Joel of Far Outliers has a post that begins with his student years:

During my dissertation fieldwork in Papua New Guinea over thirty years ago, I discovered that a bunch of Austronesian languages in Morobe Province mark their relative clauses in a manner that is pretty rare from a typological point of view: they mark both the beginning and the end of the clauses. An English equivalent would go something like, “The language [that they were speaking that] sounded vaguely familiar,” or “The language [which they were speaking such] sounded vaguely familiar.”
The only other place where I could find languages that did the same was in Central Africa, and my dissertation cited a 1976 article by the great French linguist Claude Hagège which mentioned by name two Nilo-Saharan languages, Moru and Mangbetu, and two Niger-Congo languages, Mbum and M’baka. Over the years, I lost track of anything pertaining to those languages except their names.

But recently his interest revived and he thought to ask his brother, who “had spent years working in the (at that time) Central African Empire for the US Peace Corps and USAID while I was writing my dissertation in linguistics,” and his brother asked “his linguist friend Raymond Boyd at CNRS whether he could think of Adamawa-Ubangi languages that used such markers for relative clauses,” and Boyd responded “Right off, I can’t think of one that DOESN’T.” Read Joel’s post for examples of this interesting phenomenon.


  1. Guin,

  2. I promise to keep conlang details to a minimum in my comments here. This is a feature of Lojban, too. Although almost every structure in Lojban is similarly bookended. So while it has copied features from many languages, this one is, I think, derived from a desire for easy machine parseability, & the resultant similarity to natural languages is coincidental.

  3. I agree that it’s coincidental. Note, however, that the great bulk of these “elidable terminators” can be and are elided in actual use: whenever the next word cannot grammatically appear in the construct being terminated, the terminator can be dropped.
    For example, in the full sentence le nanmu ku cu dunda lo rozgu ku le ninmu ku ‘The man gave a rose to the woman’, each instance of ku (which terminates, among other things, NPs beginning with articles) can be safely elided, because in each case the next word cannot be part of the noun-phrase: it is either an article (le, lo) or the predicate marker cu, producing the less verbose le nanmu cu dunda lo rozgu le ninmu’, which is no longer than the English.

  4. John’s comment gets me to wondering whether any of the bracketers Joel exemplified can be elided in some contexts.

  5. Yes. Most of my PNG sources say that the final bracket is often elided when it would fall at the end of the sentence. But otherwise the clause markers seem not to be elided as much as they are in English. These are all verb-serializing languages, so maybe the markers are more crucial than they might be in a language where only one verb is expected to occur in each functional clause.

  6. Lojban is also verb serializing.

  7. Reminds me very much of possessive and object pronouns in Welsh, too. For instance,
    “dy gath wen di” can be glossed as
    “your white cat you”
    In this case also the back end of the possessive can be omitted = “dy gath wen”

Speak Your Mind