It is well known that dust conditions in a coal mine vary from place to place and from time to time. Watson (1949) pointed out that even at one position on a long-wall face the concentration fluctuates wildly from minute to minute and varies even more from day to day, while the two halves of a double-unit face might differ considerably. Bedford and Warner (1943), when considering the concentration of dust in different mining operations, also noted that there was great variation in the individual dust counts and in addition large varia tions from mine to mine, so much so that it was not practicable to estimate with any reasonable accuracy an average dust concentration for each mining operation. Other difficulties which this variation causes in dust sampling work have been emphasized by several writers, and Oldham and Roach (1952) have pointed out that it must be given due con sideration in designing a sampling scheme which is efficient in terms of the effort needed to obtain a specified amount of information. Fortunately it appears that momentary variation is of little importance in the aetiology of pneumoconiosis. The arguments of Wright (1953) and Long (1953) show that an estimate of average dustiness and the period during which it operates is all that is required, an estimate which could most economically be obtained by means of automatic, long-term sampling instruments. However, so long as short-term, inter mittent sampling instruments continue to be used, the effect of this variation must be considered before deductions about average dust conditions may safely be drawn. Equally, it must be allowed for in interpreting the records of past dust conditions. Accordingly, this paper will present some of the evidence that we possess at this Unit of its nature and magnitude, and will draw some practical conclusions. In the past six years extensive sampling with the thermal precipitator has been carried out by teams from this Unit. The number of samples obtained (some 8,000) is large, although already eclipsed by the number taken by the National Coal Board's sampling teams, which are making ever-increasing use of the thermal precipitator, an accurate though inconvenient and short-term dust-sampling instru ment. However, the samples obtained by this Unit have been concentrated at a relatively small number of mines, so that the number of samples representing each different environment is large. Mines with a big range of average concentrations were well represented. In consequence our data are particularly suited to the detection of any consistent patterns in the variation of concentration, such as might be obscured in data covering a less extensive range in less detail. Few of the surveys we have carried out at this Unit have continued for more than two or three weeks, and it is clear that a statistical treatment of variation of concentration from day to day could not be satisfactory without a long series of daily samples. There are many factors determining the average level of dust conditions during a shift, some of them (such as the width of the workings from face to goaf) clearly likely to cause changes of a cyclic nature, others (the progress of a dust suppression programme, a change in the nature of the coal seam or surrounding strata) inducing more or less steady trends, while a large number will operate in an apparently random manner. Under standing of the effects of such factors will most readily be achieved by analysis of the results of routine dust sampling over long periods of time, provided that such sampling is carried out in accordance with sound statistical principles. This paper will be concerned only with the variation of concentrations about the average level during particular shifts at particular coal faces. 227