This question was asked on Stack Overflow, but closed as off-topic there. Before it was closed, it received this answer (visible only to sufficiently-privileged users, so quoted here, lightlylightly edited by me):
I had guessed that "string" was in use by mathematicians long before its adoption in programming languages. Turing machines effectively operate on strings. Turing may not have used the term, but it is used everywhere in automata textbooks, going back decades.
The earliest reference I could find was a fragment in Google books of a 1944 article "Recursively enumerable sets of positive integers and their decision problems" by logician Emil Post in Bulletin of the AMS. I think there is little doubt that he is using "string" in the conventional sense used in computer science. Page 286 contains:
For working purposes, we introduce the letter b, and consider "strings" of 1's and b's such as 11b1bb1. An operation on such strings such as "b1bP produces P1bb1" we term a normal operation. This particular normal operation is applicable only to strings starting with b1b, and the derived string is then obtained from the given string by first removing the initial b1b, and then tacking on 1bb1 at the end. Thus b1bb becomes b1bb1.
