Why did I guess that Tarski's problems
about first-order theories of non-abelian free groups
were not published in his papers or books, and he posed
the problems around 1945? Here are the reasons of that:
$\bullet$
In the papers of Kharlampovich-Myasnikov
[J. Algebra 302 (2006), no. 2, 451-552]
and Sela
[Geom. Funct. Anal. 16 (2006), no. 3, 707-730]
with solutions of Tarski's problems
there is no references to his publications;
$\bullet$
In the Kharlampovich-Myasnikov paper it is written that
the Tarski's conjectures were formulated around 1945;
$\bullet$
Lyndon in his paper [Problems in combinatorial group theory,
in: Combinatorial group theory and topology, Ann. Math. Stud., Princeton Univ. Press,
1987] formulated the problem on elementary equivalence of all non-abelian free groups
and called it a folklore problem of Alfred Tarski.
I checked Tarski's papers and books and now I know that his problems
on the elementary theory of free groups had been formulated in some of his
publications; so these problems were not `folklore'.
Also, it seems the problems had been posed not around 1945 but later.
The first Tarski's publication where the decision problem for
the theory of non-abelian free groups was mentioned
is the Tarski's book [Undecidable theories, North-Holland, 1953]. On p. 77 he informed that
he stated his result of undecidability of the elementary theory of groups
at a conference in Princeton in December, 1946.
On p. 85, he wrote: 'For many extensions of the elementary theory of
groups, e.g. for elementary theories of finite groups and non-Abelian free groups,
the decision problem remains open'.
Note that it was not a conjecture on decidability of the elementary theory
of non-abelian free groups; he just said that it was not known whether
the theory is decidable or not. Clearly, the question had been raised
between 1945 and 1953, and I don't know when exactly.
In that Tarski's book the problem of elementary equivalence
of all non-abelian free groups was not mentioned.
First time it was published in Vaught's abstract
[Bull. AMS, 61 (1955), N 2, 173-174], where he formulated his theorem implying
that standard embeddings of free groups of infinite rank are elementary and wrote:
'These investigations arose from a still unresolved conjecture of Tarski's
that any two free groups with at least two generators are arithmetically equivalent'.
That theorem was a result from Chapter 3 of Vaught's PhD thesis (Berkeley, 1954),
the supervisor of which was Tarski.
Later the Vaught's theorem and the Tarski's conjecture were published in Tarski-Vaught's paper
[Compositio Math. 13 (1957), 81-102]; see pages 82 and 98.
Also, in the introduction of the paper it is written that the decision problem for the elementary theory of
free non-abelian groups is a closely related open problem.