We make the volume element so small that no fluorescing molecule generally is inside, and you will say so what? If you want to measure a fluorescence signal you want to have as many as possible molecules in there. That's right if you measure the average single strength. But as I've said already, we measure the fluctuation, and the fluctuation is strongest if you've only a single event, because the square root of one is one. So, what does it mean now with our method? We sit there and wait until a molecule diffuses into that volume element. It's not steadily in there, but occasionally it diffuses in. And once it diffuses in it will be excited. The volume element is as large that the diffusing molecule spends about a millisecond or a tenth of a millisecond inside. And since the fluorescence lifetime is of the order of 10-8 to 10-9 seconds, it means when the molecule is inside the volume element it emits many fluorescence light quanta because it gets excited, multiply, and emits many thousands or ten thousand light quanta. So you get a burst of light quanta, and then nothing again. And this burst can be measured, but of course you want to... since a molecule spends most time outside the volume element, the burst occurs only occasionally and if you would measure it steadily you wouldn't see very much, the noise would overthrow it.
So here comes in the, we say it, fluorescence correlation spectroscopy. You do correlation, or auto-correlation, a method invented by Norbert Wiener in the '30s... a mathematician at MIT. Auto-correlation means you simply don't pick up only the intensity as it appears at your measuring device, you take the intensity at a time, T, and multiply it with the intensity you measure in the next time element, T + ΔT. So you don't pick up single intensities, you pick up products of intensities. Now you see, if you have a noise then one of the intensities is high but at the next time moment there is nothing, so the product is zero. Zero times any number is zero. But if there are a thousand or ten thousand light quanta while a molecule is inside the light cavity, the next time element will be also. And you integrate over the whole and suddenly you see wonderful element, and you see already you're observing a single molecule. The method wouldn't work if you would have to average over many molecules. At the moment where you pick up your light quanta, there's one molecule inside your volume light cavity and you can pick up. So this is a very sensitive method.
