Thirty-six owners of seasonally headshaking horses took part in a trial to compare the effectiveness of three types of nose net, a traditional cylindrical net (full net) and two forms of larger mesh nets which cover only the nostrils and dorsorostral muzzle (half nets). Baseline data relating to the overall severity of the problem and 18 specific behaviours describing the nature of the problem were recorded on a check sheet by the owners. A within-subjects repeated measures design experiment, with each net used for a week before reassessment, was then used to assess the effect of the nets on the headshaking problem. Approximately 75 per cent of owners reported some overall improvement with each net; around 60 per cent recorded a 50 per cent or greater improvement and 30 per cent a 70 per cent or greater improvement. The nets significantly reduced the overall headshaking score and the following specific behaviours: up-and-down headshaking, nose flipping, acting as if a bee had flown up the nose, shaking at exercise, shaking when excited, shaking in bright sunlight or in windy conditions (P < 0.0001), striking at the face, shaking at night, rubbing the nose when moving, rubbing the nose on objects, sneezing, shaking in the rain and shaking indoors (P < 0.05). There was no evidence of a significant effect on side-to-side headshaking, shaking at rest or rubbing the nose when stationary, but the effect on snorting was uncertain. There were few significant differences between the nets, but the half nets were reported to be significantly better at controlling 'bee up the nose' behaviour. Horses more than 10 years old were reportedly less likely to show a 50 per cent or greater improvement in 'nose flipping' and 'headshaking at exercise.