Among techniques for estimating enteric methane (CH4) emission by ruminants, open-circuit respiration chambers (OC), the use of a gas tracer (SF6), and the GreenFeed (GF) device are the most commonly used. In this study, we compared these techniques in 8 dry cows receiving a diet made of 70% hay and 30% concentrates given in limited and constant amounts, in a 15-wk experiment. Two periods in free stalls for SF6 and GF and in chambers for OC were used; in addition, SF6 was determined in chambers for 1 period. Methane emission (g/d) and CH4 yield (g/ kg DMI) were higher (P < 0.0001) for OC than for SF6 and GF (367, 310, and 319 g/d for OC, SF6, and GF, respectively). The difference between OC and GF was related to a difference in post-prandial rate of gas emission. The between-animal coefficient of variation of CH4 emission was higher for SF6 than for OC and GF (20.8, 13.5, and 12.0% on average, respectively). Correlation coefficients for CH4 emission and CH4 yield (r = 0.782 and r = 0.717, respectively; P < 0.05), but not significant between OC and GF, or between SF6 and GF. Correlation coefficients were highly significant for SF6 determined either in free stalls or in chambers (r = 0.908 and 0.903 for CH4 in g/d and g/kg DMI, respectively; P < 0.01). Carbon dioxide (CO2) emission and CO2 yield were similar for GF and OC (10,003 and 9,887 g/d, 752 and 746 g/kg DMI, respectively); CO2 data obtained with SF6 were lower (7,718 g/d and 606 g/kg DMI; P < 0.0001), but this technique is not relevant for CO2 emission determination. Correlation coefficients between OC and GF were not significant for CO2 emission and CO2 yield. This set of results shows that differences between methods are minor for average values, but that individual correlations may limit their interchangeability for determining gas emissions of individual animals. This study also shows the reliability of GF on-farm determination of CH4 and CO2 emissions for groups of animals.