Now I'd like to describe an experiment, and tell you its surprising results. In this experiment some photons of the same color—let's say, red light—are emitted from a light source (see Fig. 2) down toward a block of glass. A photomultiplier is placed at A, above the glass, to catch any photons that are reflected by the front surface. To measure how many photons get past the front surface, another photomultiplier is placed at B, inside the glass. Never mind the obvious difficulties of putting a photomultiplier inside block of glass; what are the results of this experiment?
For every 100 photons that go straight down toward the glass at 90°, an average of 4 arrive at A and 96 arrive at B. So "partial reflection" in this case means that 4% of the photons are reflected by the front surface of the glass, while the other 96% are transmitted. Already we are in great difficulty: how can light be PARTLY reflected? Each photon ends up at A or B -- how does the photon "make up its mind" whether it should go to A or B? (Audience laughs.) That may sound like a joke, but we can't just laugh; we're going to have to explain that in terms of a theory! Partial reflection is already a deep mystery, and it was a very difficult problem for Newton.
(from QED, by Richard Feynman, p. 17)