Attempts to devise an accelerated test that will predict the absolute or even the relative abilities of metals and coatings to resist atmospheric corrosion continue to be unsuccessful. For example, a very specifically described salt spray test, proposed to assess the durability, under various conditions of use, of components for radio communication equipment, may not be able to measure anything but large differences at any level of quality. It is recommended that any single test of this sort be abandoned in favor of a number of tests, each designed especially to measure a quality of interest. Results of salt water erosion tests on disk specimens of propeller materials illustrate the principle that tests of this type are more reliable in predicting relative performance of materials tested in the same way than in yielding quantitative data on rates of attack that may occur in service. In many tests, for example, galvanic corrosion and jet impingement tests, results are so influenced by the dimensions of the test specimens that it is impossible to give the data any quantitative significance. Experience with a number of other tests, among which are qualification tests for stainless steels, crevice corrosion tests, stress corrosion tests, corrosion fatigue tests, tests of corrosion by deicing salts, laboratory corrosion tests, and soil corrosion tests, contributes further to the conclusion that what is needed most is a greater documentation of the relationships beween performance in tests and performance in the services towards which tests are directed.