Using the monumental work of figures like Bloomfield (1933), Harris (1951) and Chomsky (1965), scientific linguistics has often been centrally concerned with verbal patterns. Yet such views have even older roots. Ever since writing systems first arose in Sumeria, what people do as they talk has gradually been standardised in ways that eventually gave rise to the electronic use of artificial codes. First, iconic signs became ideographic or alphabetic. Then, in Europe, mediaeval scribes introduced spaces between units (and written words) and, in the aftermath of new technologies, there came dictionaries, grammars, printing and, yesterday, computers. As a result, written language bias (Linell, 2005) has dominated philosophy, linguistics and classic cognitive science. Languages are seen as verbal systems whose words and rules are, in some sense, separate from people. Even talk is often modelled around transcriptions that invite comparison with ways of construing verbal sequences. This can be highly misleading. In fact, while ideographic and alphabetic symbols are unsponsored, speech and hearing are coordinated human activity. Language is intrinsic to action and thus partly constitutive of experience: for Sartre (1945), self must be mirrored by others. Although this is a commonplace, many explanatory models reduce doing things with language to how linguistic forms are "used' and/or "represented'. In what follows, I adopt the distributed perspective (see, Cowley, 2011b), to offer an alternative.