“Reason, Mathematics, and Modality”, p. 4:
There are secondary virtues such as interestingness, elegance, the having of applications outside [the field at hand] and so forth, and these secondary virtues are quite important. … I contend that none of the virtues requires that the [scientific] theory be true.
Field is talking about mathematical theories, but this statement seems equally applicable to other scientific theories. Scientists adopt one theory rather than another for a variety of reasons. When both theories make the same predictions, these other virtues of theories are generally used to decide which theory to adopt. Quine and Putnam go on to suggest that the adoption of a theory is taken to mean an endorsement of its literal truth, but Field seems to take a slightly more sophisticated account. Adoption of a theory is just a way to make predictions, and scientists choose which theory to use for a variety of reasons, with accuracy of predictions being paramount. If adoption were to mean taking as true, then scientists would have to have odd metaphysical beliefs to use interestingness or elegance as guides to adoption. However, if adoption of a theory is something less commital, like perhaps some sort of instrumentalism or fictionalism, then these practices make more sense.