In 53 children with acute lymphoblastic leukaemia (ALL), initial handmirror cell (HMC) count among lymphoblasts was studied in relation to the occurrence of a relapse in the central nervous system (CNS), taking into account the white blood cell count (WBC) and the immunological phenotype. The children were followed for a period limited by (1) the first CNS relapse, (2) death or (3) the closing day of this study. The median follow-up period was 30 months, range 2–106 months. HMC counts were available in bone marrow smears of 41 children and in cytospins of 35 children. Cytospins proved to give more reliable and consistent results than bone marrow smears. In the 35 ‘cytospin’ children no CNS relapses occurred in the group of 16 children with non-T-non-B-ALL and HMC counts above 10%. However, in the group of 10 children with non-T-non-B-ALL and HMC couts below 10%, and in the group of 9 children with T-ALL (HMC ≤ 11 %), 6 and 5, respectively, got a CNS relapse. The CNS relapse-free period was not significantly different between these last two groups, whereas both groups did differ significantly from the group mentioned first (p < 0.01). This was not found in bone marrow smears of 41 children, presumably because of the inaccurate counting results. A low initial HMC count in cytospins is associated with an increased risk for the occurrence of a CNS relapse in children with ALL. This prognostic factor seems to be independent of other prognostic signs such as immunological phenotype and high WBC.