Two Latin square design experiments investigated the relationship between hydrogen sulphide concentration in the rumen headspace gas of dairy cows and the early stages of protein degradation in the rumen. In Expt 1, three protein sources differing in rumen N (nitrogen) degradability (maize gluten feed (MGF); sunflower meal (SFM); and soyabean meal (SBM)) were used, whereas in Expt 2 four different batches of the same feed (MGF) differing in colour (CIE L*, a*, b* (CIELAB) scale) were used. After allowing the concentration of hydrogen sulphide in rumen gas to decline close to zero, a fixed amount of protein sources was offered to cows and the concentrations of hydrogen sulphide were recorded in rumen headspace gas at 30-min intervals. In Expt 1, the concentration of hydrogen sulphide showed considerable variation between protein sources, with MGF having the highest concentration followed by SFM and SBM resulting in very low concentrations. The N wash losses (zero time measurements with nylon bags) ranked the feeds in the same way, from MGF (highest; 61%) to SBM (lowest; 26%). There were marked differences in the degradation of cystine and methionine between protein sources, although the degradation of cystine was always higher than for methionine. MGF (Expt 2) led to increased concentrations of hydrogen sulphide, with peak concentrations achieved between 1 and 2 h after feeding. The concentrations of hydrogen sulphide were higher for MGF1, intermediate for MGF2 and lower for MGF3 and MGF4, agreeing with colour scale. Differences in the early stages of dietary sulphur degradation corresponded with differences in hydrogen sulphide concentrations in rumen gas. The results suggest that hydrogen sulphide concentrations in the rumen headspace gas could be useful to evaluate nutritional parameters not measured by the in sacco technique, contributing to a better understanding of the response of dairy cows to different protein supplements.