To return to the exercise , close this browser window.

Reliability and Validity

The Reliability of a measurement or observation refers to its repeatability. If you measure or observe the same thing twice, how close are the two measurements or observations? If they are the same or very similar, then they are highly reliable. If they are quite different, then reliability is low. For example, to be a reliable IQ test, retesting an individual with an equivalent form will give about the same score as did the first test. The correlation (measure of relationship that can go from 0 [none] to 1 [perfect]) between scores on different versions of an IQ test (WAIS) administered several weeks apart is about .85. The average difference between the two scores is about 5 points (100 is the population average).

Figure 1 shows the relation between pairs of scores that 10 individuals would get on a relaible test (left) and on an unreliable test (right). Each line connects the pair of scores for each individual measured about six weeks apart. Notice that the lines connect similar values in the left (high reliability) panel, whereas they connect rather different vaules in the right (low reliability) panel.

The validity of a measurement or observation refers to whether the measurement or observation really reflects what you think it reflects. Imagine that you use a reliable IQ test thinking that it measured personality. It would not be a valid test, because it didn't measure what you thought it did, even though it was a highly reliable test (it gives similar numbers when retested.