My first reaction to being asked to speak on this subject was that it is the user and not the producer of model data whose views we want to hear, and I am glad to say that Mr. Ellis has agreed to co-ordinate with me, so that I have some inside knowledge to draw on, if I venture to say what a designer does.

I wish in the main, however, to take the title in rather a different way, and to discuss in what sense model tests can be expected to give results useful to the designer.

It is well known that, under certain conditions, duralumin and other wrought aluminium alloys are liable to suffer intercrystalline attack, such corrosion being frequently accompanied by little visible corrosion product on the surface of the material. The necessity for careful temperature control during the solution heat treatment of duralumin and the beneficial influence of a rapid rate of cooling after such heat treatment on the corrosion-resistance of this material are now too widely appreciated to need emphasis here.