The compromise effect refers to individuals' tendency to choose intermediate options. Its existence has been demonstrated in a large number of hypothetical choice experiments. This paper uses field data from a specialties restaurant to investigate the existence and strength of the compromise effect in a natural environment. Despite the presence of many factors that potentially weaken the compromise effect (e.g., a very large choice set, the opportunity to choose familiar options), we find evidence for it both in descriptive statistics and regression analyses. Options which become a compromise after a change in the choice set gain on average five percent in market share. We also find that the compromise effect is especially pronounced in groups, while for single customers it is statistically insignificant.