Displaying images of eyes causes people to behave more pro-socially in a variety of contexts. However, it is unclear whether eyes work by making people universally more pro-social, or by making them more likely to conform to local norms. If the latter, images of eyes could sometimes make people less pro-social if pro-social behaviour is not the local norm. To separate these hypotheses we conducted a field experiment in which we explored whether manipulating a local descriptive norm altered the eyes effect. We recorded litter dropping decisions on a university campus in a 2 x 2 design, comparing situations with and without litter already on the ground (a manipulation of the local descriptive norm) and with and without large signs displaying images of watching eyes. We additionally recorded the number of potential human observers in the vicinity at the time of each littering decision. We observed a norm effect: the presence of litter on the ground increased littering, replicating previous findings. We also found that images of watching eyes reduced littering, although contrary to previous findings this was only when there were larger numbers of people around. With regard to our central aim, we found no evidence that litter on the ground interacted non-additively with images of eyes to induce increased littering behaviour. Our data therefore support the hypothesis that images of eyes induce more pro-social behaviour, independent of local norms. This finding has positive implications for the application of eye images in combating anti-social behaviour.