Peter Achinstein (Johns Hopkins Univ. Baltimore) gives a talk at the Workshop on "Why trust a Theory?" (7-9 December, 2015) titled "Scientific Speculation". Abstract: Throughout the history of science controversies have emerged regarding the legitimacy of speculating in science. Three very strong views about the general practice of speculating have emerged: One, very conservative, says “never do it, or at least never publish it.” It is the official doctrine of Isaac Newton: “hypotheses have no place in experimental philosophy.” (Of course, he violated his official doctrine on several occasions). Another, more moderate position is the official doctrine of hypothetico-deductivists such as Whewell, Popper, and Hempel: speculate freely but verify before publishing. The third, the most liberal, is suggested by Feyerabend’s principle of proliferation: speculate like mad, and publish, even when you have no idea how to test your speculations. In my talk I want to reject all three of these views. They are too simple-minded. Some speculations are good ones, some not so good. I will ask how a speculation is to be evaluated. In the process of doing so, I will consider two historically important speculations: James Clerk Maxwell’s kinetic theory speculations from 1860 to 1875, and a speculation that has been put forward by some string theorists, as well as by others, viz. that there is a “theory of everything” (whether or not it is string theory). The first, I will argue, deserves praise, the second does not.