Over the last decade or so there has been increasing recognition of the importance of measuring antibodies to neuronal proteins in the diagnosis and management of disease. Such antibodies can be classified into two main groups. First, there are those that are markers for a disease process, such as the paraneoplastic antibodies against Hu, Yo, Ma2, but which are not thought to cause the disease. This is because their targets are cytoplasmic or nuclear proteins that are not exposed on the surface of the neurons, and because the patients do not respond to immunotherapies such as plasma exchange, intravenous immunoglobulins or corticosteroids. Moreover, the antibodies do not necessarily associate with specific syndromes; for instance, Hu antibodies can be found in a wide range of disorders from neuronopathy to limbic encephalitis. By contrast, the second group includes those that are not only diagnostic biomarkers but are also pathogenic (or likely to be). These bind to important membrane proteins, their epitopes exposed on the outside of the target cells; they are usually syndrome specific, e.g. the acetylcholine receptor antibodies in myasthenia gravis and, most importantly, the patients do well with treatments that reduce the levels of the antibodies. Where do glutamic acid decarboxylase (GAD) antibodies, that are found in diabetes and stiff-person syndrome (SPS), fit into this scheme? GAD is a cytoplasmic enzyme …