A stressful interview was conducted in 52 older people to determine whether a stress provoked change in growth hormone (GH) is correlated with anxiety and "defensiveness." The interview was referred to events surrounding the potential loss of a spouse hospitalized two months previously: in 29 instances, the spouse had died before this interview. GH was measured before and, in 23 subjects, during the interview. After the interview, in addition to GH, anxiety was measured using the Taylor Manifest Anxiety Scale (MAS), and defensiveness using the Crowne Marlowe Social Desirability Scale (SD). An elevation of at least 2.0 ng/ml in GH was used to classify subjects as GH responders (Rs) and nonresponders (NRs). Although the Rs and NRs had similar mean scores on defensiveness (SD), the Rs scored higher on anxiety (MAS) (14.9 vs. 10.2, p less than 0.05). Using a median cutoff with anxiety (MAS) alone, only 59% of the Rs were correctly classified as Rs. However, combining both scales in a discriminant function analysis (DFA) improved the ability of either scale alone to distinguish the Rs from the NRs. Using the DFA, 73% of the 29 NRs were correctly classified and 70% of the 23 Rs were correctly classified (p less than 0.005). With higher levels of defensiveness (SD), relatively low levels of anxiety were needed to provoke a GH response. This finding supports previous studies of GH response to stress and provides a multivariate model for the interaction of anxiety and "defensiveness."