This paper brings to first fruition an analytic schema based on four elements which involve a conception of skills independent or particular testing devices: (1) the development and application of a class of statistical models incorporating qualitative definitions of skill, distorted in item response by errors conceived as misclassifications; (2) a critique and reformation of the concept of test validity--making more concrete and specific the implications of invalidity; and (3) an integration and fusion of these concepts which allows meaningful empirical analyses of item response data. This conception/model is exemplified as contributing to the clarification of previously intractable technical and policy issues in the testing field. (Author/GK)