Over the past several years, major advances have been made in understanding the pharmacology of anxiety, involving three broad classes of experimental approach. One approach studies the mechanism of action of drugs that are known to treat anxiety clinically, such as the benzodiazepines. A second approach uses various animal models of fear or anxiety that are sensitive to known anxiolytic drugs, to see if they will detect new compounds. A third approach involves describing the neural pathways and neurotransmitters that are active in a state of fear or anxiety; importantly, this approach is not derived from the mechanisms of known anxiolytics. In this review, Michael Davis describes such a 'neural systems' approach to the study of fear or anxiety that uses the paradigm of fear-potentiated startle.