Sputum microscopy for acid-fast bacilli (AFB) is considered to be the most appropriate method for case-finding in a tuberculosis (TB) control programme. It is usually carried out by general technicians, often after minimal training. Quality control of their results therefore seems indispensable. The methods advocated for quality control are reviewed. Controls by culture leave too much uncertainty because of big differences in technical characteristics of the methods. Sets of smears sent out by a central laboratory can only be used to assess capability. Rechecking routine smears allows daily performance to be appraised and may be a strong motivation, but feasibility may be a problem. Based on our experience, we describe the technical requirements for cross-checking of routine smears. Counter-checking slides with discordant results is crucial for accurate assessments. A sample size should strike a balance between statistical accuracy and the man-power needed. Indicators for evaluation are proposed that allow discrimination of error gradings, to be used in a phased manner with priority at first being given to false negatives and false positives that pass the threshold for clinical decision-making. Estimates of critical values with suggestions about their interpretation are placed in the context of supervising TB laboratories.