How many hours of recorded speech?

Julia Sallabank js72 at SOAS.AC.UK
Tue Aug 28 09:50:45 UTC 2012

Dear Lindsay

I agree with what has been said before, but would stress the importance of
quality as well as quantity. Not only in terms of the quality of the
recordings - which it goes without saying is essential to capture nuances
of vowel quality, tone etc. (see
guidance on best practices in language documentation can be found at

It is also of course essential to archive the recordings, metadata and
analyses in a safe and accessible archive which is regularly backed up and
updated (e.g. ELAR, DoBeS).

As well as this I would recommend thinking about the (potential) purposes
of the recordings and what kind of language would be useful for them. I
would urge you to think of future uses that community members and
their descendants might like in terms of language revitalisation/revival,
perhaps even 200 years after the last speakers, as is happening in
Australia.  As noted by Sugita (2007, and Amery (2009,,
documentation often omits the kinds of genres that would be most useful for
language revitalisation/revival, e.g. conversational gambits, childcare
language, jokes, insults, swear-words etc.

I agree with Graziano that elicitation can help help to fill gaps, but
beware that if the speakers are undergoing attrition or there is a
diglossic situation, elicitation and grammaticality judgements may elicit
incorrect data due to perceptions of 'correctness'.

I hope this is helpful


On 28 August 2012 08:32, Graziano Sava' <grsava at> wrote:

> Dear Lindsay,
> I totally agree with Seffen's comment. I would add that language
> documentation and description is not only data collection but also data
> analysis in the field. One should take time to sit with language
> consultants to understand meanings and structures of the language. The more
> you collect and the more you understand, analyse and describe, the more you
> realise what are the areas of the grammar that are unclear. In order to
> fill analytical gaps you could record texts of specific genres and topics.
> For example recipes for the imperative or idea's on one's future life for
> the expression of future or modal forms. To make sure you get what you need
> you might use basic elicitation, creating specific example sentences to be
> verified in the texts and that verify the texts.
> Hope this helps.
> All the best and good luck,
> Graziano
> 2012/8/27 Steffen Haurholm-Larsen <shaurholml at>
>> Dear Lindsay,
>> You are raising a question that is relevant for most all linguists
>> working with endangered languages, no matter wether they consider
>> themselves "documentary linguists" or not. As you are implying, and as is
>> well know to us all, the work that you are doing on this language right now
>> might be the last, and in some cases the only record of the language.
>> Regarding the amount of hours or recordings that is "enough" I think that
>> there is just no way of answering that - the easy answer would be to say
>> "as much as possible" withing the limits of your budget and time frame of
>> the project. However, it will be helpful for you to have a goal to work
>> towards, so, the practical way of going about setting the limit would be to
>> first make a small pilot project where you run the whole process from
>> recording through archiving with all of the steps in between to see how
>> long it takes you to process, say 10 minutes of conversation. This will
>> give you some kind of hint as to how long you may expect things to take and
>> thereby how much you can afford to document. But you don't want to limit
>> your recordings to what you will be able to transcribe - perhaps the bulk
>> of your record will just be provided with a "rough transcription" that will
>> tell the user what is going on in it. So, record as much as you can and
>> then select the stuff you find the most interesting to transcribe; another
>> bonus of this approach is that you will likely have different genres to
>> choose from when you decide what to transcribe and gloss in detail.
>> In terms of choosing speech genres I think you should go out into the
>> community and find out how people interact; in other words, let the speech
>> genres show themselves - this is of course easier said than done, but on
>> the other hand, you (or somebody) will have to be present in order to
>> record the speech event anyway. This brings me to my final point - it might
>> be worth considering providing a number of speakers with solid state
>> recorders and have them record to whatever extent they and their fellow
>> community members feel comfortable. This will eliminate, to whatever extent
>> possible, the influence of you as outsiders on the record. This of course
>> has its drawbacks because you loose control almost completely of the
>> recording process and equipment placement, but if you have funds for it and
>> if speakers like the idea, it certainly couldn't hurt.
>> Good luck with your project and enjoy your fieldwork experience!
>> Steffen Haurholm-Larsen
>> Phd. student - University of Zurich
>> On Mon, Aug 27, 2012 at 9:39 PM, Lindsay Marean <lmarean at>wrote:
>>> I'm helping to document a language with few first-language speakers
>>> living.  We want to record them speaking naturally (and transcribe and
>>> translate the recordings), and we hope to use this documentation as the
>>> basis for more language description in the future.
>>> I'm looking for people's opinions, experiences, and citations - how many
>>> hours of recorded speech are minimally "enough" to most likely represent
>>> the grammar of the language?  Are there particular discourse types that we
>>> should be certain to record, besides narratives and conversations?
>>> Best regards,
>>> Lindsay
> --
> Graziano SavĂ  - PhD Leiden (African Languages and Linguistics)
> Postdoc DoBeS-Volkswagenstiftung
> Based at LLACAN-CNRS
> Personal links
> Blog
> <>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the Endangered-languages-l mailing list