One of the many things we can do with language is exchange information. It is tempting to think this aspect of language is the key to understanding how it works. One way of developing this is to suppose that sentences in a language, such as English, each encode a piece of information, what we can call a proposition. When we say a sentence assertively we are putting forward the proposition it encodes. Call this picture the static view of conversation, a view consonant with the conception of meaning we find in Frege, Russell and the early Wittgenstein. As banal and truistic as the static picture seems, it is not forced on us. We can instead view individual sentences not as encoding information, but rather acting as instructions. This is the dynamic view of conversation, which has its origins in Wittgenstein's later philosophy. The dynamic and static models are different, idealized pictures of conversation: on the static view, conversation proceeds as an incremental accrual of information, on the dynamic view, conversation consists of back-and-forth instructions. The Dynamics of Conversation project is aimed at developing and comparing these two rival views.