We have carried out a meticulous time-space-analysis of the incidence of yellow fever in humans in Brazil from 1954 to 1972 and especially from 1973 to 1999. This study has added to our knowledge of the epidemiology of yellow fever and enabled us to redefine epidemiological zones and determine their geographical limits. The endemic area is located within the Amazon basin; here cases are scattered and generally limited in number. However, there are also "foci of endemic emergence" within this area, where cases are less rare, although occurrence remains irregular. The epidemic area is for the most part situated outside the Amazon basin, to the north east and particularly to the south. It has been divided into two parts according to whether the occurrence of yellow fever is cyclic or sporadic. The epidemics, which are all sylvatic, follow either a circular path (in the forest area) or a linear path (in forest-galleries of the savannah area). The study of the development of the 3 main epidemics (1972-74; 1979-82; 1986-92) in the cyclic emergence area showed that, on each occasion, the yellow fever virus appeared at a particularly active outbreak site located in the "serra dos Carajás", and from there, it followed the courses of the Tocantins and Araguaia rivers upstream, moving southwards during the "pre-epidemic phase" which may be visible due to the occurrence of a few cases, or may remain invisible. Subsequently the virus reached the emergence area, where it appeared in the form of epidemics. In this zone, it also followed privileged south-western pathways, moving from one hydraulic basin to another along the upstream courses of the rivers. Almost exactly the same pathways have been identified for each of the 3 epidemics studied. The distances travelled by the virus over a period of one year--when it goes rapidly--can reach several hundred kilometers. On the other hand, it may be stationary for a period of one or two consecutive years, occasionally three, remaining present in the area but infecting humans only rarely if at all. The virus occasionally leaves the cyclic emergence area and appears in the sporadic emergence area to the east, in the states of Bahia, Minas Gerais and São Paulo, and, as a consequence, moving onto other hydraulic basins. The small river basins in Maranhão and NorthWest states, as well as in the northern part of the state of Roraima also form part of the sporadic emergence area. The epidemics that occur here are directly linked to the endemic area and are only preceded by sometimes indiscernible epizootics and can consequently not be foreseen. Again the virus appears to use privileged pathways to reach the sporadic emergence areas where human and monkey populations are generally only partially immunised against yellow fever and where contact with mosquitoes is intense despite the fact it is limited in space and time, being restricted to the often narrow strip of trees along the water courses. Other routes used by the virus may be the Madeira, Xingu and Tapajós rivers, the scene of outbreaks observed in the state of Rondônia and in the north of Mato Grosso, where ongoing environmental changes are likely to result in an increasing number of outbreaks in the coming years. Since the discovery of the sylvatic cycle of yellow fever in 1933, not only the extent of the epidemiological areas has changed, but also their limits. Ecological modifications that are currently taking place in the Amazon basin, which is an endemic reservoir of the virus, will inevitably facilitate an increase the contact between humans and vectors. While more and more urban areas harbour populations of Aedes aegypti, the domestic and urban vector of yellow fever, it is particularly important to try to protect human populations living in emergence zones and epidemic areas and thus to prevent the arrival of the virus in towns via humans with viremia--in other words the much feared urbanisation of yellow fever in Brazil.