Heart rate variability was investigated in order to determine whether it could, in turn, assess alterations of the autonomic nervous system at different levels of excitement. Warmblood horses were given a challenging ground exercise task during which their behavioural and physiological responses could be studied. Prior to commencement of the study, the horses were evaluated at rest and during forward walking, then while made to walk backwards continuously for three minutes, according to a standardised protocol. The horses were then exposed to two training sessions, followed by a re-evaluation of backward walking. Heart rate and heart rate variability parameters, such as the standard deviation of the beat-to-beat intervals, the low (sympathetic tone) and high (parasympathetic tone) frequency component of heart rate variability, and their ratio (index representing the sypatho-vagal balance), were sampled at rest, during forward walking, during the first backward walk, and during the second backward walk. Using video recordings taken during both backward walking sessions, stress-related behaviour was determined. The first backward walk experiment induced a rise in heart rate, low frequency heart rate variability, and the low frequency/high frequency heart rate variability ratio, compared to rest and forward walking. A decrease of high frequency heart rate variability in response to the first backward walking session was noted in comparison to rest and forward walking. The standard deviation of the beat-to-beat intervals decreased from both forward walking and rest, but only the standard deviation of the beat-to-beat intervals at rest was significant. In the second backward walking session, heart rate and the parameters of the sympathetic branch of the autonomic nervous system (low frequency component of heart rate variability, and the index representing the sypatho-vagal balance) were decreased and the vagal tone (high frequency component of heart rate variability) was increased compared to the first backward walking session. According to the duration of stress-indicating behavioural patterns, a decrease of excitement after the two training sessions occurred. At that time backward walking did not differ from forward walking for any parameter measured. It was found that there were correlations between heart rate variability parameters and stress-indicating behaviour, as well as heart rate. The results of this study indicated that the heart rate variability parameters (i.e. both the low frequency and high frequency components) were valuable measures for quantifying sympatho-vagal balance. These measures allow for a more precise assessment of the responses of heart rate and standard deviation of beat-to-beat intervals to mental stress during low intensity exercise.