Sixteen children (aged 3 to 12 years) participated in a 12 month crossover study comparing bedtime with teatime insulin injections in an endeavour to reduce morning hyperglycaemia. Blood glucose values were lower at lunch and at teatime on the later injection, but higher at bedtime and midnight. There was no overall change in glycosylated haemoglobin. Despite more frequent mild hypoglycaemic attacks, parents preferred the convenience of the later injection. Analysis of individual children's glycosylated haemoglobin values showed that those whose metabolic control improved on the later injection were younger and went to bed earlier, indicating that this regimen may have a place in the management of younger children with diabetes mellitus.