Abstract
Two factors give rise to the need for an iterative approach to evidence, inexactness of measurement and complexity of the phenomena being represented. As Duhem emphasized, the first of these alone entails any number of alternatives at a comparable level of agreement with any one representation. Duhem, and van Fraassen after him, contend further that any representation of a regularity can never amount to much more than a curve-fit insofar as the evidence for it cannot exclude any number of alternative disparate representations at a comparable level of agreement. The reliance on least-squares curve-fitting in many iterative approaches to evidence – especially in representations of orbital motions – seems on the surface to support this view. The question addressed here will be the extent to which iterative approaches that rely on least-squares methods can nevertheless achieve evidence that markedly restricts the range of alternative representations.
