To define further the clinical importance of cytogenetic analysis in acute lymphoblastic leukaemia (ALL) a prospective study was performed on 139 unselected children. Analyses were considered adequate in 104, of whom 35 were normal and 69 had clonal abnormalities. Abnormalities were categorised according to banded chromosome analysis as well as chromosome count. Karyotypes were correlated with clinical and laboratory features at diagnosis and with survival. Of the successful analyses, thirty five (34%) children had no abnormalities; this group contained an excess of T cell disease. Twenty five (24%) had a "characteristic" hyperdiploid karyotype and as a group had lower presenting white counts, a tendency to CD10, and periodic acid schiff positivity of the blast cells and smaller spleens. None was an infant and only one was over 10 years old. Seven (7%) children with t(9; 22), t(8; 14), or t(4; 11) translocations were grouped together as "specific" translocations. Collectively they had a significantly worse prognosis than the remainder. Nine children developed central nervous system relapse, six of whom had either t(4; 11) or abnormalities of 9p or 19p. A descriptive classification taking into account chromosome bonding pattern is cytogenetically more appropriate and may be more clinically useful than grouping children simply by chromosome number. As knowledge and techniques improve, the classification of cytogenetic abnormalities in ALL will need to be kept under frequent review.