The judgment bias approach is a promising method for measuring optimism and pessimism in horses. The behaviour of horses during training using NR suggests more negative emotions than when using PR. Yet, horses trained using NR develop more optimistic bias than horses trained using PR, after the training stops. This optimistic bias could be due to higher feeding motivation in NR compared to PR mares, or to the cessation of a negative event. We therefore recommend training method using only PR, or combining NR and PR (Warren-Smith and McGreevy, 2007). Training methods that trigger good mental health should be as important for animal welfare as appropriate housing systems.