1 . Many studies of the doctor-patient contact, for example, focus on the way the patient's symptoms and medical problem are constituted in the doctor-pa tient interaction . In his recent work, Paul Atkinson has broadened this approach to include, for example, discussions between specialists (1995). Here also, however, the emphasis is on the constitution of a patient's medical situation through talk .

2. Many have taken similar departures before us: see e.g. Deleuze and Guattari (1988), Haraway (1991) and Mol (1993).

3. Hodgkin's Disease is a form of cancer of the lymphoid tissues. We start out with a thought experiment since we want to convey the sheer impossibility of clinical work without the record. Our fieldnotes have several instances where the record was (temporarily) not available, and then the unwillingness to undertake any potential consequential action is always prominent. For stylistic reasons, however, we preferred introducing Mr. Wood immediately, so as not to introduce confusion by presenting other patients.

4. Although they do not focus specifically on the medical record, the following classical studies provide valuabl e insights on the role of time in medical work: Roth (1979/1963); Richman and Goldthorp (1977); Zerubavel (1979); Atkinson (1981); Horobin and McIntosh (1983); Frankenberg (1992).

5. For the notion of different, co-existing zones or orders of time see e.g. Kavanagh (1995), Frankenberg (1992), and Zerubavel (1979).

6. This refers to the bone marrow treatment: the bone marrow cells are reinfused after high doses of chemotherapeutics have done their toxic yet hopefully curative work.

7 . In a study performed in Sweden it was shown that in 1975, coders disagreed with doctors' attributions of rheumatoid arthritis 38% of the time (Fagot-Largeault, p. 228).

8 . In Rewriting the Soul, Ian Hacking (1995, pp. 234-257) discusses the problem of usi ng current categories - such as child abuse - to describe past actions - such as Lewis Carrol's collections of pictures of children. When the category did not exist in social discourse, he argues, there is a real sense in which does not apply. Action, he argues, is always "action under a description". By this he means that whenever we describe a past action, we only have access to it through a set of constructed categories (a description). Thus when I seek to describe the behavior of someone I consider a "bully", the latter category only fits the action within the context of my middle class latter twentieth century definition of bullying - a nineteenth century English patriarch might have seen someone with a "bit of backbone" (a description not available today...). Hacking's concern is to demonstrate that where there was in the past no category of child abuse there was no abuse (even though, he is quick to point out, there might have been bad action...) (ibid, pp. 55-68). However his argument becomes much stronger when it is applied directly to social organization mediated by record keeping. Where there is no logging of the action of child abuse in a record within an organizational discourse, there is no child abuse. Recording something as having been so makes it so - particularly when the record is itself the only trace of the past action.

9. These three dimensions are not to be taken as exhaustive or even as necessarily central: they are dimensions in which the record's role is spelled out particular ly clearly.

10. We wish to thank Bruno Latour for pointing out the the analytical significance of the centrality of this legal aspect of record-keeping (personal communication).

11. One could say that a specific nursing record does exist - but that record is hardly ever used for research, administrative and/or financial purposes.

12 . See also Goody (1977) and Latour (1987) on the power of writing and written traces.

13. As opposed to for example an ordering by means of a list of the patient's problems, as in Weed's problem oriented record (1971).

14. With regards to the electronic medical record, it is tantalizing to assert a connection between the databases drawn upon and the work organization. There have been three major epochs in the development of databases: the hierarchical, the relational and the object-oriented (Khoshafien, 1993). All three are still in use today - in medical and other databases. The hierarchical database echoes the hierarchical organizational structure most favoured in the 1960s; the relational database echoes the more team model of the 1970s; and object orientation is the nec plus altra of radical outsourcing. Organizational theorists have certainly drawn these connections: arguing a one-to-one relationship between database structure and organization; with mismatches leading to problems. Thus it has been argued that relational databases are sometimes problematic since `they have a flat, two dimensional view of the world' (Stead et al. 1992). Although the co mplex mediations we describe here are unlikely to end up in any rigid repetition, the way a particular type of story gets told, enforced and mediated through different nodes in the network is an important issue for further research.

15 . See Abbott (19 90) for domain knowledge as a professionalizing tool; see Bowker (forthcoming) on nursing records.