A collaborative study was conducted in 14 laboratories to evaluate the performance of the hydrophobic grid membrane filter method against the official final action method, 46.016, for enumeration of colif orms in foods. Twelve artificially contaminated samples of nonfat dry milk and canned custard were analyzed by each laboratory. The random errors (Sr2) associated with the hydrophobic grid membrane filter method were significantly lower than those of the official method, and the counts obtained by the new method fell within the 95% confidence interval of the reference method. The colif orm counts obtained by tiie hydrophobic grid membrane filter method did not differ significantly from those obtained by the reference method as measured by a 3-way analysis of variance. The new method has been adopted official first action.