The development and changes in Quine's ideas on universais are analysed, and especially the interplay of the notions of attribute, set and predicate is highlighted. In a first logico-mathematical part it is shown how Quine banned attributes as a result of extensionalism, and how set-theoretic solutions for Russell's paradox disturbed the easy view of each predicate determining a class. Quine even tried to formulate nominalistic theories without universais (sets). It is further shown how linguistic considerations played a role in Quine's (...) ideas on universais. The role of predicates is scrutinised, and it is shown how they are torn between logical and linguistic demands. It is suggested that the role of attributes has been taken over by predicates. This semantic role of predicates is quite unstable. I conclude with the suggestion that Quine should separate set theory and linguistics more radically. (shrink)