Crystalline silica collected on membrane filters is commonly analyzed by X-ray diffraction (XRD). If the dust layer on the filter is very thin, the linearity of X-ray response is not affected. As loading increases, the X-ray beam is attenuated and the linear relationship no longer holds. The basic method for correction is to employ a silver filter to assess and compensate for the reduced X-ray response. No comprehensive study has yet been published that defines the dust loading beyond which the linear relationship no longer holds. As a consequence, a variety of dust loading limits are currently taken for linearity, ranging from 0.2 up to 2 mg. Members of the ISO working group for silica measurement collaborated in an experimental study on absorption for defining the critical sample mass, m_cr, above which the measured intensity of a reflection of the analyte deviates from the theoretically correct intensity by more than a set threshold. Quartz, calcite, hematite in the respirable size range, as pure minerals and mixtures, were used to test both the indirect and the direct-on-filter XRD methods. Analyses were carried out in three laboratories using CuKα radiation. The critical sample mass for a weight fraction of quartz, f_Qtz, depends on the area of deposition on the filter, A, and the mass absorption coefficient of the sample, μ_sample. For a typical f_Qtz = 5% value, if a 10 % deviation is accepted to represent the limit of linearity of XRD response, the critical sample mass derived from theoretical calculations is given by the equation: m_cr10 = 35.711 ⋅ A ⋅ μ_sample^−1.065, where m_cr10 is given in mg, A in cm², and μ_sample in cm²/g. Equations were also found for deviation values of 15 % and 20 %. A standard “average” sample of dust was defined, characterized by f_Qtz = 5% and μ_sample = 100 cm²/g, allowing the setting of a critical sample mass valid for the general case.