The avoidance of ammonia by sheep was tested using operant conditioning techniques, training 12 sheep to operate a lever to open a door that would allow them to traverse a chamber with or without ammonia and obtain a food reward. One half of the sheep had been previously exposed to ammonia at 45 ppm in a simulated ship voyage, and the other half were kept in identical conditions but without ammonia. The maximum number of lever presses, or breakpoint, achieved when the chamber contained 45-ppm ammonia was compared with the number when the chamber contained no ammonia. There was a significantly lower mean breakpoint when the sheep traversed the chamber filled with ammonia (mean: 4.2 presses), compared with the chamber filled with fresh air (mean: 6.6 presses), demonstrating that sheep exhibited a moderate aversion to 45-ppm ammonia. Although the scale of the reduction in mean breakpoint was considerable, the aversion was only clearly demonstrated in 75% of the sheep, and there was no evidence that prior exposure influenced this aversion. It is therefore concluded that sheep show a moderate aversion to 45-ppm ammonia, with no evidence of sensitivity being affected by previous exposures.