Metal adsorption by Raisin River sediments in vitro depended linearly on soluble metal concentration to adsorption densities of 6,000-9,000 microg/g with 48 hr partition coefficients of approximately 50, 30, and 25 L/g for Cu, Cr, and Zn, respectively. Partition coefficients computed from field data spanned a comparatively wider range of values in a manner consistent with the often reported adsorbent concentration effect, but other factors likely contributed, too. Desorption of Zn was complete and rapid (24-48 hr) in contrast to Cr, which was incomplete and much slower; Cu desorption was intermediate to Zn and Cr. A reversible-resistant equilibrium model (DiToro et al. 1986) could not describe the observations as Cu and Cr had not reached metastable desorption equilibria after 24 days. Metal desorption, however, could be described kinetically by distributing sorbed cations between either of two classes; rapidly desorbing and slowly desorbing cations. Sequential and simultaneous desorption models gave similar predictions. Aqueous chemical considerations suggested precipitated as well as adsorbed species could give rise to the observations, but available data did not permit adequate tests of this hypothesis. The extent to which kinetic constraints rather than irreversible reactions account for the desorption-resistant binding signifies a potentially greater metal mobility or bioavailability than would otherwise be assumed.