If I got what Martin said, it would mean that as Flow rate is constant throughout the whole loop, if you put a thinner tube you would have more velocity (but less...what? something has to decrease, nonsense otherwise)
Cross-sectional area decreases. Flow rate equals cross-sectional area times velocity, or Q=A*V. Flow rate, Q, must be constant everywhere, so by decreasing A (thinner tubing, nozzle in water block etc.), V must increase.

Yai, this doesn't make much sense in my head at the moment, I admit. Pressure is force/surface, flow is liters (or gallons)/second yet what is speed in therms of water? The speed of each particle? I can see that with a different tubing you can change the velocity of water, yet I've got no idea how to quantify it....
That's a little tricky. Individual water molecules will move in an unpredictable manner, so looking at just one won't work. What you do (one way to do it, at least) is imagine a tiny cube placed somewhere in the fluid flow and look at the average net number of water molecules entering and exiting each face of the cube. Skipping the rest of the details, this gives you a way to say in which direction, and how quickly, the water in the cube is moving on average. There's a lot more to this, but that would make a very long post not very relevant to water cooling. :P