Thanks to all who very helpfully replied.
Now I understand what I was missing:
The total psf is the product of illumination and emission psf. This is
obvious in confocal, but more subtle in WF.
In other words, the at focus peak in the intensity distribution of
illumination in the axial, z direction matters in both confocal and WF
Now I try to reexplain it to myself, any errors here?:
From a single photon in (and out) perspective :
The probability of excitation is proportional to the probability of the
photon arriving by all possible routes and constructively interfering with
itself. This probability is highest at focus by definition . The spatial
distribution of that probability in z direction is determined by the NA of
the excitation optics. Higher NA means faster drop in probability of
excitation with increasing axial distance from best focus, and vice versa.
Pseudo confocal illumination effect seen by Agard et al has bigger effect
for smaller ilumination field and higher NA,
Since the light intensity drops off faster as you go out of focus.
Widefield z resolution is improved in this case because resolution is about
contrast, and contrast requires signal. Excitation is weaker out of focus.
That gives some added contrast that a zero NA illumination system could not
have. In other words, at high illumination NA, the illumination contrast
modulation function (illumination OTF) is stronger for higher frequencies
in the axial direction than it would be for low NA illumination.