ABSTRACT
This chapter examines Coleman’s insight in terms of traditional statistics used in tests and measurements. A simple test-retest correlation may not measure true reliability because it is affected by temporal instability in a variable as well as by errors of measurement. One may attempt to reduce the effect of temporal instability on test-retest correlations by reducing the time interval between measurements, thereby reducing the amount of shifting that can occur in true scores. Once the reliability coefficient is available, it is possible to correct the test-retest correlations for attenuation and thereby obtain stability coefficients measuring the amount of change that occurred during a given interval. The reliability coefficient, unlike a simple test-retest correlation, is not attenuated in size because of changes that occur during the testing interval. These reliability and stability coefficients suggest that a sizeable proportion of the flux in family-size expectations is a matter of ambiguity in measurement rather than due to actual changes of mind.
