Four low and four high methane (CH4) emitters were selected from a flock of 20 Romney sheep on the basis of CH4 production rates per unit of intake, measured at grazing using the sulphur hexafluoride (SF,) tracer technique. Methane emissions from these sheep were monitored at grazing for four periods (P): October, November, January and February 1999/2000. All measurements were carried out on perennial ryegrass/white clover pasture under generous herbage allowance, and the sheep were maintained on similar pastures during non-measurement periods. The tracer technique was used for all the CH4 measurements and feed DM intake was calculated from total faecal collection and estimated DM digestibility. Data for liveweight (LW), gross energy intake (GEI) and CH4 emission were analysed using split-plot analysis of variance. In addition, a between-period rank order correlation analysis was carried out for CH4 emission data. Low CH4 emitters were heavier (P < 0.05) than the high emitters in all the periods, but they did not differ (P < 0.05) in their gross energy intakes (GEL MJ/kg LW0.75). Low and high CH4 emitters consistently maintained their initial rankings in CH4 yield (% GEI) throughout the subsequent periods and the correlation analysis of rank order for CH4 yield showed strong between-period correlation coefficients, although this was weaker in the last period. It is suggested that feeding conditions that maximize feed intake (e.g. generous allowance of good quality pasture under grazing) favour the expression and persistence of between-sheep differences in CH4 yield.