Richard Marshall interviews Scott Soames in 3:AM Magazine:
3:AM: You argue that a theory of meaning needs propositions but also needs to account for the cognitive stance of a person towards a proposition and also how propositions manage to represent the world and have truth conditions. Is that right? So firstly, can you explain why you find Russell’s attempts to account for them not right?
SS: Yes, a theory of meaning for a language L needs propositions that represent the world and so have truth conditions. Yes, it also needs an account of cognitive stances agents take to propositions – if L has sentences – e.g. attitude ascriptions – that predicate properties of propositions. My answer to your question about Russell is explained in chapter 9 of my newThe Analytic Tradition in Philosophy, Vol. 1. Here is the gist of it. Between 1900 and 1910 Russell believed in propositions constituted by objects and properties, but he couldn’t explain their “unity”. Just as sentences aren’t collections of unrelated expressions, but have a structural unity that distinguishes them from mere lists and allows us to use them to represent the world truly or falsely, so propositions aren’t collections of unrelated meanings of the words used to express them, but have a unity that endows them with truth conditions that mere aggregations of their parts don’t have. Russell struggled unsuccessfully to explain this unity until he rejected propositions in 1910 in favor of his multiple relation theory of judgment. Although that theory was disastrous, the insight behind it was brilliantly correct. The intentionality of agents can’t be derived from the supposed sui generis intentionality of propositions to which agents bear attitudes. Instead, the unity that brings together Desdemona and being unfaithful in Othello’s belief that Desdemona was unfaithful is provided by the sui generis fact that the agent predicates being unfaithful of Desdemona. What Russell failed to see was how this insight can be used to reconstruct genuinely unified (i.e. representational) propositions by deriving the intentionality of propositions from the intentionality of agents who entertain them.
More here.