ABSTRACT
A complaint frequently voiced in the two previous chapters is that current personality tests are low in validity. Why is this so, and what, indeed, do we mean by validity? Various definitions have been proposed, but they mostly imply, in essence, that a test is valid in so far as it indicates or measures whatever it is supposed to. We intend to argue a somewhat more unorthodox view — that a test measures only itself, but that it is valid in so far as it can be shown to correlate with other observable behaviour. That is, its validity lies in the inferences we are entitled to make from it. Before this position is reached, it is necessary to discuss the several types of validation which are customarily distinguished.
