How do you write an unwritten language?
Writing never expresses all the details of spoken language, so how do you know what is important to write and what is not? One way is for a linguist to collect lots of words from one speaker and then spend time analyzing the data and determining what is important and what is not. In a recent workshop we used another method, involving a large number of participants.
First, we had the people write lots of words on little pieces of paper, spelling them however they wanted. Then we sorted these papers by the way different things were spelled (once by first vowel, then by second vowel, etc.) and they read them all off. The speakers put them in piles where the item in question was the same. This way we were able to determine what needed to be written (distinguishing between the piles), no matter how they were originally written. For example, we found two sounds, both often written “i”, that were actually different.
We also looked at things that were changed by context. For example, “t” in front of one “i” was pronounced “ts”, but not in front of the other “i”. This helped confirm the distinction of the “i”s, but also indicated that we should always write “t”, whether it was pronounced “t” or “ts”. This same kind of change happened for many consonants in front of certain vowels: k / kf, g / gv, s / sh, etc. Writing each consonant consistently helps to simplify the system, and through this method we could confirm that this would not lose important contrasts. Another advantage of the method was that many people got practice in recognizing differences between sounds in their language and in using the new writing system.
For more details, see the report of this workshop here.