+Pieter Belmans (re)discovered a proto-drawing of Mumford's iconic map of Spec(Z[x]) in his 'red book'. The proto-pic is taken from Mumford's 'Lectures on curves on an algebraic surface' p.28 and tries to depict the integral projective line. The set-up is rather classical (focussing on points of different codimension) whereas the red-book picture is more daring and has been an inspiration for generations of arithmetical geometers.
Still there's the issue of dating these maps. Mumford himself dates the P^1 drawing 1964 (although the publication date is 66) and the red-book as 1967. 
Though I'd love to hear more precise dates, I'm convinced they are about right. In the 'Curves'-book's preface Mumford apologizes to 'any reader who, hoping that he would find here in these 60 odd pages an easy and concise introduction to schemes, instead becomes hopelessly lost in a maze of unproven assertions and undeveloped suggestions.' and he stresses by underlining 'From lecture 12 on, we have proven everything that we need'.
So, clearly the RedBook was written later, and as he has written in-between his master-piece GIT i'd say Mumford's own dating is about right.
Still, it is not a completely vacuous dispute as the 'Curves' book (supposedly from 1964 or earlier) contains a marvelous appendix by George Bergman on the Witt ring which would predate Cartier's account...
Thanks to +James Borger i know of George's take on this "I was a graduate student taking the course Mumford gave on curves and
surfaces; but algebraic geometry was not my main field, and soon into
the course I was completely lost.  Then Mumford started a self-contained
topic that he was going to weave in -- ring schemes -- and it made
clear and beautiful sense to me; and when he constructed the Witt
vector ring scheme, I thought about it, saw a nicer way to do it,
talked with him about it and with his permission presented it to the
class, and eventually wrote it up as a chapter in his course notes.
I think that my main substantive contribution was the tying together
of the various prime-specific ring schemes into one big ring scheme
that works for all primes.  The development in terms of power series
may or may not have originated with me; I just don't remember."
which sounds very Bergmannian to me.
Anyway I'd love to know more about the dating of the 'Curves' book and (even more) the first year Mumford delivered his Red-Book-Lectures (my guess 1965-66). Thanks.
Pieter maintains an "Atlas of this picture" here:
2 Photos - View album
Shared publiclyView activity