Martin (1992: 501):
In broad terms the development of work on register and context reviewed above can be interpreted in terms of a gradual de-materialising and concomitant semioticising of frameworks for relating language to situation (see in particular Hasan's 1977, 1979, 1985/1989 use of contextual features to generate text structure).
This is inconsistent with Martin's own claim (p497) that the early work of Firth (1950, 1957b, 1957c) developed context 'more abstractly as a level of language'. The confusion of semiotic context with material setting is Martin's, as demonstrated here, here, here and here. The material and semiotic are distinct orders of experience.