The Economist, while not a linguistics journal, does far better with language stories than the NY Times (which I have repeatedly lambasted in this space). Case in point: their Jan. 8 article on David Gil, a researcher at the Max Planck Institute for Evolutionary Anthropology in Leipzig (no credit given to the writer, unfortunately). The Times dealt with the issue last May, and I had to whack them again for their sloppiness; the Economist may not settle the question, but they report on Gil’s theories with intelligence and a basic grasp of what’s involved:

Dr Gil has been studying Riau for the past 12 years. Initially, he says, he struggled with the language, despite being fluent in standard Indonesian. However, a breakthrough came when he realised that what he had been thinking of as different parts of speech were, in fact, grammatically the same. For example, the phrase “the chicken is eating” translates into colloquial Riau as “ayam makan”. Literally, this is “chicken eat”. But the same pair of words also have meanings as diverse as “the chicken is making somebody eat”, or “somebody is eating where the chicken is”. There are, he says, no modifiers that distinguish the tenses of verbs. Nor are there modifiers for nouns that distinguish the definite from the indefinite (“the”, as opposed to “a”). Indeed, there are no features in Riau Indonesian that distinguish nouns from verbs. These categories, he says, are imposed because the languages that western linguists are familiar with have them.

This sort of observation flies in the face of conventional wisdom about what language is. Most linguists are influenced by the work of Noam Chomsky—in particular, his theory of “deep grammar”. According to Dr Chomsky, people are born with a sort of linguistic template in their brains. This is a set of rules that allows children to learn a language quickly, but also imposes constraints and structure on what is learnt. Evidence in support of this theory includes the tendency of children to make systematic mistakes which indicate a tendency to impose rules on what turn out to be grammatical exceptions (eg, “I dided it” instead of “I did it”). There is also the ability of the children of migrant workers to invent new languages known as creoles out of the grammatically incoherent pidgin spoken by their parents. Exactly what the deep grammar consists of is still not clear, but a basic distinction between nouns and verbs would probably be one of its minimum requirements.

Dr Gil contends, however, that there is a risk of unconscious bias leading to the conclusion that a particular sort of grammar exists in an unfamiliar language. That is because it is easier for linguists to discover extra features in foreign languages—for example tones that change the meaning of words, which are common in Indonesian but do not exist in European languages—than to realise that elements which are taken for granted in a linguist’s native language may be absent from another. Despite the best intentions, he says, there is a tendency to fit languages into a mould. And since most linguists are westerners, that mould is usually an Indo-European language from the West.

It need not, however, be a modern language. Dr Gil’s point about bias is well illustrated by the history of the study of the world’s most widely spoken tongue. Many of the people who developed modern linguistics had had an education in Latin and Greek. As a consequence, English was often described until well into the 20th century as having six different noun cases, because Latin has six….

The difficulty is compounded if a linguist is not fluent in the language he is studying. The process of linguistic fieldwork is a painstaking one, fraught with pitfalls. Its mainstay is the use of “informants” who tell linguists, in interviews and on paper, about their language. Unfortunately, these informants tend to be better-educated than their fellows, and are often fluent in more than one language. This, in conjunction with the comparatively formal setting of an interview (even if it is done in as basic a location as possible), can systematically distort the results. While such interviews are an unavoidable, and essential, part of the process, Dr Gil has also resorted to various ruses in his attempts to elicit linguistic information. In one of them, he would sit by the ferry terminal on Batam, an Indonesian island near Singapore, with sketches of fish doing different things. He then struck up conversations with shoeshine boys hanging around the dock, hoping that the boys would describe what the fish were doing in a relaxed, colloquial manner.

The experiment, though, was not entirely successful: when the boys realised his intention, they began to speak more formally. This experience, says Dr Gil, illustrates the difficulties of collecting authentic information about the ways in which people speak. But those differences, whether or not they reflect the absence of a Chomskian deep grammar, might be relevant not just to language, but to the very way in which people think.

The story goes on to describe an intriguing experiment Gil, in collaboration with Lera Boroditsky, a researcher at the Massachusetts Institute of Technology, is carrying out to test Sapir-Whorf; I’ll be interested to see what they come up with. (Via Eamonn Fitzgerald’s Rainy Day.)

