The release rate (RR) of sulphur hexafluoride (SF 6) gas from permeation tube in the rumen appears to be positively related with methane (CH 4) emissions calculated using the SF 6 tracer technique. Gas samples of breath and ruminal headspace were collected simultaneously in order to evaluate the hypothesis that transactions of SF 6 in the rumen are the source for this relationship. Six non-lactating dairy cows fitted with rumen cannulae were subdivided into two groups and randomly assigned to a two-period crossover design to permeation tubes with low RR (LRR = 1.577 mg/day) or two-times higher RR (HRR = 3.147 mg/day) RR. The cows were fed limited amounts of maize silage (80% ad libitum) split into two meals (40% at 0800 h and 60% at 1600 h). Each period consisted of 3-day gas sampling. Immediately before the morning feed and then each hour over 8 h, ruminal gas samples (50 ml) were withdrawn through the cannula fitted with stoppers to prevent opening. Simultaneously, 8-h integrated breath gas samples were collected over the same period. Ratios of concentration of CH 4/SF 6, CO 2/SF 6 and CO 2/CH 4 and emission estimates of CH 4 and CO 2 were calculated for each sample source using the SF 6 tracer technique principles. The LRR treatment yielded higher (P < 0.001) ruminal CH 4/SF 6 (by 1.79 times) and CO 2/SF 6 (by 1.90 times) ratios than the HRR treatment; however, these differences were lower than the 2.0 times difference expected from the RR between the LRR and HRR. Consequently, the LRR treatment was associated with lower (P < 0.01) ruminal emissions of CH 4 over the 8-h collection period than with the HRR treatment (+11%), a difference also confirmed by the breath samples (+11%). RR treatments did not differ (P = 0.53) in ruminal or breath CO 2 emissions; however, our results confirm that the SF 6 tracer seems inappropriate for CO 2 emissions estimation in ruminants. Irrespective of the RR treatment, breath samples yielded 8% to 9% higher CH 4 emission estimates than the ruminal samples (P = 0.01). The relationship between rumen and breath sources for CH 4 emissions was better for LRR than for HRR treatment, suggesting that tracer performance decreases with the highest RR of SF 6 tested in our study (3.1 mg/day). A hypothesis is discussed with regard to the mechanism responsible for the relationship between RR and CH 4 emission estimates. The use of permeation tubes with small range in RR is recommended in animal experiments to decrease variability in CH 4 emission estimates using the SF 6 tracer technique. © Copyright The Animal Consortium 2011.