Continuing to listen to more and more data, I'm now less convinced that the problem I had trying to tell the difference between two As is actually a problem of two different sounds.
Instead, I think that it's just one /a/ sound, but that can be made more forward in the mouth (like Australian 'rang'), or further back (like 'rung') but which can change depending on things like which other sounds that it's made near, whether there's a high or low tone, or whether it's long or short.
This kind of mind-changing throughout the analysis period (and after!) is a standard part of the process - no one's going to hear everything perfectly on the first parse - but it does lead to crazy, messy notes, which doesn't sit well with my clean-freak nature.
Other phonological problems I'm still not sure about (If you have no interest in phonetics, just stop reading now):
- There is one sound that could either be an affricate [dz](like English 'jam') or a frictive [z] (like the French 'rouge')
- This language has alveolar stops - like the t and d in english (although I think they're more dental)- and retroflex stops (where you curl your tounge under) and I can't here the difference a lot of the time.
- As well as alveolar/ retroflex distinctions there's a voiced/voiceless unaspirated/voiced aspirated distintion on all the stops. Basically, aspiration is where a little bit of air comes out when you make a sound, like English sounds p,t and k. Unaspirated is where your vocal cords don't vibrate, but you don't actually make any noise. English b, d, and g are often voiceless and unaspirated. Voiced stops are really heavily voiced. My silly English speaker ears hear all voiceless unaspirated as voice.
- Tone. I'm getting better, but you know, it's like suddenly realising reality has another dimention, but only if you look at it from the right angle...