The D6 shows a white square on the face if it found a face within the custom group area. This is displayed only when in playback and pressing the zooming in button (for a couple of seconds). I have been testing it on singles figure skating today. Basically the group area (with face priority turned on) focuses on the closest face if it can identify faces and if not, it focuses then on the closest part of the subject within the group area. When the skater's arms are extended and she is spinning, previous implementations of group area would often focus on the arms or nose instead of face or eyes. It is bizarre how Nikon can implement this great functionality and doesn't properly document it for the user.
I would say that this makes a world of a difference in sports photography. Or it would, if the cameras were out there. :/ I tried for reference shooting in 25-point dynamic area which was my previous go-to mode for figure skating and I got lots of shots focused on the background because it's so unforgiving. I can work with it but it really needs to be used like single point, the grace provided by the surrounding points is so short-lived. Shooting with dynamic area requires constantly moving the selected point around whereas custom group-area allows me to select a range of points within which I will compose the face and it does the rest with a much lower error rate than I could with 25-point dynamic. And when the skater is close and fills the frame it allows me to use the frame fully so that I don't crop a limb trying to keep the primary point on the face. It doesn't always recognize the face (the skaters can have obstructions in the line of sight and expressions can be quite different from those in typical portraits). Still a lot of the time even when the face is not designated by the white square, it's nonetheless in focus. Maybe the camera has criteria for displaying the box that the focus has to be within tolerance but since the phase-detect sensor doesn't see anything when the actual picture is captured, it is not the ultimate judge, predictive algorithms can sometimes nail the focus even when the camera thought it is not in focus.
In auto-area AF, the faces are detected (and displayed with white boxes when zooming in playback) also outside of the autofocus sensor array. I suppose the camera then uses the RGB matrix meter image to determine where the rest of the subject is within the AF sensor array and then focuses on that. In some cases the camera shows a face but focused elsewhere. I think there is some kind of balance between straight closest-subject priority and face-priority and the camera makes a decision on which is more important. It would be nice to know the algorithms in more detail and how they justify the decisions made by the camera. It does work well, but in order to apply the correct focusing mode in different situations, more explanation of the implementation would be helpful.
I'm happy with the general performance, I gather that if I allow the camera to make more decisions, the results on average are now better than they were in past models, but there are still some situations where tighter control over the focusing is needed, if the photographer has time to exercise that control. I will try to post some screen shots and examples later though I will need to take photos of the back screen of the camera to do that. It seems ViewNX-i is currently not able to display the information that the camera shows in playback and zoom; it just shows all the focus points within the group area rather than giving additional information of which focus points were actually used to focus and where the faces, if any, were detected. I am sure this information is in the raw file, as the camera is able to display it, but it would require some additional work investigating the EXIF to find out how they are encoded.