We investigated multiple-rate diffusion as a possible explanation for observed behavior in a suite of single-well injection-withdrawal (SWIW) tests conducted in a fractured dolomite. We first investigated the ability of a conventional double-porosity model and a multirate diffusion model to explain the data. This revealed that the multirate diffusion hypothesis/model is consistent with available data and is capable of matching all of the recovery curves. Second, we studied the sensitivity of the SWIW recovery curves to the distribution of diffusion rate coefficients and other parameters. We concluded that the SWIW test is very sensitive to the distribution of rate coefficients but is relatively insensitive to other flow and transport parameters such as advective porosity and dispersivity. Third, we examined the significance of the constant double-log late time slopes (-2.1 to -2.8), which are present in several data sets. The observed late time slopes are significantly different than would be predicted by either conventional double-porosity or single-porosity models and are believed to be a distinctive feature of multirate diffusion. Fourth, we found that the estimated distributions of diffusion rate coefficients are very broad, with the distributions spanning a range of up to 3.6 orders of magnitude. Fifth, when both heterogeneity and solute drift are present, late time behavior similar to multirate mass transfer can occur. Although it is clear that multirate diffusion occurs in the Culebra, the number of orders of magnitude of variability may be overestimated because of the combined effects of drift and heterogeneity. ¿ 2001 American Geophysical Union