Relativism and Reliability

Abstract Neither principle implies the other. Correctness could change spontaneously through time in a way that the scientist cannot affect, and correctness could depend on what the scientist does in the limit without changing through time. The nonexperimental inductive setting considered in chapters 1 through 13 satisfies both axioms, since correctness is taken to be a fixed relation between data streams and hypotheses in that setting. Experimental investigation of structural hypotheses also satisfies both axioms, since structural relations in the system under study are not altered by the manipulation of variables. But there are circumstances in which both axioms can fail.