Update. Mark Liberman has an interesting take on this over at Language Log.


  1. This may have something to do with Sapir-Whorf, but not much with Chomsky (or not). It sounds like an extreme form of what you see in Chinese, especially classical Chinese, where position in the sentence governs interpretation (instead of part-of-speech or markers like “the”.). Sounds like a fun language though.
    What little I ever knew of Malay seemed similiar — Riau must be an extreme case.

  2. Riau must be an extreme case
    Exactly. I’m actually more interested in learning more about Riau than in the theoretical implications (though, as you know, I’m always glad to see someone take a whack at Chomsky); I haven’t done much with Malayo-Polynesian (though I did write a paper on Toba Batak for my field method course in grad school), but I know enough to find this intriguing.

  3. I dunno. Something seems a little fishy about this to me:
    “For example, the phrase “the chicken is eating” translates into colloquial Riau as “ayam makan”. Literally, this is “chicken eat”. But the same pair of words also have meanings as diverse as “the chicken is making somebody eat”, or “somebody is eating where the chicken is”.”
    That seems a little hard to believe. You can say “The chicken is making somebody eat” just by saying “chicken eat.” ?? Just from “chicken eat” people would get the propositional content of “The chicken is making somebody eat.” ?? I find that very surprising, to say the least.
    It would seem to me that, if these claims are actually true, they would probably only be true within a very constrained discourse context. How could you get the sense of “Someone is eating where the chicken is” from just “chicken eat” without massive support from context? Outside of a lot of immediate context, you’d probably need lots more than just “chicken eat” to have people understand the propositional content of “Someone is eating where the chicken is.” I don’t see any way around this.
    “Deep grammar”? In the context in which this phrase appears, he seems to be referring to Universal Grammar. I thought he was referring to “Deep structure” at first, but that isn’t supported by the context.
    I found this more confusing than convincing.

  4. If Riau speakers really didn’t distinguish between nouns and verbs either semantically or morphologically, you’d expect a wider range of possible interpretations: ‘the food is fowl’, ‘the eater is a chicken (or maybe chickening out)’, and the like. In every interpretation cited, makan corresponds to a verb, and ayam corresponds to a noun. If Chomsky’s grammar errs by allowing impossibly abstract ‘deep’ structures, Gil errs in the opposite direction by disallowing even the minimal level of abstraction necessary to compare two languages with rather different morphological markings (or the lack of them). It looks like he’s going not just back to Whorfianism, but the extreme kind of structuralism that described each language in terms of numerically designated word classes to avoid any claim of cross-linguistic comparability. I really hate siding with Chomsky, but I’m at least 51% in his camp on this one.

  5. Ross Clark gives this article a much needed dose of the what for over at sci.lang. Most pertinently:

    I’ve heard Gil talk about Riau Indonesian a couple of times. He seemed to be trying to suggest that has no grammar at all — just words that can be strung together in any order you want. Unfortunately, since even a linguist talking to people bumps them into a higher and less grammar-less register, the only way you can experience the real RI seems to be to hang around down at the dock with David Gil. Which makes it hard to evaluate his hypothesis.

    I’m up for it, though, if anyone has any spare funding…

  6. > tones that change the meaning of words, which are common in Indonesian but do not exist in European languages
    Except, of course, Swedish.

  7. > You can say “The chicken is making somebody eat” just by saying “chicken eat.” ?? Just from “chicken eat” people would get the propositional content of “The chicken is making somebody eat.”
    What kind of sentence – or thought – is that anyway? “The chicken made Benjamin eat?” Can somebody explain the kind of action the chicken would perform to obtain this result?
    And what about “the chicken made Benjamin stop eating”? Or “the chicken made Banjamin continue eating”? “The chicken made Benjamin eat habitually”? “Ayam makan” too?

  8. The Sapir-Whorf hypothesis would suggest that this would actually effect the way people think about eating chickens. I see no evidence, in what’s given here, that the grammatical difference makes a difference in actual thought.

