Daily methane production and feed intake were measured on 160 adult ewes, which were the progeny of 20 sires and 3 sire types (Merino, dual-purpose and terminal) from a genetically diverse flock. All animals were housed in individual pens and fed a 50/50 mix of chaffed lucerne and oaten hays at 20 g/kg liveweight (LW), with feed refusals measured for at least 10 days before the first of three 22-h measurements in respiration chambers (RC). Feed was withdrawn at 1600 h on the day before each RC test to encourage the ewes to eat the entire ration provided for them in the RC. After the first 1-day RC test, the sheep were returned to their pens for a day, then given a second 1-day RC test, followed by another day in their pens, then a third RC test. After all animals had been tested, they were ranked according to methane emissions adjusted for feed intake in the RC and on the previous day, enabling 10 low and 10 high methane animals to be chosen for repeat measurement. No variation between sires nor consistent effects of LW on feed eaten (%FE, expressed as per cent of feed offered) was evident in the 10 days before the first RC measurement. However, significant differences between sires (equivalent to an estimated heritability of 41%) were identified for %FE during the 2nd and 3rd days of RC testing (2 and 4 days after the initial RC test). The analysis of all data showed that methane emissions in the RC were related to feed intake on the day of testing and the two previous days (all P<0.0005). Before correcting for feed intake on previous days, there was some variation between sires in methane yield, equivalent to an estimated heritability of 9%. Correction for feed intake on the 2 previous days halved the residual variation, allowing other effects to be detected, including effects of LW, twins reared as singles, test batch, RC and test-day effects, but estimated sire variation fell to zero. In order to avoid potential biases, statistical models of methane emissions in the RC need to consider potential confounding factors, such as those identified as significant in this study.