This paper introduces some new features in the standard experimental design of the beauty contest in order to allow communication among participants. With that aim, we use the mode instead of the mean and non-rival payoffs. This design encourages students to communicate their guessed number, with a higher probability if subjects know the 0 Nash equilibrium. The lack of communication can only be explained by subjects endowed with competitive other-regarding preferences. Experiments are run in 11 classrooms ranging from 11 to 60 students in size. Participants are given at least one week to submit their guesses and a questionnaire explaining their choice. Results indicate that: i) communication induces coordination in the responses, ii) communication does not guarantee any improvement in the average reasoning level, iii) there exist significative differences according to classroom size and duration of degree.