Human plague in the Western Usambara Mountains in Tanzania has been a public health problem since the first outbreak in 1980. The wildlife reservoir is unknown and eradication measures that have proved effective elsewhere in Tanzania appear to fail in this region. We use census data from 2002 and hospital records kept since 1986 to describe the temporal, spatial and demographic variation in human plague. A seasonal peak in cases occurs from December to February with the numbers of cases during this peak varying between 0 and 1150. Variation in incidence, calculated for each village as the mean number of cases per thousand inhabitants per year, indicates that human plague is concentrated around a group of three neighbouring, relatively isolated, high-altitude villages; Nywelo, Madala and Gologolo. However, there was no evidence that these villages were acting as a source of infection for the remainder of the focus. The likelihood of becoming infected with plague is highest between the ages of 5 and 19 and lowest for adult men. This was most clear in the ward encompassing the three high-incidence villages where the risk of plague among children aged 10–14 was 2.2 times higher than for adults aged 30–34, and among adults aged 30–34, the risk was 2.4 times higher for women than men.