At present, there are three major surgical approaches to Parkinson's disease (PD): (1) Ablative surgery (i.e. pallidotomy, thalamotomy); (2) deep brain stimulation (DBS) of the thalamus, internal globus pallidus (GPi) and subthalamic nucleus (STN); and (3) grafting fetal mesencephalic cells into the striatum. As a result of increasing understanding of the pathophysiology of the basal ganglia and the demonstration of surgical alleviation of experimental parkinsonism, surgery has regained a paramount importance in the management of PD. The aim of pallidotomy and DBS is to reduce the excessive inhibitory output from the GPi and substantia nigra reticulata (SNr). Pallidotomy and DBS of the STN or GPi aim to reverse the pathophysiological consequences of dopamine deficiency in PD, and should be considered entirely symptomatic treatments. The ideal candidates for pallidotomy are young patients in good general health in whom dyskinesias are the main reasons for disability. Patients with severe bilateral problems uncontrollable with present pharmacological tools are candidates for DBS. As yet, there are no formal data to help decide how to choose between GPi and STN stimulation. In our practice, patients are allocated to GPi stimulation when 'on' dyskinesias are extremely severe. In most other instances, we prefer to perform STN stimulation. At present there is almost no reason to decide for the thalamic stimulation since tremor is equally arrested by STN stimulation, which in addition improves all other features of PD. Equally the only indication for thalamotomy would be a patient with long-standing tremor as the main clinical manifestation, which can not be controlled with drugs. The proportion of patients in whom the thalamus will be the preferable target for either DBS or thalamotomy is small (less than 5%). Grafting aims to repair the nigrostriatal pathway and restore dopaminergic function in the striatum. In the future implants containing not only dopaminergic cells but also growth factors and a variety of other substances could become a method to not only functionally compensate the biochemical abnormalities of PD but also to arrest its progression. This technique is limited to a few centres around the world owing to the technical, logistical and ethical problems of obtaining and handling embryonic cells. At present, grafting of dopaminergic cells is perhaps best suited for patients with young-onset PD (less than 45 years old) who are at high risk of developing complications within a short time of beginning pharmacological treatment and in whom the idea of making lesions or implanting electrodes into the brain for decades seems less appealing. Consideration of surgery in any given patient should be weighed against the risks (about 1% mortality and 2-6% of severe morbidity-hemiplegia, cognitive deficit, speech problems, etc.) associated with these techniques. The development of better imaging methods and the growing expertise of multidisciplinary teams will undoubtedly make surgery for PD safer and more effective in the future.