete. So sure, stick with different from, which is universally acceptable.

    Ergative alignment seems to be stable, whereas accusative alignment is not; on the other hand, accusative alignment is both the default and metastable. Most languages are accusative in all their constructions, but some have ergative alignment in the preterite or the perfect(ive) — “split ergativity” is a misnomer, as few languages have ergative alignment everywhere. When a new construction such as the progressive arises in an ergative language, it tends to be accusative, leaving existing ergative constructions untouched.

  13. One of their points is kind of meaningless, though: they claim that grammar changes faster than vocabulary, but you can’t really compare the two.

    That was my reaction as well.

  14. I should probably preface this comment with the fact that I am not a linguist, but I am a lover of languages. I have read in several books on linguistics for the layman that grammar and vocabulary are intertwined to some extent. “Vocabulary” words turn into grammar words. Take the example of the word “like”, as in “I am like so stressed”. How does this compute into the equation? Food for thought.

  15. ““Vocabulary” words turn into grammar words.”

    It’s called grammaticalization. And example is the way “fixing to” has evolved into a marker of future reference.

    Mandarin Chinese is full of this. A large percentage of its “function words” (grammatical markers) can be identified with actual “vocabulary” words still active in the language.

  16. Reed James: That’s called grammaticalization, and it is a very important way in which languages develop their syntactic constructions. A good example is the English word will, which starts out as a main verb meaning ‘intend’, as in “This has been willed where what is willed, must be”. It then develops into an auxiliary verb showing an intention (“I will not go to Canossa”) and then as a marker of the future (“He will be President until 2021”). Similarly, the French future form chanterai ‘I will sing’ does not descend directly from its Latin equivalent cantabo, but rather from a late Latin cantare habeo ‘I have to sing’, whereby the original main verb has now become an inflectional ending.

  17. David Marjanović says:

    I haven’t yet read the paper, but…

    the lexicon changed more when new languages were created

    …does this mean something?

    I believe Georgian and Mingrelian also differ in alignment

    Yes, though only in some kind of past tense. (I happen to have read this paper, which mentions this, just a few days ago.)

  18. David Eddyshaw says:

    the French future form chanterai ‘I will sing’ does not descend directly from its Latin equivalent cantabo, but rather from a late Latin cantare habeo ‘I have to sing’, whereby the original main verb has now become an inflectional ending.

    Reminds me of something else which is at least potentially a problem with the paper.
    It’s not at all unusual for languages to change how a category (like “future”, which is in fact one of their “grammar” criteria) is expressed, while keeping the actual categories, which may themselves be far from linguistic universals.

    A classic example is Polotsky’s study of Coptic second tenses, which led to major insights into the verbal system of prior stages of Egyptian, in which similar distinctions occur but are actually expressed very differently (basically by verb flexion rather than auxiliary verb choice.)

    So you can’t always tell how stable grammatical features are, without a sufficiently profound analysis of the languages involved. I don’t know anything about Austronesian languages, but if the position with adequate grammatical studies is anything like as patchy as in Africa, I would imagine trawling through grammatical sketches to find out if they have a “future” is not going to be a very solid way of proceeding.

    Didn’t we just have a vigorous discussion about whether some fairly well described language actually has a future tense? I seem to recall that the matter was not regarded by all as absolutely settled …

  19. English, in the Learning Minority Languages thread.
    What I learned from the WALS chapter on future marking is that it’s quite common for the category of future to be fuzzier than other time categories, and that it’s often hard to say whether it’s better to describe it as a tense or as a mood.

  20. On “Philippine alignment”: yes, it’s generally reconstructed for Proto-Austronesian, because it’s found, not only in the Philippines, but in the Austronesian languages of Taiwan, and in Malagasy, over on Madagascar (and there is something that’s arguably descended from it in Indonesian, as well). Proto-Austronesian is generally held to have started out on Taiwan, so the system is ascribed to the proto-language.

  21. (Thought I posted this a few hours ago, but can’t see it in the comments, so I’m giving it another try)

    the French future form chanterai ‘I will sing’ does not descend directly from its Latin equivalent cantabo, but rather from a late Latin cantare habeo ‘I have to sing’, whereby the original main verb has now become an inflectional ending.

    Even more clearly so in Spanish, where future indicative forms are transparently INF+the inflected form of modal haber ‘have to’: cantaré=cantar+he; morirás=morir+has. Occasionally apocope confuses things slightly (e.g., podréis=poder+obsolete heis), but still.

  22. Even more clearly so in Spanish, where future indicative forms are transparently INF+the inflected form of modal haber ‘have to’

    Something like that has happened in Ukranian as well:

    Unlike all other Slavic languages, Ukrainian has a synthetic future (also termed inflectional future) tense which developed through the erosion and cliticization of the verb ‘to have’ (or possibly ‘to take’): pysa-ty-mu (infinitive-future-1st sg.) I will write. Although the inflectional future (based on the verb ‘to have’) is characteristic of Romance languages, Ukrainian linguist A. Danylenko argues that Ukrainian differs from Romance in the choice of auxiliary, which should be interpreted as ‘to take’ and not ‘to have.’ He states that Late Common Slavic (LCS) had three verbs with the same root *em- :

    • a determined imperfective LCS *jęti : *jĭmǫ ‘to take’ (later superseded by numerous prefixed perfectives)
    • an indetermined imperfective LCS *jĭmati : jemljǫ ‘to take’ (which would not take any prefixes)
    • an imperfective LCS *jĭměti : *jĭmamĭ ‘to hold, own, have’

    The three verbs became conflated in East Slavic due to morphological overlap, in particular of iměti ‘to have’ and jati ‘to take’ as exemplified in the Middle Ukrainian homonymic imut’ from both iměti (< *jĭměti) and jati (< *jęti).

    https://en.wikipedia.org/wiki/Ukrainian_language

Speak Your Mind

*