The interpersonal relationships with relatives of seriously ill patients may cause anxiety on the part of nurses and the need for adequate communication and self-control skills. To assess the efficacy of training nurses in self-control techniques and communication skills when they interact with relatives of seriously ill patients we planned a randomized, controlled trial in two parallel groups: an experimental group, with immediate training, and a control group, with training delayed for 6 months. We recruited 61 nurses from the nursing staff of a university hospital of 500 beds. The intervention consisted of training in relaxation, cognitive restructuring and some communication skills. The outcome variables were communication skills measured under simulated conditions using an observation instrument of our own, administered by observers masked with respect to the study groups, and the levels of state-anxiety measured with the self-assessment questionnaire the State-Trait Anxiety Inventory, under imaginary conditions. The two groups initially had similar scores in the scales of communication skills, and state and trait-anxiety. After intervention, compared with the control group, the experimental group showed significant improvements in the skills of listening, emphasizing, interrupting and coping with emotions (P < 0.05). State-anxiety levels did not show any changes. In conclusion, the joint training in self-control and communication improves some communication skills in nurses when they interact with relatives of seriously ill patients under simulated conditions.