Emissions of 710 ewes at pasture were measured for 1 h (between 09:00-16:30 h) in batches of 15 sheep in portable accumulation chambers (PAC) after an overnight fast continuing until 2 h before measurement, when the sheep had access to baled hay for 1 h. The test was used to identify a group of 104 low emitters (I-Low) and a group of 103 high emitters (I-Hi) for methane emissions adjusted for liveweight (CH4awt). The 207 ewes selected at the initial study were remeasured in 5 repeat tests from 2009 through 2014 at another location. The first repeat used the original measurement protocol. Two modified protocols, each used in 2 yr, drafted unfasted sheep on the morning of the test into a yard or holding paddock until measurement. Emissions of the I-Hi sheep were higher (102-112%) than I-Low sheep in all subsequent PAC tests, with statistical significance (P < 0.05) in 3 tests. Tests without overnight fasting were simpler to conduct and had repeatabilities of 51 to 60% compared with 31 and 43% for the initial and first repeat tests, respectively. After habituation to a diet fed at 20 g/kg liveweight, 160 of the 207 sheep were measured in respiration chambers (RC); 10 high (Hi-10) and 10 low (Low-10) sheep were chosen, representing extremes (top and bottom 6.25%) for methane yield (MY; g CH4/kg DMI). The Hi-10 group emitted 14% more methane (adjusted for feed intake) in a follow-up RC test, but Low-10 and Hi-10 sheep differed in only 1 of the 5 PAC tests, when Hi-10 sheep emitted less CH4awt than Low-10 sheep (P = 0.002) and tended to eat less in the feeding opportunity (P = 0.085). Compared with their weight on good pasture, Low-10 sheep were proportionately lighter than Hi-10 sheep in the relatively poor pasture conditions of the initial test. Sheep identified as low emitters by PAC tests using the initial protocol did not produce less CH4 (mg/min) when fed a fixed level of intake in RC. Correlations between estimates of an animal's CH4awt measured in PAC and CH4 adjusted for feed intake in RC were quite low (r = 0-19%) and significant (P < 0.05) in only 1 test of unfasted sheep. With moderate repeatability over the 5 yr, PAC tests of CH4awt could be a viable way to select for reduced emissions of grazing sheep. As well as exploiting any variation in MY, selecting for reduced CH4awt in PAC could result in lower feed intake than expected for the animals' liveweight and might affect the diurnal feeding pattern. Further work is required on these issues.