OBJECTIVES: To investigate whether visual assessment or microscopic grading of the cellular composition of specimens is a better strategy to screen specimen quality for tuberculosis (TB) diagnosis. METHODS: About 2643 specimens were collected from TB suspects at the Federal TB centre in Pakistan. Specimens were classified as sputum or saliva visually and microscopically using the criteria proposed by McCarter and Robinson, Van Scoy, Geckler et al., Murray and Washington and Bartlett. The acid-fast bacilli (AFB)-positivity of specimens was also assessed. Results Despite being the least 'technical' and time consuming, visual assessment rejected the lowest proportion of AFB-positive specimens (0.3%). Most microscopic grading criteria, particularly those that considered the squamous epithelial cell count, rejected a large proportion of specimens (30-66%), of which a sizeable fraction contained AFB (6-12%). CONCLUSIONS: Our results indicate that visual assessment by trained technicians is more effective and suitable than microscopic grading for screening specimens for TB diagnosis. TB control resources could be better allocated to optimizing visual screening than investing in more 'strict' microscopic grading tools.