There are significant quantities of psychological research that show that function words (and, of, with etc) are almost skipped by the reading mechanism. Their presence is usually inferred by the context. However, they are read, and their presence can be pulled from short-term visual working memory if the context is ambiguous (i.e. if the sentence does not parse correctly).
The test is made harder by the use of all-caps, which removes shape information from the words. However, the test is made easier by the presence of the question prior to the text. It prepares the brain for the test, and so the reading mechanism is altered ("primed") by the new requirement (over standard comprehension). The test would be harder if you said "Read this: <passage>, <test>".
If you're really interested in this, you should read "Visual Cognition" (Humphreys/Bruce), and "Visual Perception" (Bruce/Green).
Asking the question does make it easier, but there's really no way to know how many F's you counted without asking it. You're damned if you do and damned if you don't...
