ABSTRACT
In other chapters we present a rationale (chapter 9) and empirical data (chapter 11) supporting our contention that the GAS score is a valid measure of the change, relative to expectations, that occurs during mental health treatment. Whatever a score is intended to measure has a direct bearing on how the reliability of that score should be assessed. If a score is intended to measure a construct that is expected to vary over time, then variability in the observed score across time may be reflecting some true variance in the underlying construct rather than just error variance (e.g., unreliability due to test-retest effects). A measure of change, by definition, would be expected to provide different values across two or more scoring occasions whenever change was likely to have occurred between those occasions.
