Now transforming from the time and frequency domains to the axial
length and spatial resolution domains, we can apply Fourier's same
principles to the shape of a letter's limb. For an eye to resolve that
squared-off sans-serif character requires infinite (or at least high)
resolution. Resolving the sans-serif character is easier, because the
serifs are analogous to the ringing of a frequency-limited wave (and
the spatial frequency of our eyes is limited).
So the argument is that at lower resolutions, the serifs help the eye
determine the terminal part of the line. Hmm...
Is the difference between reading serif and sans-serif so significant that
it makes sans-serif unsuited for web use? Or is the real-world effect less
of a problem. Remember that you don't actually need to scan the whole
letter when you read! We scan words and phrases, and when we actually do
look at a letter generally Gestalt theory suggests that unless we are
specifically looking for it, we'll never even see the imperfections in
face.
If people generally never read sans-serif in the real world, ok, I can
appreciate how reading a new font is going to require more "noticing". But
no one really reads that way with a font similar to what they're familiar
with.