It goes without saying that the best type of specification is that which is most protective of the public interest and that includes both the producer and the consumer. Specifying the quality of concrete presents problems not found for most other building materials. Steel, lumber, masonry units, etc., can be tested before they are used: in the case of concrete, only its ingredients can be tested; the concrete itself cannot. Further, the user of concrete is jointly responsible with the manufacturer for its quality. A complete specification for concrete should cover the responsibilities of both parties. There are various approaches to describing the kind of concrete that is wanted but all of them fall in one of two broad categories. There is the prescription type of specification. The purchaser specifies the ingredients, the proportions, and the consistency of the concrete. Then there is the performance type. The purchaser specifies the strength and other necessary characteristics such as consistency, maximum size of aggregate, etc. Often time, the purchaser will order the performance type of specification with the guaranteed strength being the only limitation. However, a guaranteed strength specification presents many serious problems of administration and of assessment of penalties. The returns are not in until the structure is built! Nevertheless, strength tests frequently are made for the prescription type of specification. So, the prescription specification takes on many of the aspects of the guaranteed strength specification with out the attendant evils and difficulties of administration. Therefore, it seems desirable to establish classes of concrete based on minimum cement factor, maximum water ration, range in slump, maximum size of aggregate, range in aggregate proportions, etc., which have been shown to produce the required strengths under approximately minimum conditions. Then see to it that these limitations are adhered to! Strength tests should be made as frequently as necessary to check on compliance with these minimum requirements and the results should be interpreted in the light of a knowledge of variations reasonably to be expected.