<p>I develop an inferential account of collective self-knowledge. Starting with the assumption that groups have (at least propositional) attitudes I look at desiderata for any account of collective self-knowledge of such attitudes. Any such account has to explain the features that group avowals have in our ordinary linguistic practice. Moreover, any account ought to be compatible with as many views of group attitudes as possible. I propose a new account that looks at attitude-forming processes of groups as evidence for collective self-knowledge. It is based on forward-looking inferences in contrast to the backwards-looking conception commonly found in inferential accounts of self-knowledge in individuals. I then show that groups have a minimal form of privileged and peculiar access to their own attitudes because they have easy access to their own production of attitudes as evidence.</p>