In this paper we propose and estimate a model of theater participation using the data contained in the 2002 Survey of Public Participation in the Arts from the USA, a dataset widely used to study the determinants of cultural participation. Our contribution relies on the use of an estimation technique that respects the count data nature of the attendance variable (number of theater performances that an individual attended) and allows for heterogeneous behavior. By using a Zero Inflated Negative Binomial Model, we can characterize two distinct behaviors for the observable attendance: a group of never-goers (who never participate) and a subpopulation that has a positive probability of attending. For this latter group, we can estimate the effect of certain personal variables on the probability of highest frequency. The results suggest that the proposed model is appropriate for estimating cultural participation. © 2008 Springer Science+Business Media, LLC.