The aim of this study was to provide a comparative description of geographical variations and time trends in the population-based survival of European children with acute lymphoblastic leukaemia (ALL). Data on 13344 newly diagnosed children (0--14 years) with ALL were included in the EUROCARE study and were collected were collected by 34 population-based cancer registries (four comprising only childhood malignancies), operating in 17 countries (four in Scandinavia, two in Southern Europe, three in Eastern Europe, six in Continental Europe and two in the UK). Age-specific crude survival rates were estimated for boys and girls according to country for the period 1985--1989 and in adjusted form to attain comparability. Overall pooled and weighted rates were estimated as European standards. Children dead at diagnosis or diagnosed only through a death certificate were excluded. Geographical variation was also estimated by calculating the relative death rate with respect to the pooled overall European rate. After adjustment for age, gender and country, a Cox regression analysis was used to estimate time trends in survival. Survival was compared with that in the USA, Japan, Canada and Australia. During 1985--1989, the 1-year survival rate varied from 99 to 79%, the 5-year survival rate from over 80 to 56% (with the exception of Estonia; 34%; 95% confidence interval (CI) 20--52) among the various countries; the European weighted means were 90 (95% CI 87--93) and 72% (95% CI 69--75), respectively. Survival was particularly favourable in (south) Sweden, Finland, Germany and The Netherlands and rather unfavourable in Estonia and (surprisingly) France, where only 4% of its population was covered by the participating registries. Compared with the period 1978--1981, the hazard ratio for the period 1986--1989 decreased to 0.59 (95% CI 0.54--0.64) and -- in a smaller set of registries -- to 0.49 (0.45--0.55) for 1990-1992, an annual decrease in this rate of approximately 3.5%. During 1985--1989, the 5-year survival rates for European children were largely similar to those found in the USA, Canada and Australia, but markedly better than those in Japan. Higher survival rates were found for countries with 'good' access to centrally organised diagnostic and treatment facilities which stimulated 'aggressive' treatments according to a protocol. However, a subdivision according to risk profiles, e.g. according to the initial white blood cell count at diagnosis, could not be made and this might have explained partially the geographical differences in survival, because a positive association appeared between incidence at age 1--4 years and 5-year survival in most countries.