Antibodies to acetylcholine receptor (AChR) were measured in a group of patients with myasthenia gravis (MG), some of whom had previously been classified as 'antibody negative' using the standard anti-AChR radioimmunoassay (RIA). AChR antibodies were measured using the rosetting assay, a new detection method which utilizes protein A-coated red blood cells and live BC3H-1 cells, a murine cell line which expresses muscle nicotinic AChR. The results of the rosetting assay were compared with those obtained in the anti-AChR RIA. 76% of all myasthenic sera tested showed rosetting at titers higher than any of the control sera (from patients with non-myasthenic neurologic disease and normal individuals). Of the myasthenic patients previously classified as 'antibody negative' in the RIA using human AChR, 71% demonstrated positive rosetting. There was no correlation between the anti-AChR antibody titer obtained in the rosetting assay and that obtained in the RIA using either human or denervated rat AChR. The results suggest that the rosetting assay may measure a subpopulation of antibodies that differs from those detected in the RIA.