Via bulbul’s Facebook feed, Ulrike Mosel’s Grammaticography: The art and craft of writing grammars (pdf). It’s an interesting overview of the topic; I’ll quote some salient bits:

Probably every grammarian has had the experience that collecting and analyzing language data and writing a grammar are two different things: you think your analysis is perfect, you know how the language works, you might even speak it fluently, but when it comes to writing up the grammar you are faced with unforeseen problems. How are you ever going to get all you know about the language into a single book? […]

Although collecting data, grammatical analysis, and writing up the chapters of a grammar are different tasks, they cannot be entirely separated, because once you start writing, you will discover gaps or inconsistencies so that you need to collect and analyze additional data. I often questioned my capacity as a fieldworker when I realized that my data were not sufficient. But now I think that the reason also lies in the very nature of writing, because – at least to some extent – the process of writing shapes and reshapes your thoughts which inevitably leads to changes in your analysis. […]

The low prestige of text editions (if they were more highly valued, linguists would no doubt publish more) can be attributed to several factors:
1. the politics of mainstream linguistics departments, some of which do not even recognize descriptive grammars as Ph.D. theses;
2. the fact that linguistic typology concentrates on the investigation of grammatical phenomena which manifest themselves in single sentences;
3. the fact that many typologists work with large samples of languages which does not allow the time consuming in-depth study of texts.

For scientific reasons, however, this relegation of texts to marginal appendices is not justifiable. […]

In a world where specialists – and linguists are no exception – know more and more about less and less, it becomes increasingly important to develop methodologies of making specialized knowledge accessible to non-specialists. Only the identification, analysis and description of the essentials of the structure of languages will enable us to connect specialized knowledge of various linguistic areas and advance our understanding of language. For this very reason the old tradition of grammar writing is gaining more importance than ever.

Makes sense to me.


  1. David Eddyshaw says

    Interesting throughout. Thanks.

    I was struck especially by the stuff about the Lingua questionnaire and the grammars that resulted from it (Croom Helm, originally.) I’ve always found them to be extremely user-unfriendly, and also often a great missed opportunity, in that they force languages into a Procrustean descriptive bed (a large Procrustean bed, admittedly) regardless of the idiosyncrasies of the particular language. This results all too often in section after section consisting of things like “[Language X] does not have [feature Y]”, on the one hand, and on the other, frustratingly disconnected and incomplete information about things like morphology. The better grammars in the series work around this with vast “appendices” to put all the stuff in that doesn’t go in the one(-big-)size-fits-all model.

    The Mouton Grammar Library series is a great improvement in general, precisely because it doesn’t impose a uniform model.

    To be fair, the Croom Helm works were really intended for typologically-orientated readers, so I shouldn’t moan that they don’t fit my preconceived ideas of what I want from a language description. And they’re a vast improvement on the painting-by-numbers tagmeme-orientated grammars that SIL used to produce in the Time of Ignorance. I’ve never come across one of those that wasn’t a right royal pain to extract any kind of useful idea of a language from at all.

  2. The SIL questionnaire has been found very useful by conlangers who want to make sure their language isn’t missing something huge and obvious, so not so much for description as for (sub-)creation.

  3. David Eddyshaw says

    MAK Halliday in his Introduction to Functional Grammar (from memory, as my copy has gone missing) talks about meeting a Chomsky chela who said something like “I assume the function of linguistics is to establish the structure of the human linguistic organ.” This prompts Halliday to produce a whole page’s worth of different potential functions of linguistics. I don’t recall conlanging being in there, but I am absolutely certain Halliday would have put it in if he’d thought of it.

  4. Very glad to hear it!

    (tl;dr: much talk about my own work, such as it is.)

    Having read the Mosel paper now, I see that I followed its prescriptions pret-ty closely in The Complete Lojban Language (a misleading title, for it had neither dictionary nor chrestomathy, though the need was recognized at the time). It is not an “I shall express the accusative case by a prefix!” grammar, for I had nothing to do with inventing the language.

    Instead, I read lots of grammar sketches and notes by creators and grilled them on the finer points insofar as they remembered them, which was in a sense rescue linguistics: “Thou met’st with things dying, I with things new-born.” I also did a fair amount of corpus work, as well as picking the brains of other speakers/writers. Lots of linguistics papers and some books, notably Larry Horn’s A Natural History of Negation. And of course, I consulted my own intuitions as well.

    I was also fortunate in a few other ways: I was able to use Li & Thompson’s Mandarin Chinese: A Functional Reference Grammar as a model. I was also able to incorporate a grammar sketch by one of the creators which helped solve the problem of the traditional descending order (phonology before morphology before syntax before discourse) that Mosel talks about, as well as a study of the semantics of compounding by Hattic-friend Nick Nicholas, a genuwine certified linguistician.

    Fortunately, there was already a well-established native linguistic terminology, which meant that I could finesse the noun/verb distinction, for Lojban has only verbs: a noun (other than a name) is simply an article followed by a verb and means “that which is/can be the subject of the verb”. Needless to say, “S = NP + VP” has no meaning whatsoever!

    Of course, though written in 1995-96, it is already somewhat out of date. My entire theory of the articles, namely that they mark not definiteness but specificity. is no longer applicable, for speakers have moved on to using the unspecific article as the general article almost at the expense of the specific article. (A reference is specific if only the speaker knows the reference for sure, whereas it is definite if the listener is expected to know the reference for sure.)

    Similarly, there was originally a prohibition on names (even foreign names) containing the syllables la, lai, doi which has been relaxed because it was too hard to use in real time, with the compensating provision of initial and final pauses to set them off.

    In addition, Nick wrote a paper in a refereed journal showing from corpus data that people were adhering to the original prescription that reflexives refer to the matrix rather than the current clause, though my grammar said otherwise. (That is, in the literal translation of “John said that James approved of himself”, the reflexive refers to John not James.) This is like Chinese but very unlike English and most other L1s and L2s of the people in the corpus.

    (Insert something to avoid a graceless close here; time is running out.)

  5. For David Eddyshaw on another thread I can’t locate:

    “Restrictive and non-restrictive, as commonly used, are notional rather than formal categories; this means, not that they should be discarded as useless or even dangerous, but that they should be used with caution and hedged around with due reservations. ”
    (Jakobsson, 1963)

Speak Your Mind