A proficiency test to assess the capabilities of laboratories to determine melamine in a milk powder and a baking mix, representing starch-containing foods like bread and biscuits, was carried out in January 2009. The need for such an interlaboratory comparison arose from a health scare in China about melamine-tainted powdered milk in the second half of 2008. Laboratories in 31 countries, including Australia, China, India, Japan, New Zealand and the USA, and 21 of the 27 Member States of the European Union participated and reported back 114 results for the milk powder and 112 for the baking mix test materials. The reported results were compared to reference values determined by exact-matching double isotope dilution mass spectrometry. The so-determined assigned values were 10.0 +/- 0.6 mg/kg melamine in the milk powder and 3.18 +/- 0.17 mg/kg melamine in the baking mix. A coverage factor k of 2 was applied to calculate the expanded uncertainties. Three quarters of all reported results for both materials had associated z scores which were satisfactory (z <or= |2|). Of the reported results, 90% was accompanied by a measurement uncertainty statement, and the majority of the measurement uncertainty ranges were reasonable. A number of laboratories were found to underestimate their measurement uncertainties. Methods that involved the use of stable-isotope-labelled melamine were shown to be clearly advantageous with regard to the accuracy of the results. However, no significant influence by other method parameters could be identified.