The aim of this paper is to explore the influence of peer behaviour on student marijuana consumption. Our hypothesis is that, in contrast to the traditional measures of peer group effects carried out at class or school level, the use of a closer peer group, which we relate to the group of friends, is more relevant in the explanation of marijuana consumption. On the basis of the data provided by the 2004 Spanish Survey on Drug Use in the School Population, we estimate a probit model in which two alternative peer variables are introduced. The results show that, once the effect of the closer peer group is controlled for, the effect of classmates' behaviour on the student is insignificant. Moreover, the closer peer group effects are asymmetric in their magnitude. (Contains 5 notes and 3 tables.)