The effect of intra-cerebrally infused compounds in animal models of anxiety were reviewed. A large body of evidence suggested that benzodiazepine agonists in different brain regions--including areas of the raphe, hypothalamus, periaqueductal gray, septum, hippocampus, and amygdala--produce reasonably consistent anxiolytic effects in a variety of animal models. However, evidence regarding the effects on anxiety of 5-HT1A agonists, 5-HT2 compounds, and 5-HT3 antagonists was somewhat less extensive, both anatomically and behaviourally, and more complex. For example, establishing receptor specificity for 5-HT ligand effects was often complicated by the lack of 'silent' and/or selective antagonists. Neuropeptides had significant effects on anxiety, but these were shown in a smaller number of animal models and in a limited number of brain regions. Regardless of the compounds tested, however, there seemed to be a surprising number of double dissociations (brain site by behavioural test). In fact in some instances, different fear reactions appeared to be controlled by distinct receptor subpopulations within particular parts of the limbic system. These results suggest that the neural control of anxiety might be analogous in organization to sensorimotor systems, i.e., anxiety is controlled by complex systems of multiple, distributed, parallel pathways.