In a laboratory one-shot public good game, Fischbacher, Gachter and Fehr (2001) classify 50 percent of the subjects as conditional cooperators. Outside the lab, using a student sample, Frey and Meier (2005) find that people behave pro-socially, conditional on others' behavior. This paper tests for conditional cooperation and social comparisons in a natural field experiment, using decisions from a sample of cross-country skiers in Sweden on the issue of voluntary cash contributions to the preparation of ski tracks. Two test procedures are used. First, testing for correlation between beliefs about the contribution of others and own behavior and second, experimentally varying the beliefs about others' behavior. Using the latter approach, I find the share of subjects giving a contribution to be significantly greater in the group receiving information about others' behavior than in the group that does not. Regression analysis cannot reject that subjects are affected by social comparisons and express a behavior classified as conditional cooperation.