How sensitive do facial measurements need to be?
There are multiple aspects of facial expression that can be measured. It is possible that technological advances in imaging and computation can allow measuring each aspect in very fine grained units. This possibility raises the question of how much precision is required for facial measurements to enable some answers to the puzzles posed by behavioral scientists. It would be wasteful to put effort into making measurements more precise than necessary. How precise measurements will need to be to answer most questions is not known. But illustrations exist of how precise measurements need to be to describe some phenomena that have been studied. These are summarized in the table below and elaborated further in the text.
In the time domain, we know that the fastest actions cannot be accurately tracked at 30 fps NTSC video. Blinks, for example, are ambiguous as to whether the eyes close.
For the measurement of intensity, the 5 point scale used by the FACS scorer can be inadequate to capture the important relative activation of muscles in combinations. Also, a continuous measurement of intensity across time could be helpful.
Regarding which actions must be measured, consider the Felt versus Unfelt smiles. A number of research studies, including work on brain activity and self-reports of felt experience, indicate that one difference between emotional smiles and non-emotional smiles is whether the eyes crinkle or squint slightly during smiling. This crinkling is caused by orbicularis oculis, or AU 6 in FACS. (The video also shows the action of the inner orbicularis oculi or AU 7.)