This paper describes CoMFA and CoMSIA studies for affinity and selectivity of a series of indole ligands to cannabinoid CB1 and CB2 receptors. The developed models have proven to be predictive, with average q(2) of 0.675 and average r(2) of 0.855, demonstrating a good statistical validation. The obtained results have helped us to understand the structural motifs that are responsible for the affinity and selectivity of some of these derivatives towards each subtype of cannabinoid receptor and have demonstrated that the exploited 3D-QSAR methods could be useful tools for the design of new safer analogues presenting better affinity and selectivity profiles.