Two experiments were undertaken using the in vitro gas production technique of Theodorou et al. [Anim. Feed Sci. Technol. 48 (1994) 185] to compare rumen liquor (RL) and faeces (FA) as inocula for fermenting gramminaceous forages over 96 h periods. Experiment 1 used 12 forages of differing in vivo digestibility (ammonia treated wheat straw, field-cured hay (Lolium perenne) and 10 artificially dried grasses (L. perenne) harvested at different maturities). Experiment 2 used seven maize-silage based forages (whole plant, stover, leaf, lower stem, middle stem, upper stem and husk). In both experiments, rumen liquor and faeces were obtained from two cows in early lactation, each fed daily with 9.4 kg DM of grass silage and 9.0 kg DM of concentrate. Rumen contents were sampled through the fistula, before morning feeding; faeces were sampled from the rectum, immediately afterwards. Rumen liquor (250 ml) was prepared by straining contents through two layers of muslin, adding the solids after blending with 250 ml of buffer and re-straining. Faeces were prepared by mixing (300 ml) with 150 ml of buffer and straining through two layers of muslin and adding a homogenate of the solids and 150 ml of buffer after straining. Data were fitted to the model of France et al. [J. Theor. Biol. 163 (1993) 99]. All model parameters showed FA to have a poorer fermentation capacity than RL. In both experiments, potential gas production volumes (A) were lower (on average 52.9 ml (18.5%)) and lag times longer (on average 2.9 h) for FA compared to RL. Fractional rate of fermentation at half asymptote (T/2) was generally greater for RL than FA (overall means, 0.042, 0.028) and the time required to T/2 being less (overall means, 21.9, 35.4 h). However, potential gas production (A) was highly correlated between RL and FA: Experiment 1 (r2=0.94, 11 forages, excluding ammonia treated straw) and Experiment 2 (r2=0.83, six forages, excluding middle stem). In Experiment 1, organic matter digestibility in vivo (OMDIV) was also highly correlated with both OMDFA (r2=0.77, 11 forages) and OMDRL (r2=0.89, 11 forages); OMDRL and OMDFA were also highly correlated (r2=0.81). Similar correlations occurred in Experiment 2. It is concluded that faeces have potential as an alternative inoculum to rumen liquor for in vitro gas production techniques, but methods of overcoming the longer lag phase with faeces require further research.