I have devoted so many pages to accentuating the difference between traditional oral and more recent literate ways of communication, because we have to bear in mind the differences between them when we want to find out what the study of language can and can't do. In our society we find, side by side, primordial speech situations and metalinguistic reflections of arguably great depth. But while it is impossible to reduce a primordial speech situation to a written text, this can easily be done with oral discussions that do not involve extensions, such as a shared reality in visible reach of all participants. Hansard, for instance, is, in most cases, considered commensurate with, and as authentic as, the oral parliamentary proceedings of which it is a (sometimes quite heavily post-edited) transcript. Yet transcripts of pub chats normally leave a lot to be desired. They will almost always contain bits and pieces we cannot, with the best of intentions, make sense of.

The problem with the primordial speech situation is that it is impossible in principle to separate what is said from the setting in which it takes place, and one cannot distinguish the linguistic from the non-linguistic symbolic interaction that takes place in it. Not even the most detailed annotation of these non-linguistic factors can replace the immediacy of being part of the situation itself.