John McCarthy, who for a time used CAS---as well as CAR and CADDR, which survived---in LISP, thought that the instruction came first, i.e., that the instruction was the model for the arithmetic-IF statement. It is my guess that he was right, but it may now be too late to resolve this particular chicken-egg conundrum. What is clear is that the two were closely linked.
It is known, was indeed emphasized by Backus, is that he was much concerned not to deprive 704 assembly-language programmers of facilities they prized. His situation was very different from ours today. He wanted to open up programming to others, but he crucially needed to convince 704 assembly-language programmers of the time---There were no others---that FORTRAN was not a performance dog; and to this end he tried very hard to make analogues of the crucial 704 machine instructions available in FORTRAN. I looked shortly after his death in 2007 to see if there were any oral-history recordings of interviews with him, but found notjhing but some sanitized, hagiographic IBM materials. There may now be more out there, and I understand that there is a biography in the works. John Gilmore, Ashland, MA 01721 - USA
