ABSTRACT
Connections With Other Chapters
The previous chapters have analyzed three important concepts in validity theory: measurement, causation, and meaning. This chapter presents an integrative model of validation, and ties together the concepts by showing how they play out in the larger scheme of validity theory and validation research.
One can summarize the arc of the previous chapters as follows: Measurement involves an inference from observable response behavior to the property measured (chapter 2). The same holds true for almost all cases of assessment and testing. In psychometric models, such inferences are commonly cast as inferences from observable to latent variables (chapter 3). To support such inferences, item responses must carry information about the measured property. This means that these responses either depend on that property (reflective model; chapter 6), or determine it (formative model, chapter 6), or are part of an estimate of it (behavior domain theory; chapter 5). The inference from observation to property thus requires a dependence relation. We understand this relation as either being causal (chapter 6), or as resting on other causal assumptions (chapter 5) but leave the door open for a fully non-causal account of assessment should one emerge. Within the causal interpretation, the nature of the causal relation is not fixed. Several ways of constructing that relation are open to the researcher (chapters 7 and 8).