[Arthur Macmillan]

To make it clearer,  if you are using a standard 85f1.8 not even f1.4, at 1.5 meters depth of field is 2cm before or after subject, focus differs whether camera focuses on eyes or nose...
So if you have mirrorless without eye autofocus or an SLR without AFMA done perfectly well and autofocus point perfectly chosen you have very high chances of having out of focus portraits.

I am not trying to take sides on DSLR or MILC this time.  I know some features are available with mirrorless cameras.  I was just trying to figure out the method as outlined which stated that the AF was set for a single point, the center point.  I'm not trying to dissect the Imatest and how they come up with sharpness.  I guess I was just surprised that one can really have a focus problem with a 50mm lens at 12 feet.  I mean, are we testing focus ability or the intrinsic characteristic of the lens?  And honestly, how could could you possibly beat live view manual focus?  I could tether the camera to a million to one contrast 24 inch monitor then zoom in 10x, and the cameras contrast focus can beat that?  I just was not able to get it focus and sharpness only go as far as the lens goes.  If you want a baseline for the test you have to be able to achieve the best focus manually, otherwise how do you have a standard by which to judge the autofocus systems.  I think Roger C. was right there were problems in the methodology because changing the lenses only proved that the autofocus tests were not appropriate.  For one thing, in general you would always calibrate the MFA for a F/1.4 or F/1.2 lens on a DSLR.  But I still think I am missing something.  And looking ahead at Klaus answer he mentioned something that has bothered me because I happened to be confronted with them in related situations.  So it leads me to another question, that again, I hope is not off topic because it came up when Roger C. decided to change the Micro motor 50/1.4 for the USM 50/1.2.

Just wondering...I do get your point about eyes.  It is a game changer.

-Mac