Glossing

I have to bite the bullet now and decide how to interlinearise my Bardi examples in the grammar. Most of my corpus isn’t interlinearised. I don’t need it, and I’ve been avoiding this for a while because there’s no good solution that I can see. These are the issues:

  • rampant post-affixal morphophonology (harmony, epenthesis, etc) which obscures prefix boundaries and which makes segmentation arbitrary
  • prefix-root interactions
  • long-distance allomorphic conditioning
  • null roots
  • secondary morphological exponence

First up in this list is vowel harmony. It’s not nice Turkish-like harmony, it’s a messy sytem that’s got a bunch of unprincipled exceptions. inoonggoonboon shows it, but inamboona doesn’t, for example. In the Laves collection, there’s ingarroogoolina ‘they broke it’, Modern Bardi ingoorroogooloona. The harmony affects epenthetic vowels too. Epenthetic vowels trigger deletion of certain segments (there was a historical sound change that deleted a number of intervocalic consonants). One of the deleted consonants was /w/, and it happens that the Proto-Nyulnyulan for ‘give’ was *wa. In Bardi this ends up as a null verb root. In other cases, the epenthesis is the only relic of the morpheme (that’s what I mean by secondary exponence here). inamarran is ‘(s)he was cooking it’ and inmarran is ‘(s)he’s cooking it’. The a in the past form is probably epenthetic. In a verb like ingoorrooloonganirr, we could divide the form i-ng-oorr-oo-loonga-n=irr, i-ng-oorr-ooloonga-n=irr, i-ng-oorr-oo-loong-an=irr, and probably other divisions too. The root is underlyingly –jooloonga-, with the first /u/ conditioning harmony and the /j/ deleting following the transitivity marker a-.

All these difficulties in segmentation might lead us to give schematic glossing only. Ingoorrooloongan=irr could be glossed as ‘collect.3augS.pst.tr-3augO’. But to do that would be to deny the pretty obvious segmentation for a large part of the lexicon. Inyjalgoon ‘(s)he fell’, for example, is straightforwardly i-ny-jalgoo-n. Ilamanka ‘3sg/3pl is listening’ is ambiguously 3sg or 3pl, and the sound rule that produces that is regular and found elsewhere. The segmentation problems are localised, but they are fairly frequent. So, not to segment (or to show only major morpheme boundaries) would be to overlook what is a pretty regular system overall.

One solution is to ‘unpack’ the morphemes a bit. For ingoorrooloonganirr, the verb could be glossed as follows:

i- ng- arr- a- [joo]loong[a] -an =irr
3- PAST- AUGMENT- TRANS- collect -CONT =3AUG.DO

That would make it easier to analyse the morphemes (which is, after all, the point of interlinearisation) but does abstract away from the surface form. That doesn’t worry me too much for verbs like this.

I did this in my PhD, and it produced unreadable sentences. It crams the verbs full of abbreviations and makes the grammar pretty hard to use. It also gets a bit abstract for some other verbs, and it raises questions about how much to ‘undo’. For instance, the /ŋ/ past tense marker assimilates to a following stop (cf inyjalgoon above); should that be represented with ng? How about the lenition of b > w in iwirrilin 3sg-flying? Should I just add back deleted segments and leave the rest? That would make my example verb look like this:

i- ng- oorr- oo- [joo]loong[a] -an =irr
3- PAST- AUGMENT- TRANS- collect -CONT =3AUG.DO

I’m leaning towards the following at present:

  • Some version of ‘unpacking’ for chapters where it’s crucial to know what the morphemes are.
  • Clitic boundaries marked always
  • Schematic glossing for examples where the full morphemic splitting would obscure the point. In the chapter on particles, for example, for an example like arranga maalanirr irr “without seeing them”, the important point is that arranga means ‘without’ and that it can occur with a gerund; it’s not important that the gerund’s structure is ma-jala-n=irr GER-see-GER=3AUG.DO.

