Abstract Dairy cattle often make poor use of protein when offered diets comprising high proportions of alfalfa (Medicago sativa L.) hay or silage because nonprotein N formed during forage conservation and ruminal fermentation exceeds requirements for rumen microbial protein synthesis; however, condensed tannins (CT) may reduce proteolysis in the silo and in the rumen, thereby potentially improving the efficiency of crude protein (CP) use in ruminant diets. Two harvests, yielding 12 hays and 12 silages made from alfalfa and birdsfoot trefoil (Lotus corniculatus L.) that varied in concentrations of CT, were evaluated for in situ disappearance kinetics of CP in 6 ruminally cannulated lactating Holstein dairy cows (627±56.3kg). Prior to conservation, alfalfa contained no detectable CT, whereas CT in fresh lyophilized birdsfoot trefoil ranged from 1.16 to 2.77% of dry matter, as determined by a modified acetone-butanol-HCl assay. Percentages of CP remaining at each incubation time were fitted to nonlinear regression models with or without a discrete lag time. Effective ruminal disappearance of CP (rumen-degradable protein, RDP) was calculated by 3 procedures that included (1) no discrete lag (RDPNL), (2) discrete lag (RDPL), and (3) discrete lag with a lag adjustment (RDPLADJ). Regardless of the calculation method, RDP declined linearly with increasing CT concentrations (R2=0.62 to 0.97). Generally, tests of homogeneity showed that conservation type (hay or silage) or harvest (silage only) affected intercepts, but not slopes in regressions of RDP on CT. A positive relationship between lag time and CT suggests that the RDPLADJ approach may be most appropriate for calculating RDP for legumes containing tannins. With this approach, regression intercepts were mainly affected by conservation method, and RDPLADJ averaged 77.5 and 88.7% of CP for hay and silage, respectively, when no CT was present. Greater estimates of RDP for silages were related to extensive proteolysis in laboratory silos resulting in conversions of protein into nonprotein N forms, which readily washed out of Dacron bags. When RDPLADJ and CT were expressed on a CP basis, regression slopes indicated that each unit of CT protected 0.61 units of CP from ruminal degradation in hays and silages. Applying this relationship to a typical mid-maturity forage legume containing 21% CP suggests that a CT concentration of 3.8% of DM would be required to reduce RDP from 81% to a 70% target considered optimal for improving protein utilization and milk yields by dairy cattle.