Sixteen experienced and 15 inexperienced observers watched videotapes showing participants of a back school program perform simple tasks to assess the functional adequacy of their posture. They had received basis or more extensive observer training. Intraclass correlation coefficients were computed as a measure of intra-and interobserver reliability. When individual judgments were compared, it was found that the judgments of the experienced observers were more reliable than those of the inexperienced observers. The additional observers training did not lead to a significant augmentation of reliability. By selecting the most reliable items two alternative measures of overall posture were constructured whose reliability coefficients ranged fromr=0.89 to 0.93 for the group of experienced observers. The reliability of the judgments by the more extensively trained inexperienced observers was also satisfactory. The observation method presented here can thus be regarded as a reliable and potentially valid instrument for assessing the outcome of back school programs.