Hey guys sorry for the late reply.
I hope this answers to both you guys’ questions.
I haven’t used any specific reference. Having a mirror and a human skull on my desk most of the time has been of great use 
I just purely analize the machanics under a physical and anatomical point of view, adding some artistic liberty.
I always try to devide my observation time from my work time. I basically try not to use reference for as long as I can.
Most important is to understand before observing I think. So take in consideration what DaVinci said about muscles:
muscles work as ropes and pullies. Take the fact that they always pull and never push and that skin more or less acts like cloth sliding over bones and then you’re done!
The idea of imagining skin to behave like curtains really helped me to also make sure I would keep the volumes. Once the UVs were been moved around
I would use “Project All” on a duplicate geometry of the head which was the best tool for that purpose (I will make a video of that).
This to underline the importance of using ZBrush for this process. I did this kind of work before in maya and it becomes tedious and tricky. The use of subtools really makes it super fast.
Maps are set in a similar way, but it’s how you set the shader that does the magic. Instead of using pre-subtracted maps, I use 3 displacement maps baked out of zbrush
from extreme expressions (positive or negative or corrective).
It’s the alpha maps that are going to be applied to the transparency of one displacement map comped onto the other. If you have the displacement maps one over another as layers in photoshop,
you can use alpha maps to reveal this or that area of the maps. The video you saw it was rendered in 3Delight which has a very similar layering system as photoshop has.
I’m creating the shader for Vray now so I’m going to post it as soon as I’m done with the render in Vray.
So it’s the shader that is connects the alpha maps to the blendshapes. When a blendshape goes from 0 to 1, the value of the alpha goes from 0 to 1.
Btw, with positive, negative and corrective I intend to say that the morphing motions are the ones to be positive and negative and so on. The shapes are a lot not only three. The displacement maps are only three
because they are baked from more then one layers turned to 1.
Maps are so named after the layers. So I decided that my positive motion shapes would have been the ones that would have pulled the skin upwards.
Assigning a negative value to each of those layers would get me new shapes, that due to the nature of the linear interpolation of the morph would give me bad looking shapes.
So I resculpted those “negative” shapes storing them into new layers that I would actually recognize as those which would give me the negative motion. Funny enough the positive shapes are also
ones of positive emotions most of the time!
In few words if the brow goes up I consider it to be creating a positive motion; a negative motion would be given by the same brow going down.
Under an anatomical point of view, those muscles who would pull the brow up would give me a shape with details that would be totally different from the ones produced by
those muscles who pull the brow down. Basically one pisitive shape is not a mirrored negative shape.
This is because organic motion does not happen on a line. Ok too long. Hope it wasn’t boring.
Check also this thread, I posted some work I’ve done with my friend Justin Holt.
http://www.zbrushcentral.com/showthread.php?161471-Banjo-Man
small_orange_diamond
!