More than a decade after the first description of HIV DNA in the nervous system the pathophysiology of HIVD remains largely enigmatic, with data supporting a number of potential mechanisms for the development of neuronal dysfunction. Nevertheless, a few key findings have considerable support in the literature devoted to this subject: 1. HIV dementia is caused by HIV itself; no other pathogen has been consistently found in the brains of patients with HIVD. 2. In comparison with other viral encephalopathies, there appears to be a significant discordance between the amount of virus being produced in the brains of patients with HIVD and the degree of neurological deterioration. 3. The key cell types responsible for viral production within the CNS are the resident macrophages or microglial cells. 4. Other elements within the CNS, particularly astrocytes, are probably infected with HIV as well, but all of these infections are highly restricted in terms of production of virus or viral structural proteins. 5. At least one component of the pathogenesis of HIVD may be the generation of neurotoxins by infected microglia, although the type of neurotoxin, and the specific compound most likely to be involved, are quite controversial. Advances with combination antiviral therapy have successfully reduced plasma viral load in a high proportion of individuals, leading to the speculation (previously almost heretical) that it may be possible to eradicate HIV completely from the systemic immune system. If that were the case, potential "sanctuary" sites such as the immunologically protected CNS might remain as important reservoirs for reseeding of lymphoid tissues. Microglia may be particularly suited for this purpose because they are long lived, can produce HIV for several weeks (at least in culture), and they are apparently relatively immune to virus-induced cytopathology such as syncytium formation. One can speculate about several scenarios resulting from the continued presence of replication-competent HIV within brain. In the worst case, a smoldering infection of the nervous system could lead to neurological deterioration without reinfection of systemic immune cells. The epidemiological data indicating that HIVD is a disease primarily associated with immunodeficiency suggest that the systemic immune system plays a role in maintaining virus residing within the CNS under control. Thus it is quite possible that this scenario would not occur for many years after the systemic infection is controlled. Alternatively, virus could be transported from the CNS by circulating lymphocytes and monocytes and reinfect systemic organs. This would necessitate restarting therapy for those individuals who were previously thought to be cured, but presumably virus within the CNS would not have developed resistance to antivirals. In either case, the techniques currently available do not permit an accurate assessment of CNS HIV load in living people, and this question will remain unanswered until antivirals are discontinued in a few individuals with persistently negative tests for systemic virus. In addition to this most critical question, the relationship between viral levels and HIVD is largely unexplored, as is the possibility that some strains are particularly virulent or neuroinvasive. Furthermore, the potential contribution of host genotype in the development of dementia is unknown. In view of the strong influence of major chemokine receptor (CCR5) truncations on HIV replication, it is entirely possible that more discrete genetic polymorphisms have a subtle effect on either brain invasion or virulence.