If negative interpretational bias causes emotional vulnerability, reduction of this bias should reduce proneness to anxiety. High trait-anxious volunteers were trained over four sessions to resolve descriptions of ambiguous events in an increasingly positive manner. This group subsequently made more positive interpretations of novel descriptions than did those in a test-retest control condition. Furthermore, trait anxiety scores reduced more in the trained group than in untrained controls. These results confirm earlier findings that modifying interpretation biases produces congruent changes in emotional vulnerability, and suggest a possible role for similar training methods in controlling pathological anxiety.