ABSTRACT
This chapter argues that computing a total score for each person by adding together scores on several items should show a large correlation with the true score because all of the random errors will tend to cancel out. Memory may also be a problem for personality scales and if the test content is interesting or unusual and is thus processed deeply on the first administration. Problems arise if people's levels of a trait change between two administrations of a scale. In order to measure states, other methods would need to be devised, as the test-retest reliability of a state should be close to zero. Like any other form of reliability, test-retest reliability will vary depending on who is being assessed and why. It is possible for systematic errors of measurement to creep into any scale measuring personality, mood, emotion or attitudes. Systematic measurement errors, such as social desirability and acquiescence, can also add 'noise' to scores on psychological scales.
