QUOTE(Dulciana @ Nov 20 2008, 12:46 AM)

She can bring this 'voice' down quite low, and can take the other, harder voice up quite high. So what we were wondering was - how do we move naturally between the two without it sounding like we're changing instrument mid-phrase? Because it really does sound like two different singers! We tried some experimentation, and she can sing almost a whole verse of 'Away in a Manger' in either mode, but not quite - she needs to change 'voice' for either the highest or lowest notes.
This is very common in untrained voices. Basically your vocal folds can be thick, thin or stiff. If they're thick, you get the rich voice sound you're talking about at the bottom, if they're thin, you get the sweet sound you're talking about on the top - if they're stiff, you get an airy falsetto type sound, like a voice breaking. Any one of these vocal fold configurations can be used in any part of the voice (which is why she can bring the light sound down and the thick one up) - it's more dangerous, without proper supervision, for her to bring the thick one up, however.
In thick fold singing, the larynx posture is natural, as in speech. In thin fold singing, the chords are longer and more taught and the larynx actually moves a little to make this happen (the thryoid cartilage, or adam's apple, tilts downards over the cricoid).
To move between the two modes without a sudden change, she simply needs to find the point where her voice does it automatically (what note) and then make the cartilage move consciously a few notes earlier in her range. You can easily teach her to do this by getting her to put a little whimper into the voice as she goes up - this will make the cartilage move in the necessary way and stop any sudden changes in voice happening. Kids love to touch their adam's apple (or the bigger of the two larynx lumps in her case) and feel it move down when they make a crying sound - adults find it gruesome, but kids love it! Teach her to make the move, and then - very important - keep it there for all the notes above that area (her first reflex will be to do it, then let it go straight away - she'll get there in the end). A load of sirening on 'ng' at a very low volume over her entire voice will get her used to when she has to do it - eventually it will happen automatically.
Once she's got used to this, you can play a resonance game with her - get her to feel how her low notes feel like they resonate in her chest (they don't - the chest can't resonate, being full of organs and stuff, but it feels like they do and that's useful as feedback for singers) - then get her to feel how the high notes seem to resonate in the head area. Get her to imagine that the resonance is a lift and have her put it on the ground floor and sing a low note, then on the top floor and sing a high note. Then you want her to do some sliding siren exercices on the sound 'ng' from the bottom up to the top, with the elevator moving all the way from the bottom to the top - tell her she mustn't miss any floors out. Repeat on the long i vowel (as in heat), then on the words of the song.
This will balance the resonance and move her away from the mechanical thing of tilting her larynx and help her to apply the tilt in a musical way - basically it will get her back to singing rather than mechanics, and the lift resonance thing will help her tone to be even, whilst forcing the larynx to continue to make the changes necessary.
These two devices should help balance the two registrations out - but it may take more time than you have for her to get comfortable with it all. If that's the case - just have her sing the whole thing in the sweet light mode, with a little cry in it - it will sound light and pretty and won't do her any harm (like taking the thick folds up without professional supervision could). Just one last thing, my advice is based purely on your description - without hearing her, I can't be certain that this is the solution, but your description describes an almost text-book jump between thick and thin fold singing that occurs in loads of voices, so it probably will work.
Let me know if you have any questions !
Allan