This study investigated the relationships of assessed social support and experimentally provided social suppprt to anxiety and task performance in different stress situations. A 2 (assessed support: high- and low-SSQN) SMITH×2 (high and low stress) ×2 (support manipulation: provided and not-provided) mixed-design was used. Dependent variables were pulse rates as a physiological anxiety index, STAI as a psychological anxiety index, and correct word puzzle solutions. The subjects were 51 college students. There was a significant interaction between support manipulation and stress condition on physiological anxiety. The provided support in a high stress situation tended to reduce anxiety. This tendency was consistent with the idea of buffering effects against stress. The provided support in the high-SSQN subjects in the high stress situation had a negative effect on task performance. Suggestions as to why the provided support in the low-SSQNs did not have facilitative effects on task performance are given