Measurement sources & issues
Existing records are a data source
Observation should be unobtrusive & verifiable (more than one observer)
Observers should not know when treatment begins
Paper & pencil measures are reactive; multiple testing effects
Self-reports correlate poorly with obj. indicators