Gravar-mail: Stimulus generalization as a mechanism for learning to trust