Richly structured reward predictions in dopaminergic learning circuits - World Wide