  9. If Riau speakers really didn’t distinguish between nouns and verbs either semantically or morphologically, you’d expect a wider range of possible interpretations: ‘the food is fowl’, ‘the eater is a chicken (or maybe chickening out)’, and the like.
    There’s very little to go on, here, but I don’t think that’s necessarily the case. First of all there’s no indication that the translations quoted exhaust all possibilities. Secondly, it’s possible that the language makes no lexical or morphological distinction between nouns and verbs, but can distinguish a predicate by word order – there is at least a case to be made that such a situation exists in e.g. Tagalog. I would find such an interpretation to be consistent with what Gil is quoted as saying in the article, although I’m not sure that the theoretical implications are really as great as is suggested.

  10. Gil is very much against the idea of intonation as a grammatical feature of Riau Indonesian. He has quite amusing comparisons of what the reverse would be for English.
    I can’t speak for his theory; I’ve heard him talk, and it wasn’t entirely convincing. It also doesn’t seem to work for other dialects of Indonesian.

  11. I haven’t got too much truck with linguistic universlism either, but I smell a fish here too. I did put forward – in my distant and misspent youth – a theory about Inuktitut grammar that involved abandoning traditional grammatical categories altogether, but even then there were still lexical categories and a grammar.
    My idea was that each fully inflected word in Inuktitut could be treated as a noun of sorts, and that when more than one word appeared in a sentence, the speaker was indicating that the the two nouns refer to the same thing. Unfortunately, my idea didn’t hold up too well with my fluent Inuktitut speaking prof – there was a fairly important class of lexemes I had neglected to think about. I still occasionally think the idea has a shred of merit to it. The most common class of morpho-lexical structure in Inuktitut is able to function as a noun in any context. Even when inflected, it is possible to just add more infections to it and make it part of a whole new word. But my case becomes more and more forced as I try to apply it to other features of the language.
    An example:
    Nunaqaqtunga Iqaluimmi.
    Land-have-I Iqaluit-LOC
    == “I live in Iqaluit.”
    I proposed to understand this sentence as:
    “A place where I have land” = “A place in Iqaluit”

  12. In classical Chinese the parts-of-speech ambiguity is always there. George Kennedy did a study putting various full words on a continuum from mostly-verb (90-10) to mostly-noun (10-90), whith plenty of words in the exact middle. However, sentence structure usually disambiguates things adequately, and authors who want to avoid ambiguity have resources to do so. As a matter of style, though, in poetry especially authors often choose ambiguous structure. (The Tso Chuan history is also written in an especially difficult gnomic style). But classical Chinese really is not exactly a natural language.
    Is this language actually a “dialect of Indonesian” or just a “language of Indonesia” spoken there. It seems off that the author might master Indonesian and then be baffled by a near relative. I do remember than Indonesian / Malay in fact did not distinguish N and V in any systematic way.

  13. Scott: Your idea is not so terribly off-key; I have seen a related language explained as each word being, in essence, an English sentence. Something like “There is land I have. It is in Iqualit.” More specifically this was done for subjects: “There is a man eating fish. He is John.” This was as part of a semantics course, from also a fluent speaker, but not a native one. I’m not sure how much was “this is how the language behaves” and “here is an illustrative example of one way out of many in which the language behaves”, though.
    Zizka: Even in English, most nouns can be (made into) verbs and vice versa. So there is some ambiguity there, mediated by strict word order.
    The ethnologue says that Riau is a dialect of Malay. Gil calls it Riau Indonesian. He wasn’t baffled by it in the “I don’t understand a word anyone’s saying” sense, but rather in the “This is so different from what I’ve been taught” sense. (This is from a conversation I had with him 2 years ago, so it may be somewhat warped.)
    Tim: I don’t think Tagalog works very well as being undifferentiated between A/N/V in the sense of it being unclear which they are in a sentence. Nouns do nouny things with determiners; verbs get all that weird inflectional stuff. (There is a nominalising prefix, and many verbs have nominal roots, but this is similar to English and other languages — Itelmen is like it, in that many roots are unspecified for type, but are clearly N or V within the sentence based on affixation and epenthesis patterns.) This is quite unlike Indonesian.

