Transposition with minimal errors during training trials