Two experiments with the gas pressure transducer technique were conducted to investigate the effects of different types and levels of tannins. Forage samples were incubated in batch cultures together with rumen fluid and digestion medium in order to model rumen fermentation dynamics in vitro. In Experiment 1, graded levels (0, 25, 50, 75 and 100 mg/g of forage dry matter (DM)) of purified condensed tannins (CT) obtained from four tanniniferous tropical shrub legumes, Flemingia macrophylla, Leucaena leucocephala and two provenances of Calliandra calothyrsus (San Ramón and Patulul), were added to a CT-free mixture of the grass Brachiaria humidicola and the herbaceous legume Vigna unguiculata. The statistical models used for data evaluation included the two main effects (origin and level of CT) and their interaction. Apart from dose-dependent effects (P<0.001) there were clear plant-specific effects (mostly P<0.001) in fermentation dynamics, in vitro DM degradation and apparent degradation of N-containing compounds. Tannins from C. calothyrsus had the most pronounced effects on fermentation dynamics and nutrient degradation, followed by F. macrophylla and L. leucocephala. Within C. calothyrsus, CT from the provenance San Ramón suppressed asymptotic gas production and DM degradation to a larger extent (P<0.05) than CT from the provenance Patulul. Low addition (25 mg/g DM) of CT from C. calothyrsus was only slightly detrimental while high CT levels (75 and 100 mg/g DM) had clearly negative effects on asymptotic gas production and DM degradation. In a first incubation in Experiment 2, the grass in mixtures with foliage of the different tanniniferous shrub legumes was incubated with graded levels of polyethylene glycol (PEG) corresponding to 0/3, 1/3, 2/3 and 3/3 of the contents of CT in these mixtures. In a second incubation, all mixtures based on tanniniferous shrub legumes and mixtures containing non-tanniniferous shrub legumes ( Cratylia argentea, Desmodium velutinum) or a herbaceous legume ( V. unguiculata) were tested without PEG and with 80 g/kg supplementary PEG. In general, addition of PEG had no effect (P>0.05) on fermentation parameters in non-tanniniferous legumes. With tanniniferous species, response curves in fermentation parameters were mostly non-linear reaching a plateau at a PEG:CT-ratio between 1:3 and 2:3. One exception was that undegraded dietary N decreased (P<0.05) with increasing PEG addition until the highest level. In conclusion, the study demonstrated that to a large extent the effects of CT-rich plants on in vitro ruminal fermentation can be explained by the specific CT properties. Furthermore, apart from binding CT, PEG seems to have no major side-effects on ruminal fermentation, and for most fermentation variables and species, PEG supplementation equivalent to the CT content in the forage mixtures seems sufficient to prevent negative effects of CT on fermentation parameters.