I’m worrying about this for two main reasons. One is, of course, the grammar itself needs to be consistent and usable for non-Nyulnyulanists, since there are only two of us. I’d like other people to be able to read it without spending a year or so internalising Bardi verb morphology. The other reason is that if this grammar is going to be at all useful to Bardi people, the examples need to be readable and interpretable. It’s aimed at an academic audience, that’s why there’s a learner’s guide aimed at teachers and community members, but I have the feeling that if the glossing system is completely opaque it’ll be completely unusable, whereas if the examples can be followed more or less it should be possible to get some use out of it. Otherwise, I might as well use a different orthography too.

If you’re still reading, I’d welcome comments.

Advertisements

4 responses to “Glossing

  1. I think what I would find most useful and informative (speaking as a linguist, but not a Nyulnyulanist) would be a five-line presentation, something like this:

    1. Bardi sentence, unsegmented.
    2. Idiomatic English translation.

    3. Bardi sentence, segmented into morphemes but with no phonology undone.
    4. Underlying forms of morphemes, aligned with above.
    5. Morpheme-by-morpheme English gloss, aligned with above.

    In other words, I want it all! The last four lines would be really useful to morphologists, phonologists, and syntacticians, but the first two lines would be more readable for people who already know Bardi. You could use boldface to highlight the morphemes that are most important in each specific example.

  2. I agree completely with Q. Pheevr. Not having the underlying forms presented is one of the most aggravating things for me in working with a certain scholar’s writings on Tlingit. I can usually figure it out, but it’s still frustrating and slows me down a lot.

    I approve of reducing your segmentation in contextually appropriate ways. For example, in Tlingit there’s a sequence of verb prefixes which produce the future tense, but they undergo some complex morphophonological changes. Rather than glossing all of the prefixes individually, I usually gloss them as a “prefix cluster” which is simpler. The section on the future tense describes how to unpack the future prefix cluster, and gives all the possible forms depending on the ergative prefix. I only use the unpacked form like ga-u-gha-xha- where I’m specifically discussing one of the prefixes, otherwise I leave it as an opaque cluster like kkhwa- glossed as 1SG.ERG.FUT.

    One thing that you might consider is including an IPA transcription for examples where you discuss morphophonology. This can help make the effects clear to linguists who are not entirely familiar with the orthography. I’m not recommending including it for every example, but it could help for ones where the rules are complex, span syllable boundaries, have long distance dependencies, etc.

    Basically, the more the better. Q’s suggestion for putting the unadorned forms first is great, since the nonlinguist can still access the connections between the two without having to drag their eyes over all the segmented glossing. But don’t fall into the trap that linguists can “figure out” what is going on just from the natural forms. In this electronic world space is cheap, so help the reader with everything you can.

  3. 5 lines? You guys don’t want me ever to finish this grammar! :)

  4. As a non-linguist who enjoys reading and collecting grammars on out-of-the-way languages, I humbly suggest: (1) if yours is the only available grammar on the subject, in any language, please have at least some of your examples reflect commonly-used vocabulary and syntactic patterns. Despite owning two books on Quechua (naturally no two were on the same dialect), I had to do quite a bit of cross-comparison and educated guesswork to figure out how make “I want to … ” constructions before a trip to Peru. I still haven’t figured it out in Aymara. Too many of these books are either infuriatingly cursory (you can’t say “to want to” or form relative clauses), or else focus so much on curiosities (“I am prepared to raffle the goat”) that the practical and most commonly-uttered sentence patterns (i.e., the ones drilled into your head by a good self-learning course like Pimsleur) end up being ignored. Thackston’s Persian Grammar is nice and thorough in this way, and I sometimes use this as a guide to studying other languages. (2) Have at least some of your illustrations of a given grammatical principle be exhaustive enough that the interested reader can accurately generate their own solutions to expressive problems you haven’t anticipated. (3) If nouns or verbs have overwhelmingly complex inflections and/or phonetic rules (Nahuatl or Navajo, say), consider an introductory chapter breaking generation and interpretation into step-by-step processes so the practical user has some useful idea of how to tackle the problem. I really like Jacobson’s Practical Grammar of the Central Alaska Yup’ik Eskimo Language, Faltz’s The Navajo Verb, and Lockhart’s Nahuatl as Written for their handling of polysynthetic or complex phonological issues. I guess this is kind of long, but as a non-linguist who’s working on quite a few languages, those are the sorts of things I look at when trying to decide if a book will be helpful.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s