[en] A method is described for determining ²²⁸Th (and hence ²²⁸Ra) in small samples of human bone when accompanied by many times as much ²²⁶Ra. Such determinations are needed for the radiation dosimetry of skeletally-deposited radium 30-50 years after intake of mixtures in unknown proportions of the two isotopes of radium. The only preparation of the bone sample is ashing and dissolution in nitric acid. The short-lived radioactive gas, ²²⁰Rn (54.5 seconds), is de-emanated from the solution with nitrogen, which carries it into a 3.8-litre collection chamber. Positively charged ions of ²¹⁶Po and ²¹²Pb produced by the decay of the ²²⁰Rn during its stay in the chamber are collected on a negatively charged electrode in the form of a copper disk, to which is cemented a zinc sulphide phosphor-coated Mylar disk. De-emanation and ion collection continue for up to 24 hours. The electrode is then removed and presented to a second phosphor disk on the window of a photo-multiplier tube. Alpha particles from the collected radioactivity are thus counted with almost 4π geometry, and an initial efficiency is obtained of about 0.3 count per disintegration of ²²⁸Th in the solution. Positively charged ions of ²¹⁸Po and ²¹⁴Pb from the decay of some ²²²Rn in the chamber are also collected. The overall efficiency for this is much lower than for the collection of the daughters of ²²⁰Rn but, because of the much higher activity of the ²²⁶Ra, the initial counting rate is primarily due to its daughters. To make maximum use of all the counting data, they are analysed by a computer method of least squares. In routine use, de-emanation for 1400 minutes, followed by counting for 1400 minutes, permits the measurement of a few tens of fCi of ²²⁸Th. The need for correct weighting of very low counting rates is demonstrated. (author)