form that it did. I have worked on the assumption that the relevant system-builders (here the design team) sought to resolve a set of interrelated problems of varying degrees of obduracy, by creating technical and social components that, when appropriately juxtaposed, would solve those problems. In particular, I have suggested that their assessment of both these problems and their most satisfactory solutions was a function of the relative malleability of the different elements that they had to bring together. The aircraft as an artefact may therefore be seen as the physical expression of what we might call the geography of malleability of a sociotechnical system. It relates the least malleable elements of that system together in physical form thereby creating links that in turn become obdurate. As a consequence, as decisions are made, a network of sociotechnical elements is gradually built up. The durability of this network increases as more and more of the obdurate elements in the system are shaped and brought into it. If all goes well, what was malleable borrows durability from its neighbours, and the whole is turned into a self-bracing structure within which a sociotechnical artefact is born.
In this paper I have tried to exemplify a particular approach to the analysis of technological innovation, an approach that is broadly consistent with analysis of large technological systems that is advocated by Hughes ( 1979; 1983; see also MacKenzie, 1987). I have also, however, made free use of another similar body of work -- that of the sociology of translation ( Callon, 1980; 1981; 1986a; 1986b; 1987; Latour, 1983; 1985; 1986a; 1986b; Law, 1984; 1986b; 1986c; 1987; Law and Whittaker, 1986) -- and the analysis has, in addition, been informed by symbolic interactionist work on the dynamics of 'going concerns' ( Gerson, 1976; Hughes, 1971a; 1971b; Star, 1983). I have tried to show that it is possible to analyse the growth of sociotechnical systems, and hence technological artefacts, by considering a number of factors. First I have suggested that it is important to attend to the concerns or problems of the system builders. I do not think that this is very controversial, though I have argued that such concerns should not necessarily be equated with the social interests that are so often used in explanation by sociologists. I have also suggested that it is possible to understand the ways in which concerns are generated by looking at the relationship between system builders and other actors. Thus, though I have not sought to press the point home fully here, I have tried to suggest that system builders are not, so to speak, primitive