There are a few mentions in the tutorials on this excellent website about working in 16bit mode even for 8 bit images.
With regard to Photoshop or After Effects, putting 8bit images into a 16bit project appears to allow far better manipulation of the 8bit image.
Is this all to do with higher precision calculations, as the 8 bit image still only has 256 levels to work with.
Or does Photoshop do something else, like throw in some very discrete gaussian noise to each channel to spread the values out into a wider range of levels available in a 16bit format.
I read somewhere that from Photoshop 5, when 16bit mode was introduced that it was 15bit image + 1 bit noise, whatever that means. :-)