7. Conclusions, broader implications

Repair is seen by CA as an important part of spoken interaction, but repair in CMC had not previously been investigated in any depth. Building on investigations of repair in spoken conversation, this research has been able to gain insights into the operation of repair in email discussions. Repair in email discussions differed from that found in spoken conversation where most of repair centres on the effects of real time spoken interaction, such as dealing with simultaneous speech and the self-repair of a term or construction as soon as it has been produced. But email discussions do have instances of all four types of self- and other-initiated self- and other-repair

CA also offers us the concept of "deviant case analysis". The data included a clear example of a deviant case in the BE sample, and this provided a useful point of comparison with the rest of the email data, enabling us to identify features of refusal to repair and pseudo-repair, features which deserve research in spoken conversation.

CA offers a flexible approach for investigating spoken interaction. From the beginning it used multi-party data, and from the beginning Sacks et al (1974) set up an expectation that their work on conversation would be used for comparative purposes. Sacks et al believed that conversation was the basic form of spoken interaction and that other speech exchange systems would be transformations of this. By starting with the least controlled form of spoken interaction then available, conversation, they were able to devise an approach flexible enough to be transferable, both to other kinds of spoken interaction, of which they mention some in their paper, and to kinds of data not then available, such as CMC. This research has shown that it is indeed appropriate to investigate repair in email discussions using a framework based on CA, and suggests that the framework could usefully be used to investigate repair in other forms of CMC.