When deciding whether to invest in an improvement to animal welfare, farmers must trade-off the relative costs and benefits. Despite the existence of effective solutions to many animal welfare issues, farmers' willingness to pay for them is largely unknown. This study modelled pig farmers' decisions to improve animal welfare using a discrete choice experiment focused on alleviating aggression between growing/finishing pigs at regrouping. Eighty-two UK and Irish pig farm owners and managers were asked to choose between hypothetical aggression control strategies described in terms of four attributes; installation cost, on-going cost, impact on skin lesions from aggression and impact on growth rate. If they did not like any of the strategies they could opt to keep their current farm practice. Systematic variations in product attributes allowed farmers' preferences and willingness to pay to be estimated and latent class modelling accounted for heterogeneity in responses. The overall willingness to pay to reduce lesions was low at £0.06 per pig place (installation cost) and £0.01 per pig produced (running cost) for each 1% reduction in lesions. Results revealed three independent classes of farmers. Farmers in Class 1 were unlikely to regroup unfamiliar growing/finishing pigs, and thus were unwilling to adopt measures to reduce aggression at regrouping. Farmers in Classes 2 and 3 were willing to adopt measures providing certain pre-conditions were met. Farmers in Class 2 were motivated mainly by business goals, whilst farmers in Class 3 were motivated by both business and animal welfare goals, and were willing to pay the most to reduce aggression; £0.11 per pig place and £0.03 per pig produced for each 1% reduction in lesions. Farmers should not be considered a homogeneous group regarding the adoption of animal welfare innovations. Instead, campaigns should be targeted at subgroups according to their independent preferences and willingness to pay.