Antibodies that immunoprecipitate (125)I-alpha-dendrotoxin-labelled voltage-gated potassium channels extracted from mammalian brain tissue have been identified in patients with neuromyotonia, Morvan's syndrome, limbic encephalitis and a few cases of adult-onset epilepsy. These conditions often improve following immunomodulatory therapies. However, the proportions of the different syndromes, the numbers with associated tumours and the relationships with potassium channel subunit antibody specificities have been unclear. We documented the clinical phenotype and tumour associations in 96 potassium channel antibody positive patients (titres >400 pM). Five had thymomas and one had an endometrial adenocarcinoma. To define the antibody specificities, we looked for binding of serum antibodies and their effects on potassium channel currents using human embryonic kidney cells expressing the potassium channel subunits. Surprisingly, only three of the patients had antibodies directed against the potassium channel subunits. By contrast, we found antibodies to three proteins that are complexed with (125)I-alpha-dendrotoxin-labelled potassium channels in brain extracts: (i) contactin-associated protein-2 that is localized at the juxtaparanodes in myelinated axons; (ii) leucine-rich, glioma inactivated 1 protein that is most strongly expressed in the hippocampus; and (iii) Tag-1/contactin-2 that associates with contactin-associated protein-2. Antibodies to Kv1 subunits were found in three sera, to contactin-associated protein-2 in 19 sera, to leucine-rich, glioma inactivated 1 protein in 55 sera and to contactin-2 in five sera, four of which were also positive for the other antibodies. The remaining 18 sera were negative for potassium channel subunits and associated proteins by the methods employed. Of the 19 patients with contactin-associated protein-antibody-2, 10 had neuromyotonia or Morvan's syndrome, compared with only 3 of the 55 leucine-rich, glioma inactivated 1 protein-antibody positive patients (P < 0.0001), who predominantly had limbic encephalitis. The responses to immunomodulatory therapies, defined by changes in modified Rankin scores, were good except in the patients with tumours, who all had contactin-associated-2 protein antibodies. This study confirms that the majority of patients with high potassium channel antibodies have limbic encephalitis without tumours. The identification of leucine-rich, glioma inactivated 1 protein and contactin-associated protein-2 as the major targets of potassium channel antibodies, and their associations with different clinical features, begins to explain the diversity of these syndromes; furthermore, detection of contactin-associated protein-2 antibodies should help identify the risk of an underlying tumour and a poor prognosis in future patients.