ABSTRACT
To survey the results of studies with objective tests, including performance and self-report, would be a Herculean task. Ellis (1946), and Ellis and Conrad (1948) published valuable critical surveys of personality questionnaires, showing how frequently they fail to agree with external criteria of the traits they are supposed to measure. Some of the weaker instruments may have dropped out of use since then, and some better ones have been constructed. But the trend of innumerable subsequent investigations is that they cannot be trusted. The median validity coefficient — probably around 0·4 — is by no means negligible, though hardly sufficient to justify using the tests for individual diagnosis. But the trouble is rather the variability — tests working fairly well in one investigation, not in another. The validities and stabilities of performance tests are, if anything, even poorer. Considering all the effort and skill that has gone into producing better tests, from the late 1920's onwards, it is surprising how little progress has been made. The object of this chapter is to bring out one of the main reasons why.
