Three tropical legumes, namely Leucaena leucocephala, Sesbania sesban and Cajanus cajan, were subjected to chemical analysis plus in vitro, in situ and in vivo evaluations. Three different assays were used to determine total tannins: adsorption to polyvinyl pyrrolidine (PVPP-tannins), radial diffusion (RD-tannins) and protein precipitation capacity (BSA-tannins). Total phenols, total tannins and condensed tannins were highest for Sesbania. RD-tannins were correlated with total phenols ( r 2 = 0.93), PVPP-tannins ( r 2 = 0.92) and condensed tannins ( r 2 = 0.99). The protein precipitation capacity of Sesbania, Leucaena and Cajanus were 25.9, 6.13 and 4.05 μg BSA/g DM, respectively. Gas production at 24 h was negatively correlated with total phenols ( r 2 = 0.99), PVPP-tannins ( r 2 = 0.99) and condensed tannins ( r 2 = 0.91). The RD-, PVPP-tannins and the response to polyethylene glycol (PEG) in the gas production assay appeared to be useful as a first screen for tannins. In situ degradability did not reflect any adverse effects of tannins. However, in vivo experiments showed that the apparent DM digestibility of Sesbania and Leucaena was lower than the basal diet. The apparent protein digestibility was lower for all legumes compared to the basal diet. Most treatments caused a negative nitrogen balance. The problems associated with browse feeding were not only related to tannin contents, other factors such as inherently poor digestibility and low energy intake may also have lead to the poor animal performance on these diets. We propose, given the limitations of current tannin assays, that it is not possible to predict beneficial or harmful nutritional effects from total tannin concentrations per se.