Diagnosis of tuberculosis in a field setting depends on the quality of specimens submitted for smearmicroscopy. Macroscopic assessment (sputum or saliva) of the specimen for suitability for further examination is common practice in routine care. We examined whether macroscopic assessment could correctly identify sputum specimens based on four published algorithms using microscopic features in the setting of active case finding in a community survey. The study included 901 randomly selected adults who reported cough for 3 weeks or more in the national tuberculosis prevalence survey in Bangladesh. A single specimen of each was assessed with microscopy and microscopy (Gram-stain) to classify it as either sputum or saliva. The primary outcome was the agreement between the two assessment methods (Kappa statistic) From 901 specimens, 561 (62%) were macroscopically classified as saliva and 340 (38%) as sputum. From these, 888 Gram-stained slides could be examined for microscopic features. The agreement between the macroscopic assessment with any of the four microscopy algorithms for sputum was very poor (all Kappa’s below 0.1). While macroscopic assessment of submitted specimens might be of value in routine care, it is not warranted in a setting of active case finding in a community survey. Submitting a specimen in the first place should be the primary goal in this setting. © Banu et al.