can be gathered, the choice between alternative models often remains a matter of personal taste. But taste is unsuitable as an arbiter of the value of scientific theories; and, if better touchstones cannot be found, the current vogue for cognitive theorising will go the way of its forerunner a century ago. To provide such a touchstone there is at present no alternative to the test of experiment, nor is there likely soon to be one. Thus we need a way of specifying the inside of the black box which will be open to experimental test; and if this cannot be done by behavioral observations alone, then these must be supplemented by other forms of observation. It is in this context that we do well to remember that cognition is a product of the brain. For, if we specify the processes which control behavior and the brain mechanisms which mediate these processes, it becomes possible to test the resulting theory by means of observations on the brain as well as on behavior; and, more powerfully still, by means of joint observations of both these kinds (for some examples, see Gray, 1982b). It is such a joint specification of cognitive process and mediating brain mechanism that my theory of anxiety, in its second stage ( Gray, 1982a; Gray, 1982b), attempts to construct. I shall not try here to summarize the data bases (stemming from anatomical, physiological, pharmacological and behavioral experiments) on which this theory rests (see Gray, 1982a), but instead present it as a fait accompli.