We investigated quantitative EEG in 19 diabetic children (mean age 14.2 (SD 1.4) years, mean HbA1c 9.8 (SD 1.2)% and 17 non-diabetic children (14.3 (1.1) years) during and after a gradual reduction in plasma glucose with the glucose clamp technique. The amplitudes of the EEG frequency bands at each glucose level were compared to the registration prior to hypoglycaemia. At plasma glucose approximately 4 mmol l(-1), a small increase in delta (p < 0.05) and theta (p < 0.01) amplitude appeared in the combined diabetic and control group. At approximately 3 mmol l(-1), and at glucose nadir, a further and widespread increase in low-frequency EEG activity was observed. In diabetic and non-diabetic subgroups with similar glucose nadirs, the diabetic children had more delta (p < 0.01) and theta (p < 0.01) activity, and more epileptiform activity (p < 0.05), than the non-diabetic children. In the non-diabetic subjects, but not in the diabetic subjects, the increase in delta and theta activity correlated with a hypoglycaemic symptom score (r = 0.75, p = 0.001 and r = 0.77, p < 0.0005, respectively). In conclusion, EEG changes are detectable already at plasma glucose approximately 4 mmol l(-1) in children. EEG deterioration during hypoglycaemia is more pronounced in diabetic than in non-diabetic children. The increase in cerebral disturbances is not accompanied by an increment in hypoglycaemic symptoms in diabetic children.