A large amount of modulation recognition algorithms has been reported in literature for Single-Input Single-Output (SISO) communications. But, to our knowledge, none of them have dealt with Multiple-Input Multiple-Output (MIMO) commu- nications. The issue addressed in this paper is the modulation recognition forMIMO communications under the assumption of a perfect symbol timing. In the first part of this paper, we develop the optimal solution based on Average Likelihood Ratio Tests. Although this solution is optimal, it requires the knowledge of the propagation channel which is usually un- known in a non cooperative environment. To overcome this problem, in the second part of this paper we propose an al- ternative solution based on Hybrid Likelihood Ratio Tests. Compared to the ALRT classifier, simulations show that the HLRT classifier presents good performances.