Halliday & Matthiessen (2014: 27-8):
Why then do we refer to them as different things? We can see why, if we consider some recent arguments about global warming, the question is asked: ‘Is this a long-term weather pattern, or is it a blip in the climate?’ What this means is, can we explain global warming in terms of some general theory (in this case, of climatic change), or is it just a set of similar events? An analogous question about language would be if we took a corpus of, say, writings by political scientists and asked, are these just a set of similar texts, or do they represent a sub-system of the language? The climate is the theory of the weather. As such, it does have its own separate existence – but (like all theoretical entities) it exists on the semiotic plane. It is a virtual thing. Likewise with the system of language: this is language as a virtual thing; it is not the sum of all possible texts but a theoretical entity to which we can assign certain properties and which we can invest with considerable explanatory power.