Two hundred and forty blood glucose readings were made on 5 Glucochek machines by 2 different operators. The results on all machines correlated poorly with equivalent results from an automatic glucose analyser with major discrepancies at all glucose concentrations. There was a wide gap in performance between the 2 operators. Analysis of variance suggested that errors were not only due to inter-operator and inter-machine differences. Possible sources of error were shown to include time of reaction and the amount of blood on Dextrostix. Dextrostix colour changes measured by eye were as accurate as those by Glucochek without the rogue values found at higher glucose concentrations with Glucochek.