The release rate (RR) of sulphur hexafluoride (SF6) gas from permeation tube in the rumen appears to be positively related with methane (CH4) emissions calculated using the SF6 tracer technique. Gas samples of breath and ruminal headspace were collected simultaneously in order to evaluate the hypothesis that transactions of SF6 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 Immediately before the morning feed and then each hour over 8 h, ruminal gas samples (50 ml) were withdrawn through the cannula fitted with stoppers to prevent opening.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 CH4/SF6, CO2/SF6 and CO2/CH4 and emission estimates of CH4 and CO2 were calculated for each sample source using the SF6 tracer technique principles. The LRR treatment yielded higher (P < 0.001) ruminal CH4/SF6 (by 1.79 times) and CO2/SF6 (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 CH4 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 CO2 emissions; however, our results confirm that the SF6 tracer seems inappropriate for CO2 emissions estimation in ruminants. Irrespective of the RR treatment, breath samples yielded 8% to 9% higher CH4 emission estimates than the ruminal samples (P = 0.01). The relationship between rumen and breath sources for CH4 emissions was better for LRR than for HRR treatment, suggesting that tracer performance decreases with the highest RR of SF6 tested in our study (3.1 mg/day). A hypothesis is discussed with regard to the mechanism responsible for the relationship between RR and CH4 emission estimates. The use of permeation tubes with small range in RR is recommended in animal experiments to decrease variability in CH4 emission estimates using the SF6 tracer technique.