Both Dummett and Tugendhat seem to conclude that Frege's thesis that truth values are objects which are signified by certain sentences is an assumption which was unjustified even for Frege. In this paper I wish to show that Frege's thesis was one of several assumptions which led Frege to a complex semantic theory for the first order predicate calculus which is surpassed only by Tarski's truth and satisfaction definitions. As such, this thesis receives its justification by being an essential part (...) of a theory which as far as Frege knew was an adequate semantic theory for his version of the first order predicate calculus. This justification, now, of course, would not support the view that there are these objects. The reason for this is that Tarski's truth definition gives us a way of providing a semantics for a class of sentences which properly includes the class of sentences for which Frege's semantics is adequate without the need to assume the existence of truth values as objects. (shrink)