Previous studies have shown that in comparison to healthy people, patients with social anxiety (SA) identify fewer emotions of happiness in vocal expressions.

1. (1) To repeat previous studies on emotion recognition in SA patients;

2. (2) To examine the effect of training on emotion recognition in SA patients.

1. (1) To examine the effect of training in emotion recognition of non-verbally vocal musical improvisations on the ability of SA patients to identify happiness in verbal vocal spoken language.

2. (2) To create a preliminary procedure for improving this ability in SA patients. Methods: 41 SA patients and 39 healthy controls aged 24–40, were examined. SA diagnosis was conducted according to the norms of the Liebowitz (1987) questionnaire. Half underwent an intervention that focused them on happiness recognition in non-verbally vocal musical improvisations, and half did not. The four groups were then compared on level of precision of emotion recognition (happiness, fear, anger, sadness and surprise) in spoken language sentences.

A Multivariate analysis of variance showed that SA patients significantly identified fewer emotions of happiness in vocal expressions than healthy controls. SA patients who were trained demonstrated a similar precision level of happiness recognition in woman spoken language sentences as healthy controls (ps < 0.05).

Our findings demonstrate that short exposure (20 min) to non-verbally vocal musical improvisations immediately improves the ability of SA patients to recognize happiness in spoken language. Future studies can refine our procedure and examine its impact on SA patients over longer time periods (e.g., months).