What were the initial motivations of the use of the proper forcing.?

I agree with Andres that this is a very ambitious question. Let me throw in a tiny little bit of information: As far as I know, Jensen's construction of a model of CH without Souslin trees was one of the first uses of both countable support iteration and master conditions (conditions generic over a countable elementary submodel of a sufficiently large initial part of the universe). One of the first publications using proper forcing seems to be Shelah 100, with the hilarious title "Independence results" (JSL 45 (1980), 563573). 


What I remember is that first Laver solved the Borel conjecture by an countable support iteration which added a real at every stage (1976). That such an iteration can be of any use, was very surprising at the time. Then Baumgartner introduced the very general notion of property A, which included most (all?) standard forcings known then which added reals and showed that countable support iteration of them behaves nicely (1978). Then came Shelah, who gave the proper definition (1980). This was again a surprising thing, as Baumgartner's definition was combinatorial (containing combinatorial properties of the poset, eh, almost) while Shelah simply required that P should preserve all stationary subsets of all sets of the form $[A]^{\aleph_0}$, that is, a semantic definition. Notice that Shelah's new theory gave new and elegant proofs to old theorems, as Baumgartner's consistency of that any two $\aleph_1$dense sets of reals are isomorphic or Mitchell's consistency of the tree property of $\aleph_2$. 


Roslanowski once asked Shelah about this, and has kindly typed down the answer he got: http://www.unomaha.edu/logic/papers/essay.pdf 


My understanding is that part of the early motivation was the observation of the attractive features of two main classes of forcing:
And what was wanted was a class of forcing notions in which these two kinds of forcing could be mixed together in iterations, while still preserving $\omega_1$. That is, what is wanted is a class of forcing that contains all ccc forcing, all countably closed forcing, which all preserve $\omega_1$ and which support some kind of iteration theorem. The full class of all $\omega_1$preserving forcing does not fit the bill, since it is not closed under iterations. But meanwhile, the class of proper forcing does have the desired features... 

