Rumen cannulated sheep (n = 12) were tested at each of three feeding levels: 0.8, 1.24 and 1.6 times maintenance. The ration (lucerne chaff) was provided in eight equal portions to emulate morning and afternoon grazing. After at least 10 days adaptation to each feeding level, methane production was measured in an open circuit calorimeter for 22 h using the same feeding regime. During measurement, 10 rumen samples were taken for volatile fatty acid (VFA) analysis by an indwelling rumen probe with a sampling tube that passed outside the calorimeter. Feed intake was strongly correlated with daily methane production (DMP, r = 0.87). Both methane production and VFA concentrations showed bimodal patterns related to the feeding cycle, but feed intake had a much smaller effect on VFA concentrations than on methane production rate. Average VFA concentration was a poor predictor of DMP. The best predictor, propionate concentration, explained 26% of the variance in DMP. The weakness of the association between VFA concentrations and methane production could be a consequence of differences in rumen volume and differences in VFA absorption associated with feeding level, although the possibility of accumulation of alternate fermentation end products or re-fermentation of VFA cannot be excluded. It is concluded that none of the suite of VFA parameters assessed offers a useful tool to predict daily methane production in grazing sheep.