The islet cell antibody test is increasingly used to predict insulin-dependent diabetes before it is clinically apparent. Four international workshops on the test have shown wide variations in reproducibility, sensitivity and specificity. This illustrates the need for a more clearly defined method. We report a technique which offers both increased sensitivity and reproducibility. Using this method 94% of newly diagnosed diabetic children have ICA levels greater than or equal to 10 international units, whereas only 1.5% of normal children are positive at his level, and 7% of first degree relatives (aged less than 20 years) of IDDM patients.