I describe a piecemeal approach to the question whether we should be realists about well-
tested scientific theories (i.e. believe they are true) that is based on a criterion for having a good reason to believe
a claim is true. Though the criterion is realist in spirit, it delivers an anti-realist conclusion about our current (and past)
knowledge at the level of abstract theories. Nevertheless, we have evidence that satisfies the criterion for many of our
low-level empirical generalizations, and even for some claims that go beyond observables. Thus, the criterion justifies a
good deal of epistemic modesty about actual science, but doesn't make us into skeptics or even strict empiricists.
The anti-realism that follows from this approach has the advantage over anti-realist views based on empiricist
assumptions (e.g., Constructive Empiricism) of setting no limits a priori on what science can and cannot ever achieve.