In a learning experiment 10 Ss as receivers judged looking signals of a sender. Providing feedback about the real direction of gaze produced learning in the direction of an improvement of discrimination performance; the improvement was independent of visual acuity of receivers and was not correlated to extraversion and neuroticism. Performance at the beginning was poorer and learning progress better for fixation points which were further away from the face. Results are discussed with reference to improving accuracy of observers on the variable ‘looking behavior’, important in studies on nonverbal communication. A pre-training of observers which allows explicit feedback about the real direction of gaze is proposed.