Excess fluoride (F-) in drinking water poses a health threat to millions of people around the world. In the present work, activated alumina (AA) has been used as an adsorbent. Data obtained from batch experiments were fitted to the (i) pseudo-first order, (ii) pseudo-second order, and (iii) Langmuir kinetic model. Model (ii) performed better than model (i), and fitted the data well. However, the rate constant for adsorption ka had to be varied as a function of the initial concentration of F- in the liquid phase c0. A more satisfactory approach is provided by Langmuir model, which fitted the data reasonably even though ka was independent of c0. Shreyas (2008) developed a model for the batch adsorption of F- onto porous pellets of AA. Some errors were detected in his computer program were corrected. The parameters of the model were estimated by fitting predictions to data. The parameter values suggest that the adsorption process is likely to be diffusion limited. Column experiments were conducted as follows. The pellets were soaked in deionized water for a time ts before they were loaded into columns. A feed solution having a fluoride concentration cf = 3 mg/L was fed to column and the concentration of F- in the exit stream ce was measured at regular intervals. Breakthrough was deemed to have occurred when ce exceeded the permissible limit (= 1 mg/L). Constant values of the bed height H, and the empty bed contact time tc were used in the experiments. The volume of treated water V, scaled by the volume of the bed Vb, varied strongly with the soaking time ts, with a maximum at ts = 24 h. To understand the possible reasons for this behaviour, XRD, FESEM, and FTIR were used to characterize the surface of AA. Though the concentrations of the surface hydroxyl groups may influence the adsorption of F-, FTIR studies show there is no direct correlation between V/Vb and the concentrations of these groups. The FESEM and XRD studies indicate that fresh AA consists mainly of boehmite, which gradually converted to gibbsite during soaking. For fixed values of H and tc, the dimensionless volume of treated water V/Vb was a maximum at D = 45 mm. This behaviour may be caused by wall effects for small values of D and by occurrence of quasi-static regions near the wall for large values of D. The cost of treated of water was Rs. 0.42/L. It decreased slightly to Rs. 0.37/L after one regeneration cycle, but increased to Rs. 0.41/L after two cycles. The volume of treated water after two regeneration cycles was 595 L/kg. The concentration of Al3+ ions ca in the treated water increased and exceeded the permissible limit of 0.2 mg/L as the number of regeneration cycles increased. The concentration of F- in regeneration effluent cre was in the range 32-70 mg/L. The effluent was subjected to solar distillation, leading to a distillate whose fluoride concentration F- was in the range 9-12 mg/L. The distillate can be discharged into the public sewers, as the permissible limit is 15 mg/L.