2. Background
2. BackgroundThe data for this research are email discussions. Although email discussions are carried out asynchronously and in writing, a frequent perception of participants in email discussions is that they are carrying out some form of conversation. This has led to debates about the relationship between the language of these discussions on the one hand, and spoken or written language on the other. It has also prompted a few researchers to comment on the nature of turn-taking in email discussions, and to suggest that this does not operate in the same way as in spoken conversation (e.g. Murray (1985), p.212).
Email discussion lists were chosen for study because:
Email discussion lists usually focus on one particular topic or theme. This can be recreational, practical, professional, or academic. The lists used in the current study are open lists (i.e. anyone can join) and are publicly available from the JISCmail site (formerly from the Mailbase site). From this site one can join a list, or access the archives of the list. All of the lists used in this study have a professional or academic theme. The ostensible reasons for the existence of these particular lists are detailed in the list information and introductions (see Section 3.2 below). Here it will be seen that their overt purpose is primarily transactional, but an examination of the data will show that they are in fact significantly interactional. Herring (1996), Matthews (2000) and Knapp (1997) have observed a similar tendency. Thus Matthews (2000, p.85) comparing two discussion lists, finds that the "Computer Professionals Forum dialogues did not match the reported communication intention of the members", i.e. that of information exchange, and that the incidence of "socially-oriented messages" was similar to that found in an overtly social group, "the New Age Forum", while Knapp comments (1997, p.183) on the merging of the public and private on email lists, and Herring concludes:
My results suggest that both women and men participate in discussions on electronic mailing lists to exchange opinions, beliefs, understandings, and judgments in social interaction with other human beings, with the pure exchange of information taking second place. (Herring, (1996, p.104))
The finding that email discussions are interactional supports the decision to use the interactional framework of CA for this research.
Conversation is generally unplanned and is produced and managed in real time, with the result that it is subject to innumerable minor problems, as suggested by Sacks (see Section 1, above). However, a particularly satisfying feature of the turn-taking system is that it is used to deal with its own problems. Sacks et al declare that the turn-taking system is used both for the repair of problems in turn-taking itself, and for the repair of other types of troubles in conversation:
The compatibility of the model of turn-taking with the facts of repair is thus of a dual character: the turn-taking system lends itself to, and incorporates devices for, repair of its troubles; and the turn taking system is a basic organizational device for the repair of any other troubles in conversation. (Sacks et al (1974, p.724))
Sacks et al identify "repair devices …directed to, and designed for, turn-taking problems" including "Who me?", interruptions, false starts, etc. (Sacks et al (1974, p.723)). They also identify a device for the repair of simultaneous talk where one speaker might self-repair by stopping. That this is a repair rather than a conventional hand-over of the turn is signalled by the speaker stopping before reaching a transition relevance point. (Sacks et al (1974, p.724)).
Hutchby and Wooffitt define repair:
This is a generic term which is used in CA to cover a wide range of phenomena, from seeming errors in turn-taking such as those involved in much overlapping talk, to any of the forms of what we commonly would call 'correction' - that is, substantive faults in the contents of what someone has said. (Hutchby and Wooffitt(1998, p.57))
Repair includes the strategies which participants use for dealing with errors or with the need to adapt their conversation to the recipient(s).
Repairs to problems in turn-taking in conversation include dealing with interruptions and simultaneous speech (Sacks et al (1974, p.723-4)). As noted above, simultaneous speech is often resolved by one speaker stopping before the end of a turn construction unit. Another technique is identified by Schegloff (1987, passim) who discusses "recycled turn beginnings" which are widely used in the event of overlapping turns. In email discussions these would not be an issue, because interruption of one participant by another is not possible, and simultaneous speech is a feature of the system. So is turn-taking repair necessary and observable in email discussions?
Within spoken conversation, many other kinds of problems are observable, in addition to turn-taking problems. Sacks (above) suggests some sources of trouble; Hutchby and Wooffitt indicate a slightly different range:
There is a wide variety of problems in conversation: incorrect word selection, slips of the tongue, mis-hearings, misunderstandings and so on. (Hutchby and Wooffitt (1998, p.59))
Much repair in spoken conversation takes place immediately after the source of the trouble, in the same turn and by the same speaker (Sacks et al (1974, p.724)). Thus a speaker may make several attempts to produce a word, or may make an immediate correction to a word that has been produced. These problems too, are largely absent from email discussions because the interaction takes place in writing. If participants take several attempts to produce a word, if they decide to change a word or expression, they would normally delete the earlier attempts. However informal the interaction, participants do not leave all of their errors for the other participants to read in order to achieve a conversational effect - the result would be as confusing to the recipient as reading a conversational transcript. But if such errors are edited out before messages are sent, if turn-taking errors of the type discussed above are not applicable to this medium, and if mis-hearings are likewise absent, what kinds of trouble sources, if any, are found in the data? This will be addressed in the analysis below.
