The present article deals with what we feel to be an unexplored path in present day sociolinguistics and a viable way toward making sociolinguistics more cognitive and cognitive linguistics more social. Naturally, sociolinguistics has a strong interest in issues of variation and change, variation being typically conditioned by social factors, such as age, gender, ethnicity as well as the discursive regimes of language use present in a community of practice. However, just like much of linguistics can be accused of not paying enough attention to the social (or communicative) side of language (cf. Croft 2009), sociolinguistics may be criticized for its relative lack of attention to cognition. We advocate that a true sociolinguistics should bridge this gap, and construct a theory about how social life influences cognitive knowledge. Against this background, we propose to combine a narrow linguistic notion of norm (Section 2) with a discourse-based one (Section 3) and we do so by portraying normativity as a process that operates on a range of elements that form a continuum of complexity. Our point here is that linguistic usage and discourse patterns are subject to the same normative pressures and regulations. First, both are sensitive to the purely cognitive effects of usage: entrenchment of what occurs commonly, leading to internal norms. Second, both are subject to external norms, as groups develop ‘ideologies’ about the way in which something should be done, in other words 'what is normal and what is not'. For some, this will support their own internal norm, but for others there may be a big discrepancy. And, finally, in both cases, is it ultimately up to the speaker (or ‘agent’) to follow the norms or not. We will illustrate this with two examples, one in the domain of traditional linguistics; the other taken from discourse-based linguistic ethnography.