While one of the original underpinnings of the dopamine theory of schizophrenia was the paranoid psychosis which often develops during the binges or speed runs of chronic amphetamine addicts (and, more recently, in cocaine addicts), neurochemical studies of such drug abusers or from animals given continuous stimulants in an effort to model stimulant psychoses have not played a major role in the further evolution of this theory. One clear persisting alteration produced by continuous amphetamine is a neurotoxicity to dopaminergic innervations in caudate. Yet continuous cocaine administration apparently does not induce a similar neurotoxicity and this makes this effect a poor candidate for an underpinning of stimulant psychoses. However, it has recently been found that both continuous amphetamine and cocaine induce a strong pattern of degeneration which is highly confined to the lateral habenula and its principal output pathway, fasciculus retroflexus. This finding has led to a reconsideration of the role of these structures in psychoses. The habenula, as the chief relay nucleus of the descending dorsal diencephalic system (consisting of stria medullaris, habenula and fasciculus retroflexus), is an important link between limbic and striatal forebrain and lower diencephalic and mesencephalic centers. Studies of glucose utilization have consistently shown the habenula to be highly sensitive to dopamine agonists and antagonists. Lesions of habenula produce a wide variety of behavioral alterations. The dorsal diencephalic system has major and predominantly inhibitory connections onto dopamine-containing cells and it mediates part of the negative feedback from dopamine receptors onto dopamine cell bodies. It represents one of the major inputs in brain to the raphe nuclei and has anatomical and functional connections to modulate important functions such as sensory gating through thalamus, pain gating through central gray and raphe and motor stereotypies and reward mechanisms through substantia nigra and the ventral tegmental area. It is argued that alterations in these pathways are ideal candidates for producing the behaviors which occur during psychosis and that future considerations of the circuitry underlying psychoses need to include this highly important but relatively neglected system.