Glucose reflectance meters equipped with memories were used in 21 insulin-dependent diabetic patients to assess reliability of self blood glucose monitoring records. Results reported in logbooks showed widespread discrepancies when compared with corresponding meter-held records. Discrepancy rates of at least one in five were observed in 11 patients. Unreliable reporting was usually recognized by clinical staff, and associated with high glycosylated haemoglobin levels and poor technique in measuring blood glucose at home.