  14. Zizka: I remember one of my profs telling me that Classical Chinese is like Yoda writing telegrams – “Do not do no try.”
    wolfangel – “Words” like “Nunaqaqtunga” can certainly stand on their own as sentences. The difficulty arises in words like “Iqaluimmi” where sometimes you can treat them like full propositions and sometimes you can’t. The “classical” analysis is that structures like that are no more independent than saying “v Moskve” in Russian – the “-mi” suffix is no more than a case marker. This treatment usually analyses “nunaqaqtunga” as just a conjugated verb, where “-tunga” is the marker of the first person singular. This sentence would have the same structure as the Russian “zhivu v Moskve”.
    But this analysis performs very poorly. First, “nuna-” and “-qaq-” are morphologically productive, so this analysis completely neglects that whole dimension of morphology. The idea that “-tunga” is a marker of conjugation like the “-o” at the end of a verb in Spanish or “-u” in Russian makes some sense, since no proper noun can appear in that morphological slot.
    However, this doesn’t work all the time either. The first obvious example is that every composite word like “nunaqaqtunga” can instantly become a noun and be placed in the first position in a composite word.
    The only example I can reproduce from memory is the word “Inuktitut” itself, and I don’t trust my memory too well, so don’t use the example without checking it. The word glosses something like this:
    Man-“be spoken by”-it
    Trans: “It is spoken by people.”
    This is a whole proposition, but it is also the substantive noun for the language itself. So we can make a sentence like:
    Trans: I speak Inuktitut
    (I think “-niq-” is the verbal morpheme for “to speak”, but don’t trust me on it.)
    Or we could gloss it as a subclause:
    Man-“be spoken by”-it-speak-I
    Trans: I speak the language that is spoken by [the] people.
    This is the major form of lexical creativity in Inuktitut. The problem is that while you can turn any proposition into a noun morpheme, you can’t do that for verbal morphemes and the final position is always reserved for the pronoun/conjugation marker. The other problem is that sometimes you can turn a word like “Iqaluimmi” into a noun morpheme, but most of the time you can’t. I can’t remember any examples to clarify it. At any rate, this made a mess of my hypothesis.
    I think you can make a good case that Inuktitut does not fit Latin-driven lexical categories very well and that some out-of-the-box thinking is required. One of the million or so projects I’ve long thought might be worth doing if I had a free decade would be to take a closer look at propositional structure in Inuktitut to see if I couldn’t find a set of grammatical categories better suited to the language.

  15. David Cerezo says

    I’m a computer programmer recently interested in linguistics and the Sapir-Whorf thesis.
    Any webpage listing curiosities about languages like Riau? Or a survey on how languages influence thoughts?
    Thank you!

  16. There seem to be three related but distinct points in this article.
    1) Riau Indonesia doesn’t seem to have lexical categories that correspond closely to “noun” or “verb”, even in the broad sense. While this isn’t a direct assault on Chomsky, it does have implications; the wider your lexical categories, the shakier he looks.
    (Of course, a devout Chomskyite could just claim that Riau is an odd exception of no particular importance, like that South American Indian language that uses object-verb-subject word order; a peculiarity where the “switches” happened to get flipped to a statistically unusual position. But the more anomalies like this pop up, the more it all starts to look like epicycles.)
    2) Methodological problem: it’s harder to get a good handle on an obscure language _as it’s actually spoken_ than is generally realized. This is because (2a) interlocutors are often “infected” by the speech patterns and categories of other languages, and (2b) everybody starts talking more formally whan dealing with strangely dressed foreigners who are asking questions about language.
    Neither of these are /new/ points, but arguably they’ve not been given enough weight, especially with difficult and obscure languages spoken by small groups of people… IOW, the clear numerical majority of the world’s languages.
    3) Gil’s perceptual experiment, which is not included in the cited excerpt.
    IMS it went something like this: Gil and Iela showed subjects three pictures. One showed a man about to kick a ball. The second showed a man kicking a ball. The third showed a /different/ man about to kick a ball. Subjects were asked which two of the pictures belonged together.
    Native English speakers tended to say the first two — same man completing an action. Native Indonesian speakers, though, usually grouped the first with the third — same /state/ (man about to kick a ball). And native Indonesian speakers who were fluent in English tended to fall statistically in between the two.
    It’s hardly a clincher for strong Sapir-Whorf; but it’s interesting.
    Doug M.

