Ultrafiltration methods have a twofold usefulness :—(1) As a general means of fractionating disperse systems, and (2) in providing data enabling the size of dispersed particles to be estimated. Their successful application requires an understanding of the physical processes involved. An ultrafilter membrane does not behave like an ordinary sieve in which coarseness of mesh alone determines whether or not a non-deformable particle shall pass. As the meshes become smaller the relative importance of the thickness of the sieve increases, until eventually, as in the case of a membrane, the length of a pore becomes very much greater that its diameter. The liquid traversing the pores of a membrane is in contact with a very large surface, and hence surface phenomena may be expected to play an important part in an ultrafiltration process. The ratio, area of pore surface/pore volume, varies inversely with the first power of the pore radius and hence becomes very large for ultrafine capillaries. Before proceeding to consider experimental evidence upon the course of filtration with typical disperse systems under varied conditions, it will be well to form a working conception of the structure in ultrafilter membranes. Collodion films which are the most commonly used ultrafilters have been studied from the point of view of their structure (Elford, 1930), and the evidence of optical examination indicated an arrangement of aggregated particles somewhat analogous to a pile of shot. This implies lateral as well as vertical permeability, while the degree of porosity will be determined by the size of the particulate units and their arrangement and closeness of packing in successive planes. The analogy cannot be regarded as in any sense a strict one, however, since the colloidal nitrocellulose particles are not spherical, appearing in the ultramicroscope to be slightly elongated, while their mutual orientation is determined by fields of molecular forces that are strongly polar in nature. The effective channels through such membranes approximate to long capillaries, and since the flow of liquid through the membrane will occur most readily by the shortest path from face to face, these capillaries are relatively straight. Evidence may be cited in support of this. Poiseuille’s law is found to govern the flow of water through the membranes over ranges of low pressure where no distension of the membrane occurs ; also when a dye suspension is filtered through a membrane supported upon a perforated plate, the circular areas, through the pores of which the dye passes, become deeply stained through the entire membrane thickness. The corresponding stained circular areas visible on the under face of the membrane remain sharply defined even after long periods of filtration. Lateral diffusion between the mainly effective capillaries therefore occurs very slowly. Thus the ultrafilter membrane may be regarded from the point of view of its performance as a porous structure, the effective pores being long and relatively straight channels, formed by the communicating interstices, between the elements in superimposed strata of the aggregated nitrocellulose particles.