Various magnetic parameters are in common use for estimating the grain size of magnetic particles. Among these, the ratio of the intensity of anhysteretic remanent magnetization (ARM) to that of isothermal remanent magnetization, as well as their alternating field (AF) demagnetization curves are used as an indicator of the domain state of the particles. Several models have been proposed to describe physically the acquisition of ARM in a biased AF field. Jaep <1969> first developed a semiquantitative theory based entirely on the thermal fluctuation analysis developed by N¿el <1949, 1954, 1955>. Significant discrepancies were found between his model and experimental results on magnetite. A new, general theory of ARM based on the work of Jaep is presented here, with particular regard to the influence of various parameters like grain size, coercivity, and mineralogy on ARM intensity. An analytical expression for ARM intensity in the special case of very fine particles was derived from this theory, and a good agreement with experimental results and data from the literature was found. A new estimation of the atomic reorganization time was obtained from ARM measurements on a sample of the Yucca Mountain Tuff, which has well-known mineralogy and grain-size distribution. The results are in agreement with the value proposed by McNab et al. <1968> for magnetite. Some authors considered magnetic interactions as the key to understand the ARM in fine particles, and this is certainly true for strongly interacting samples. In this case, ARM would be useless for the characterization of magnetic grains. However, many sediments have a very low concentration of well-distributed magnetic grains. For these samples, the explanation of an ARM in terms of intrinsic properties of the grains, as qualitatively proposed by other authors, is more suitable.