Capacitors, especially when used for HF decoupling need to be looked at in terms of C, R, and L...

If you plot the impedance curve vs frequency across a ceramic cap you will see a falling slope starting at LF dominated by the real C, at some high frequency it stops falling and flattens out that region often small is dominated by the series R, above that the impedance rises again dominated by inductance.

The trade off with caps at HF is to get low impedance where you need it. The .1uF may start at a lower impedance and fall lower than a .01uF because it has more C, but it also has more R and L, so while a .01uF may start at a higher impedance, it will often stop dropping at a higher frequency then the larger cap so actually be more effective at very HF.. This is why we often use both larger and small value caps in parallel.

http://www.signalintegrity.com/Pubs/straight/resonance.htmGo to the link for an actual article by a real person... Imagine overlaying a plot for a .01 and .1, with the .01 shifted to the right and higher. If it is still falling while the .1 is already rising, the .01 can be better for HJF decoupling.

JR