On models and mechanisms