  17. Doug: That’s my take on it as well. (And I love the Chomskyan response to exceptions to their “universals” — epicycles indeed!)

  18. I wish I had gone farther in Mongol studies. Mongol agglutination can change a verb to a noun with a suffix and then change the new word back to a noun with a new suffix. “Deverbalization” and “denominalization” are routine processes. Each new word picks up particular meanings though. Somewhat the way “sidewalk” means only one thing, and not really any the various other things that it might mean such as catwalk, side-trip, sidle, etc. (Though in fact, those usages might well be understood).
    From looking at dictionaries and various other aids it seemed that all but a few suffixes and suffix-clusters were very rarely used and that there were a lot of blank spaces. I wanted to learn the language and ask people what analogical made-up words not in the dictionary meant, to see if they were able to construe them at all, much less give meanings to them. Rather a fanciful plan, since many or most Mongol words are not actually pronounced in a way to make the agglutination audible.
    This of course is an opposite extreme from Malay and only relevant in that sense.

  19. Those surprising glosses of ayam makan, are they as a complete sentence, or are they in context? A context would be interesting, but a lot more understandable. If for example we had
    ayam makan Hasan nasi = ‘the chicken feeds Hasan rice’
    ayam makan Hasan duduk = ‘(where) the chicken eats, Hasan is sitting’.
    Note for those who don’t know any Malay/Indonesian, the standard language is very thinly inflected too. makan = ‘eats, is eating, ate, will eat’ etc., which can be made explicit by adverbials akan ‘will’, sudah ‘already’ etc.; and nouns have no articles, so ayam = ‘(the/a) chicken’, which can be optionally marked with a demonstrative: ayam itu = ‘that/the chicken. So ayam makan is perfectly good standard Malay/Indonesian.

  20. wolfangel – I think you’re certainly right that there’s no confusion as to what role a word is playing in a Tagalog sentence. What Himmelmann’s saying in the paper I linked to above is that there are five basic syntactic functions in Tagalog[1], and that any root can occur in any of these positions. Furthermore, that the various morphological operations which can be performed on roots, while derivational in nature, do not alter the distribution of the word syntactically, they don’t change its part of speech. This includes the voice affixes and accompanying inflection for mood and aspect. What it looks like he’s saying to me is that all Tagalog predicates are similar to what you’d call nominal predicates in a language with a noun-verb distinction, and the voice affixes are something like nominalizers deriving nouns from the various roles of a verbal root.
    Now, I’ve no idea whether he’s right. I don’t speak Tagalog, and my acquaintanceship with Austronesian linguistics is of the most amateur kind. All I’m saying is that, based on the very little we know from the article, it’s possible that when Gil says (something paraphrased as) “there are no features in Riau Indonesian that distinguish nouns from verbs”, he’s claiming something similar to what Himmelmann’s claiming about Tagalog. In which case he wouldn’t be denying that it’s possible to distinguish different syntactic roles in the Riau sentence.
    [1] Predicate, ang phrase (subject), ng phrase (genitive, and non-subject core arguments), sa phrase (locative), and modifier.

  21. I want to know about possesive pronoun in Batak language especially in Simalungunese but in English
    Thang you

  22. If anyone’s still interested in this thread, there’s a prepublication draft of a paper by Gil on Riau here:
    Word Order Without Syntactic Categories: How Riau Indonesian Does It
    As you might imagine, this supplies a more precise statement of what Gil is claiming about Riau than the Economist article.

  23. Wow, thanks!

  24. Charles Hartman says

    (sorry, this is from months ago, Jan 13, by zizka)
    “Mongol agglutination can change a verb to a noun with a suffix and then change the new word back to a noun with a new suffix. “Deverbalization” and “denominalization” are routine processes.”
    I’m no linguist, but isn’t “deverbalization” an English example of exactly that sequence of changes?

Speak Your Mind