The impact of the human immunodeficiency virus (HIV) pandemic on childhood tuberculosis (TB) is unclear because of inconsistent and often contradictory findings in different types of studies. We review the evidence which supports or refutes the likelihood that HIV infection in children predisposes them to TB, and conclude that, on balance, HIV during infancy increases the risk of developing TB. Surveillance shows an association between rising TB rates among children and the HIV epidemic in some parts of the world. A number of cross-sectional studies which have taken children with TB as their starting population, have yielded high rates of association with HIV (11%-64% HIV prevalence). Similarly, cross-sectional studies of hospitalised children with HIV show that many also have TB. These rates of association are all over-estimated because of the uncertainty of diagnosis of TB. Birth cohorts of perinatally HIV-infected infants and children prospectively followed up for a few years have generally failed to detect a higher incidence of TB than anticipated. The few TB cases identified in these cohorts were usually over 15-18 months of age. In acute progressive lung disease there is no excess of TB in HIV-infected over non HIV-infected children. These inconsistencies are discussed and attributed mainly to study design and statistical artefact. However, maternal factors in HIV-positive women which might affect transmission of TB to their babies are assessed, and infant immunoparesis due to HIV which may adversely influence resistance to TB is considered.