Problems in diagnosing sporadic inclusion body myositis may arise if all clinical features fit a diagnosis of polymyositis, but the muscle biopsy shows some rimmed vacuoles. Recently, immunohistochemistry with an antibody directed against phosphorylated neurofilament (SMI-31) has been advocated as a diagnostic test for sporadic inclusion body myositis. The aims of the present study were to define a quantitative criterion to differentiate sporadic inclusion body myositis from polymyositis based on the detection of rimmed vacuoles in the haematoxylin-eosin staining and to evaluate the additional diagnostic value of the SMI-31 staining. Based on clinical criteria and creatine kinase levels in patients with endomysial infiltrates, 18 patients complied with the diagnosis of sporadic inclusion body myositis, and 17 with the diagnosis of polymyositis. A blinded observer counted the abnormal fibres in haematoxylin-eosin-stained sections and in SMI-31-stained sections. The optimal cut-off in the haematoxylin-eosin test was 0.3% vacuolated fibres. Adding the SMI-31 staining significantly increased the positive predictive value from 87 to 100%, but increased the negative predictive value only to small extent. We conclude that (1) patients with clinical and laboratory features of polymyositis, including response to treatment, may show rimmed vacuoles in their muscle biopsy and that (2) adding the SMI-31 stain can be helpful in differentiating patients who respond to treatment from patients who do not.