Many types of repair have been identified, for example Jefferson (1987) classifies repairs according to whether they are "exposed" or "embedded", i.e. a repair can be explicitly signalled or it can be an unconscious adjustment of talk in progress. The written equivalent of adjustment of talk in progress is editing, but the reader has no access to either the editing process or the original text - we see only the end product, the message as it was sent, and cannot even know whether the text has been edited. So in email any observable repair is usually exposed and explicitly signalled by the writer or another participant.
As Hutchby and Wooffitt note, the properties of the organisation of repair allow "participants to deal with a whole range of trouble sources" and thus "the term 'repair' is preferred to, say, 'correction', or any other term that refers to a specific kind of trouble source" (Hutchby and Wooffitt (1998, p.59)). Indeed repair may be carried out "even when there is no error or mistake in the conversation" (Hutchby and Wooffitt (1998, p.60)), for example to produce a word more suited to the needs of the recipient. The observation that troubles are addressed by use of the turn-taking system strengthens the analysis of the turn-taking system itself, and so repair has been the focus of much CA work.
Discussions about repair usually characterise the repair according to who initiates it, and who carries it out. Ten Have offers a clear explanation of this:
The initiative can be taken by the speaker of the repairable, which is called a 'self-initiated repair', or others can take such an initiative, 'other-initiated repair'. And the repair itself can be done by the original speaker, 'self-repair', or by others, 'other repair'.(Ten Have (1999, p.116))
This results in the four combinations listed by Hutchby and Wooffitt:
Self-initiated self-repair…
Other-initiated self-repair…
Self-initiated other-repair…
Other-initiated other-repair
(Hutchby and Wooffitt (1998, p.61))
Examples of self-initiated self-repair include self-correction in the middle of an utterance (Hutchby and Wooffitt (1998, p.61-2)), or some kind of corrective action when it is clear from the listener's next turn that s/he has misunderstood (ten Have (1999, p.116)).
Examples of other-initiated self-repair include use of a "next-turn repair initiator (NTRI)" (Hutchby and Wooffitt (1998, p.62)) such as "huh?" or "what?" (ten Have (1999, p.116)) which prompts the original speaker to repair. Or the next speaker may "offer a candidate understanding of a target utterance, possibly in a format like 'you mean X?'" (ten Have (1999, p.116)).
Self-initiated other-repair would include situations such as the speaker attempting to remember a name and asking the listener(s) for assistance (Hutchby and Wooffitt (1998, p.61)).
Other-initiated other-repair "is closest to what is conventionally understood as 'correction'" (Hutchby and Wooffitt (1998, p.61)).
There has been very little work on repair in CMC. Even those studies which have identified problems in the interaction have not then extended this to include a discussion of how the troubles are repaired. For example, McLaughlin, Osborne, and Smith (1995) list factors that generate complaints in newsgroups, including technology-related problems, breaking of established conventions, "Inappropriate language" and "Factual errors" (McLaughlin et al (1995, p.96-7)). McLaughlin et al do not link these offences to conversational repair, but find that they generated "reproaches".
Cherny (1995) makes brief references to instances of repair in her MUD data. One example can be found in Cherny (1995, p.219), in which the author observes in the form of asking "clarifying questions" (other-initiated repair), and instances of when a participant deems it necessary to clarify what is being responded to (self-initiated repair), for example when another message intervenes and adjacency had been assumed (Cherny (1995, p.221)).
However, these mentions of problems and brief references to repair-like features do not constitute an investigation into repair in CMC. There is therefore a need to develop an understanding of how repair is carried out in CMC. This paper will begin to address this need by investigating repair in the email data.
This paper will deal with 1) repair in email discussions, 2) what participants in email discussions can do if they feel a turn is in some way inappropriate, and 3) metadiscussions, which can arise as a result of behaviour which is perceived by some participants as inappropriate.
