Point-of-use (POU) drinking water treatment technology enables those without access to safe water sources to improve the quality of their water by treating it in the home. One of the most promising emerging POU technologies is the biosand filter (BSF), a household-scale, intermittently operated slow sand filter. Over 500,000 people in developing countries currently use the filters to treat their drinking water. However, despite this successful implementation, there has been almost no systematic, process engineering research to substantiate the effectiveness of the BSF or to optimize its design and operation. The major objectives of this research were to: (1) gain an understanding of the hydraulic flow condition within the filter (2) characterize the ability of the BSF to reduce the concentration of enteric bacteria and viruses in water and (3) gain insight into the key parameters of filter operation and their effects on filter performance. Three 6-8 week microbial challenge experiments are reported herein in which local surface water was seeded with E. coli, echovirus type 12 and bacteriophages (MS2 and PRD-1) and charged to the filter daily. Tracer tests indicate that the BSF operated at hydraulic conditions closely resembling plug flow. The performance of the filter in reducing microbial concentrations was highly dependent upon (1) filter ripening over weeks of operation and (2) the daily volume charged to the filter. BSF performance was best when less than one pore volume (18.3-L in the filter design studied) was charged to the filter per day and this has important implications for filter design and operation. Enhanced filter performance due to ripening was generally observed after roughly 30 days. Reductions of E. coli B ranged from 0.3 log10 (50%) to 4 log10, with geometric mean reductions after at least 30 days of operation of 1.9 log10. Echovirus 12 reductions were comparable to those for E. coli B with a range of 1 log10 to >3 log10 and mean reductions after 30 days of 2.1 log10. Bacteriophage reductions were much lower, ranging from zero to 1.3 log10 (95%) with mean reductions of only 0.5 log10 (70%). These data indicate that virus reduction by BSF may differ substantially depending upon the specific viral agent. © 2008 Elsevier Ltd. All rights reserved.