We were beginning to study quantum field theory on the mass shell, looking at scattering amplitudes on the mass shell, coming from all Feynman diagrams to all orders, and presumably therefore getting results that were true, non-perturbatively, non-perturbitively on the mass shell. So by 1955 we had crossing and we had dispersion relations, not only forward dispersion relations for a massless particle being scattered, but non-forward relations for a massive particle as well. And to that we then added a generalization of unitarity for scattering amplitudes, namely unitarity on the mass shell, for amplitudes on the mass shell, but carried away from real momenta into the complex momenta that you could get on the mass shell in… in these problems. Well, putting all of those together I  suggested then at the Rochester meeting in 1956, having thought about it for quite a while, since ’55, I suggested that these three principles; the crossing, the dispersion relations, and the generalized unitarity could be used to characterize scattering amplitudes on the mass shell. One would get, they would be non-perturbatively correct. But of course, so as not to be pretentious, what I said at the meeting was that they were true to all orders. But what I meant of course was that they could be used non-perturbatively, and that therefore with the addition of a Born approximation to get started, or a condition at infinity or something like that, one could actually get a set of equations that would determine the scattering amplitudes in a given theory–but on the mass shell, without ever going off the mass shell. So this was a way of doing the whole of quantum field theory but staying on the mass shell. Well, I've looked at it that way ever since: it is quantum field theory on the mass shell. In the course of my talk at Rochester in early ’56–of course by that time I was at Caltech, but it was part of this same work–in the course of the talk I referred almost jokingly to Heisenberg and his use of the Breit-Wheeler S matrix. Because Heisenberg said, ‘Well you know, we don't have to derive the S matrix, we might guess it’. And, I was sort of, I thought it was sort of a joke to refer to this Heisenberg S matrix program in a sentence or two.