An in vitro, pepsin-pancreatin digestibility method was used to estimate the bioavailabilities of amino acids in plant feedstuffs, including two cultivars of four cereal grains and three high-protein feedstuffs. As a technique for separating hydrolytic products from substrate, trichloroacetic acid (TCA) was compared with three ultrafiltration techniques. Solutions were filtered continuously by a membrane of 500-Dalton pore size (cUF) and terminally by membranes of 500 Dalton (tUF) and 10,000 Dalton (CX). Additionally, bioavailability of amino acids was determined in vivo by the true amino acid availability (TAAA) method and utilization of lysine was determined by chick growth assay (CGA).Correlations of in vitro and in vivo estimates of availabilities were ranked CX = cUF > tUF > TCA for total amino acids and were ranked similarly for most individual amino acids. When determined using the superior technique, correlation between in vitro digestibilities of amino acids and TAAA were moderately high for the combined cereal grains (r = .92), were lower when data included wheat middlings (r = .90), but were unreliable when data included soybean meal and corn gluten meal (r = .29). Prediction of CGA values from in vitro digestibility was less reliable (r = .42) than that using TAAA values. There was agreement between values for lysine availability when determined for six grain samples by CGA and TAAA (r = .79). Evaluation from the correlations of the in vitro with the in vivo estimates of availabilities indicated that terminal filtration by a membrane of medium pore size (10,000 Dalton) or continuous filtration by a membrane of very low pore size (500 Dalton) made separations superior to those of terminal filtration (500 Dalton) and TCA precipitation.