The present study describes the responses of cortisol, prolactin and growth hormone (GH) to emotions elicited during sessions in which an hypnoidal state was induced. The purpose of the study was to provide answers for the following questions: 1) Do sessions with an emotional content have more hormonal surges than baseline, relaxation-only, sessions? 2) Does the induction of a fantasy of pregnancy and nursing elicit a prolactin response? 3) Are there any associations between surges of different hormones? 4) Are hormonal responses related to the intensity, type, or mode of expression of the emotions? For this purpose, thirteen volunteers and twelve patients with minor emotional difficulties were studied during sessions under hypnosis. The period of observation lasted for about three hours. Heart rate (HR), skin conductance (SC) and vagal tone (VT) were monitored. Serum cortisol, prolactin and growth hormone were sampled every 15 minutes. The volunteers had three types of sessions- "blank", consisting of relaxation only (12 sessions), "breast feeding", in which a fantasy of pregnancy and breast feeding was induced (12 sessions) and "free associations" in which the subjects were encouraged to evoke experiences or feelings (17 sessions). The patients had only sessions of free associations (38 sessions). Sessions of free associations had more hormonal surges than "blank" and "breast feeding" sessions. This was true for cortisol (8/17 v.3/24; p < 0.03), prolactin (7/17 v. 3/24; p < 0.05) and GH (9/17 v. 4/24; p < 0.02). During the 55 sessions of free associations (volunteers plus patients) there were 32 surges of cortisol, 18 of prolactin and 28 of GH. Cortisol and prolactin surges were negatively correlated (p < 0.03). GH had no significant association with either cortisol or prolactin. Visible emotions were positively associated with GH surges (p < 0.05). but not with cortisol or prolactin. Cortisol surges were correlated positively with evocations of real events (p < 0.01) and negatively with evocations containing defensive elements (p < 0.01). Cortisol correlated positively with shock and intimidation (p < 0.02) and negatively with rage (p < 0.04). The AUC of the cortisol peaks during shock and intimidation was significantly higher than that of the pool of all other cortisol peaks (12.4 micromol x min x l(-1) v. 7.1 micromol x min x l(-1); p < 0.005). Rage had a marginally significant positive association with prolactin surges (p=0.07). The distribution of GH surges did not show any significant association with types of emotions. The present study provides evidence that cortisol, prolactin and GH respond to psychological stress in humans. However, they are regulated differently from one another. Cortisol and prolactin surges appear to be alternative forms of response to specific emotions. GH surges depend on the intensity of the emotion, probably as a consequence of the associated muscular activity. The current paradigm of stress, implying corticotrophin-releasing hormone (CRH) as the initial step of a cascade of events, is insufficient to account for the diversity of hormonal changes observed in psychological stress in humans.