THE recent publication1 of an examination of Svedberg's theory of molecular weight distribution, in which the theory is considered to be invalid, suggests that a treatment, similar in some of its details, might be of interest. This was prepared in 19442, and the conclusions agree with those of Johnston, Longuet-Higgins and Ogston1, and with those of Bull3. It seemed clear that the explanation of the imperfection of his theory given by Svedberg4, namely, that the mean molecular weights of the amino-acid residues in the different proteins were not the same, is inadequate, since, for example, the mean residue weights of zein and ovalbumin are in a ratio not far different from 1:1.1, while the molecular weights of the proteins themselves are in a ratio of approximately 1:1.3. Moreover, the arbitrary omission of gliadin, hordein and two haemocyanins from Svedberg's classification4 needs a better reason than the maintenance of his hypothesis. Further, the existence of proteins of lower molecular weights than the basic unit of 17,600 (such as ribonuclease and cytochrome-c at 13,000) is a serious complication to the theory, and there are certain groups, such as Svedberg's 'x 96' group, where the actual mean of the molecular weights cited is not even approximately that of the characteristic 'group value'.