Three fingers bind to nine base pairs, six fingers will bind to 18 base pairs. Now if you simply go and add another set of six fingers... three fingers to the three fingers, just join them on with the same linker you find it gives you hardly any increase in affinity. Now the reason is very simple and I suspected this from the crystal structures and again from experience. The periodicity of the fingers doesn't quite match the periodicity of the DNA bases. So they get out of register, so you have to restore the register. So when we started making... adding three fingers we put in either some extra amino acids or we left gaps in the DNA and started to do protein engineering, or on the target basically we'd leave a gap and we developed different forms of combination of fingers. Now the most important one came later, where instead of having two times three fingers, we had three times two fingers. And the reason we did that was that the three times two fingers... these pairs of fingers had this cross standard interaction which gave you another tenfold boost in the affinity. So... between them we also... I had a favourite topic which was to say, 'Well, I'm going to find the site which is so rare that a single finger could recognise', and that was to take two DNA sites, that's a distance apart, so ten base pairs apart and you have a finger, two fingers which recognise each of these, linked by a fairly semi-stiff link, it has to be semi-stiff otherwise you loose entropy when you do the binding and in fact we showed that you could get enhanced binding this way if you had something with seven or eight base pairs apart, so it's a very nice laboratory exercise but it turned out to be, and we spent a lot of time on it, but that's the way of things, it turned out to be unnecessary although it's pretty but not useful. But that's what happens when you're doing this kind of research. So we found that when you got... if you put in the correct... did the correct protein engineering putting in things like glycines, serine glycine, things of that... purely empirical. What we found was that putting in a glycine, serine glycine in the zinc finger skipped one base. And that's how we worked out rules of this sort. Putting in a single glycine didn't skip a base, but readjusted, made things better. And so, we had a set of sort of know-how, which we began to build, libraries of six finger proteins.