Just to expand on Mikes excellent answer.
Mike uses what's called a third order network on the tweeter without going into why he did that or what even a third order network is. For this purpose you can think of it, calling one speaker terminal positive, and the other negative, as a capacitor followed by another capacitor in series on the positive terminal going to one terminal on the tweeter (actually a small network as I will explain later - but forget that for now - simply think of it as connected to a terminal on the tweeter) then the other terminal on the tweeter going to the negative speaker terminal. Where the two capacitors connect you have an inductor going to the negative terminal. Now its well known as much as possible you should try to avoid signals going just through capacitors - they 'harm' the signal worse than any other component. The other two capacitors in the speaker are one on the woofer which has an inductor in the signal path like the capacitors in the tweeter and a capacitor like the inductor in the tweeter - not directly in the signal path going from the inductor to negative. Since its not directly in the signal path it does not do as much harm as the ones on the tweeter. The other is a small capacitor connected in parallel with a resistor to feed the tweeter - this is a small network Mike uses to get the flattest frequency response - because some goes through the resistor it is not as critical as the two where the entire signal goes through them. Experimentation has shown those other two capacitors certainly make a difference, and more than those into this sort of stuff like me and Rawl thought it would, but its not as big as those main two capacitors. Your best bang for the buck improvement is to use the best capacitors you can afford for them.
One way to improve a capacitors performance is whats called bypassing - that is in parallel with the capacitor you put a smaller but better quality capacitor. On his other speakers Mike uses Erse capacitors in the base model - good quality capacitor - but not up to either the standard of the Jantzen Superior or of course Duelund - but of course a cheaper. Mike in what he called his plus R used Erse as the base capacitor and bypassed it with Mundorf silver or silver and gold - cant remember which and it made an appreciable difference. I nagged Mike - he is the Duelund guy after all so why use Mundorf - so he switched to Duelund for bypassing the Plus R. Recently Duelund has released a capacitor specifically made for bypassing - the Duelund Silver Bypass (from Humble Homemade Hi Fi):
Technical specifications (according to manufacturer): "Precision Speaker Capacitor; Pure Silver Foil; handmade in Denmark".
Sound: In one word: amazing! I have used these 0,01uF pure silver foil capacitors as bypass capacitors in many different systems. In just about every case there was an improvement in overall coherency and realism of tone. With realism of tone I mean that acoustic instruments and vocalists seemed much more tangible, more intimate. The Duelund Silver Bypass Capacitor brings out a certain harmonic richness in overtones, for example with brass wind instruments the tone gets a sort of "glow" making them so much more realistic. Big-band music is bright and warm at the same time if you know what I mean, just like in real-life. And now for a cliché: the Duelund Silver Bypass Capacitor really does remove a curtain from the sound, creating more depth, insight and it generally does a very good job in cleaning things up. It does this without getting harsh nor adding artificial emphasis on any part of the spectrum, something that can happen with certain types of bypass caps. They even scared me once: my personal reference system uses a 1-inch beryllium compression driver connected to a tractrix horn that flares out to 29 inches. This "tweeter" has no resistors in the signal-path and is driven by a first order network, so only one capacitor in the direct signal-path. It works like an acoustic magnifying glass. I was fiddeling around as usual, tuning the value of this capacitor in steps of 0,10uF but the sound didn't quite seem to fall into place how I wanted it to. Then I decided to listen if the 0,01uF Duelund Silver Bypass Capacitor could help out here. Bingo! Makiko Hirabayashi's recording "Surely" (that I know very well) has some small bells that are used on several of the tracks. I thought I knew how they sounded but with the bypass capacitor in the network the bells were so realistic, it scared me :-) I am so pleased with these universal bypass capacitors that I have decided to keep them in stock. Verdict: add about 2 points to whatever capacitor you are bypassing.
Now they are quite small value and the technical side of Mike was skeptical because being so small you would not expect much signal to go through them so not make much difference. But as an experiment some were got and take my word for it the above is all true - it was tried with a lot of different capacitors - the best overall bang for buck was using it with the Jantzen Superior. I liked it with the much more expensive Jupiter Copper - but it is slightly bloomy - the Jupiter Aluminum was just a smidgen better than the Janzen for but there was hardly anything in it. The only caveat is if you are using really expensive capacitors like Duelund VSF sometimes it produced too much treble detail and depending on your upfront gear in that case you may be better without them. Using them on Duelund VSF copper made a Killer DAC sound worse but a squeezebox sound better. But its a no brainer with the Jantzen superior - it makes a marked difference.
I know it's $600 extra but hopefully what I wrote above explains why they are worth it - look on it as going to the Plus R version. Mike for the ML5, which is the best speaker he currently makes, has switched from Erse in the base model to Jantzen Superior.