It is shown that failure to take into account the nonuniformity of stress in particles near a surface may lead to gross errors when stress measurements are made by x‐ray diffraction on crystallites of one phase dispersed in a matrix of another phase. In particular, the measured strains may depend on the wavelength of the radiation used, and on the particle size of the dispersed phase even though the actual strains in particles not too close to the surface are not dependent on particle size. The measured strains may differ by a factor of two or three from the strains in particles below the x‐ray penetration depth. Experimental evidence on a vitreous china (quartz—glass) system is presented, showing that the observed variation of strain with wavelength and particle size is of about the magnitude predicted by calculation on a simple model system.