Traditional methods of developing tests have been driven by item content specification and have relied on the use of summed ordinal scores in an attempt to create a quantitative index of ability. There are fundamental problems with this approach to test development. The first problem is that the summing of qualitative ordinal counts to create a total score does not result in a number that is a valid means of making quantitative comparisons of performances. The second problem is that reliance on content experts does not ensure that the test items indeed test the construct in question. As a result, the development of functional assessments based on Rasch measurement models is becoming a preferred method among rehabilitation professionals for constructing tests. Rasch measurement models offer an alternative approach to instrument development that results in unidimensional linear measures based on additive numbers. Rasch analyses also generate goodness-of-fit statistics that can be used to perform confirmatory construct validity analyses of the constructed scales based on (a) formulating expectations about what should happen when a group of persons take a test, and then (b) confirming that the test items fit that model.
This paper describes how the many-faceted Rasch model was used to develop the motor scale of the Assessment of Motor and Process Skills (AMPS) (Fisher, 1992). The process of beginning test development with a theory of expectations related to a constructed scale and then confirming that the scale conforms to these expectations also is demonstrated.