Now, of course lots of people said you'll never be able to tell from the wiring diagram whether a synapse is inhibitory or excitatory. Well, it should be possible to take nervous systems where you have a wiring diagram and compute with them that if you assumed these were excitatory and these were inhibitory you will find that certain states of these systems jam. They're incompatible. And, in fact, it seems to me to be very reasonable to start to ask what things, given this kind of general structure, are impossible. Because exclusion is always a tremendously good thing in science, if you can say it is impossible for this to be inhibitory; we actually could show that in one small sub-case, which we did. And I think when people said, how would you actually do this, then the argument was: it would be done by simulation. And it was during that time that I developed the concept I called then a proper simulation as against an improper one. Now, let me just give a definition of that. If you take these little worms, they crawl around on the surface of an agar plate, and anybody can write a program on a Macintosh or any computer that has little worms crawling around on a screen - little squiggles - and they squiggle around and they... they look as though they're doing the right thing. And you say, there you are, I've simulated the movement of the worm. But I call that an improper simulation, because all you've done... so I ask you, please show me the program, and I look at the program and the program is full of sine theta, cos theta, okay? So what you've done is you've... you have noted, you have described the behaviour of the worm in a rather strange language, which happen... may happen to be Fortran or C, and which implements sine theta, cos theta, right? It's a... it's a description. What I would have expected to find in the program were lists of neurones, lists of connections, lists of things that have to happen, right. And so a proper simulation must be done in the machine language of the object being simulated. And I think that that makes the simulation crystal clear what you have to accomplish. And I think that if you do it that way, then I argue the computer program is the explanation. Now, if you have that, you can explain everything about the nervous system, all right. Now, you need a complete wiring diagram because modellers... there's always... are always confronted by the sceptic. So if you come and explain to the sceptic, 'I have modelled this behaviour, we got this oscillator, this interacts with this, it's coupled in this way and it does it', and he says, 'That's very nice, but how do you know there isn't another wire which goes from this point, you know, comes, goes right around the back and comes in at the side again?' And what you need to be able to say in all of these is: there are no more wires - we know all the wires. All right. So you must saturate the... the kind of database, if you like, so that nothing can be added to it, and then I think you have a real test to proceed. And we then said: 'Well, let us get the structure of the nervous system and then we can solve the problem of behaviour, because we can make perturbations, we can do experiments, and indeed those experiments can be done.' And what is also known is we know what the system has to produce by the genes, you see, because the reason a nematode has stereotyped behaviour, has so to speak a... it's the only... well, it's one of the few organism where memory is encoded in nucleic acid sequences, but it's a species memory, it's not an individual memory. And these nematodes simply remember in their DNA to build the same nervous system, which then has the same behaviour, right, because there's no learnt behaviour. So it's a bad use of the word 'memory', but it does say what you have to do in order. If you wish to encode complex things in DNA, you must have the read-out system. And then the read-out system for encoding behaviour in DNA is the capacity to build a nervous system of this particular sort.