Summary: The diffusion constant of C-reactive protein determined by the method of Allison and Humphrey is not the same for all samples of pathologic serum, indicating that C-reactive protein may exist in different sera in different states of aggregation. It is unlikely that it is strongly associated in native serum with β-lipoprotein. Results of manipulating the ionic milieu and the nature of the supporting medium in which diffusion was carried out, indicated that C-reactive protein (CRP) shows a strong interaction with the charged groups of agar. The anomalous behavior of CRP on immunoelectrophoresis may result in part from this interaction. The relevance of these observations to the quantitative determination of C-reactive protein (and probably some other antigens) by immunodiffusion techniques in agar were pointed out.