A method to determine total phosphorus (TP) in the same digests prepared for total Kjeldahl nitrogen (TKN) determinations is desribed. The batch, high-temperature (block digester), HG(II)-catalyzed digestion step is similar to U.S. Geological Survey methods I-2552-85/I-4552-85 and U.S. Environmental Protection Agency method 365.4 except that sample and reagent volumes are halved. Prepared digests are desolvated at 220 degrees Celsius and digested at 370 degrees Celsius in separate block digesters set at these temperatures, rather than in a single, temperature-programmed block digester. This approach is used in the method escribed here, which permits 40 calibrants, reference waters, and smaples to be digested and resolvated in about an hour. Orthophosphate ions originally present in samples, along with those released during the digestion step, are determined colorimetrically at a rate of 90 tests per hour by an automated version of the phosphoantimonylmolybdenum blue procedure. About 100 microliters of digest are required per determination. The upper concentration limit is 2 milligrams per liter (mg/L) with a method detection limt of 0.01 mg/L. Repeatability for a sample containing approximately 1.6 mg/L of TP in a high suspended-solids matrix is 0.7 percent. Between-day precision for the same sample is 5.0 percent. A dialyzer in the air-segmented continuous flow analyzer provides on-line digest cleanup, eliminated particulates that otherwise would interfere in the colorimetric finish. An single-channel analyzer can process the resolvated digests from two pairs of block digesters each hour. Paired t-test analysis of TP concentrations for approximately 1,600 samples determined by the new method (U.S. Geologial Survey methods I-2610-91 and I-4610-91) and the old method (U.S. Geological Survey methods I-2600-85 and I-4600-85) revealed positive bias in the former of 0.02 to 0.04 mg/L for surface-water samples in agreement with previous studies. Concentrations of total phosphorus detemrined for ground-water samples by both methods were statistically equivalent.