Paraneoplastic limbic encephalitis (PLE) is a rare disorder characterized by personality changes, irritability, depression, seizures, memory loss and sometimes dementia. The diagnosis is difficult because clinical markers are often lacking, and symptoms usually precede the diagnosis of cancer or mimic other complications. The frequency of antineuronal antibodies in patients with PLE has not been investigated. We examined the neurological symptoms and the causal tumours in 50 patients with PLE to determine the utility of paraneoplastic antibodies and other tests. The diagnosis of PLE required neuropathological examination or the presence of the four following criteria: (i) a compatible clinical picture; (ii) an interval of <4 years between the development of neurological symptoms and tumour diagnosis; (iii) exclusion of other neuro-oncological complications; and (iv) at least one of the following: CSF with inflammatory changes but negative cytology; MRI demonstrating temporal lobe abnormalities; EEG showing epileptic activity in the temporal lobes. Of 1047 patients with neurological symptoms, whose sera or CSF were examined for paraneoplastic antibodies, 79 had the presumptive diagnosis of limbic encephalitis, dementia, cognitive dysfunction, or confusion. Fifty of these patients fulfilled our criteria for PLE. Pathological confirmation was obtained in 12 patients. The commonly associated neoplasms were of the lung (50%), testis (20%) and breast (8%). Neurological symptoms preceded the cancer diagnosis in 60% of patients (by a median of 3.5 months). Twenty-five of 44 (57%) patients with MRI studies had signal abnormalities in the limbic system. Thirty (60%) patients had antineuronal antibodies (18 anti-Hu, 10 anti-Ta, 2 anti-Ma), and 20 were antibody-negative or had uncharacterized antibodies (n = 4). The combination of symptoms, MRI findings and paraneoplastic antibodies established the diagnosis of PLE in 78% of the patients. Patients with anti-Hu antibodies usually had small-cell lung cancer (94%), multifocal neurological symptoms (78%) and a poor neurological outcome. Patients with anti-Ta (also called anti-Ma2) antibodies were young men with testicular tumours (100%), frequent hypothalamic involvement (70%) and a poor neurological outcome. In the group of patients without anti-Hu or anti-Ta antibodies, the tumour distribution was diverse, with cancer of the lung the most common (36%); 57% had positive MRI. Fifteen of 34 (44%) patients with a median follow-up of 8 months showed neurological improvement. Treatment of the tumour appeared to have more effect on the neurological outcome than the use of immune modulation. Improvement was observed in 38% of anti-Hu patients, 30% of anti-Ta patients and 64% of patients without these antibodies.