An analytical approach is proposed to explain the appearance of unwanted low-frequency artifacts in halftoned images when using random dithering. The research is based on a theorem that relates the correlation of the input (continuous) gray-level signal to the correlation of the (halftone) binary output signal. This second-order statistical analysis alludes to the claim that the introduction of low-frequency artifacts is inevitable, being an intrinsic property of the dithering process rather than of individual images or masks. In addition, high-frequency information in the continuous image is attenuated more than low-frequency information. This effect is enhanced for mean gray levels farther from mid-gray.