Abstract
This research studies participants’ self-repair practices in instant messaging (IM) interactions. IM is an online forum that allows two people to converse with each other through the exchange of text messages. Same turn, self-repair practices in IM are particularly interesting because here the message composition experience is uniquely poised between the written and oral contexts of communication. The composition is private and textual yet it is for an immediate audience. Drawing on the comparative approach used by Drew, Walker and Ogden (2013), the analysis compares the initial and final versions of participants’ messages and observes the repair operations they perform (Schegloff, 2013), the technology used to effect the repair and the actions thereby achieved. The data used for this research comprise 15 chats between 30 participants. The initial versions of the messages were recorded using a key logging software and these were compared with the final versions retrieved from the chat logs. A total of 671 messages were analyzed to study the ways in which participants self-repair their messages to design turns that are best suited for particular interactional needs. The analysis suggests that the repair operations that IM participants perform are not very different from their oral counterparts and include replacing, inserting, adding, deleting, aborting, searching and sequence jumping. IM participants engage in such repair to achieve significant actions such as upgrading or downgrading their claims, displaying sensitivity to epistemic authority, searching for the right referent amongst others. Besides issues related to turn design, self-repair in IM is used for another crucial function: to address trouble arising out of sequential placement of messages and to construct messages that are most appropriate for an ever-changing semantic environment.
Introduction
Repair, in oral conversation, has been identified as a set of practices, by which speakers address troubles of speaking, hearing or understanding, so that inter-subjectivity, when fractured, can be restored and the interaction can progress (Schegloff, Jefferson, & Sacks, 1977). In any exchange, speakers monitoring their own speech as well as that of their co-interactants, may encounter a segment of talk which is a source of trouble and therefore in need of repair. Talk could be a source of trouble for a variety of reasons, such as when it is not audible, or is imprecise in identifying a referent, or inappropriate for a given recipient, amongst numerous others. On such occasions, conversation analysts have shown that speakers halt the ongoing interaction and address the trouble source. In the episode of repair which begins where the ongoing talk is halted and ends with its resumption (Levelt, 1983; Schegloff, 2000, p. 207) speakers employ various mechanisms to address the trouble source and produce the repair solution. This article is an attempt to investigate participants’ repair practices in a textual and nearly synchronous medium like instant messaging (IM). Specifically, the task is to observe the sources of trouble identified by the participants as they produce their posts, the means available to repair these and the repair practices that they adopt.
IM is an online platform which allows two people who are on each other’s contact list to converse in real time using text messages 1 (unlike webchats where a large group of people can chat anonymously). The two participants who are typically spatially separated, but online simultaneously, interact through the continuous back and forth of text messages. The resultant interaction is uniquely poised between speech and writing (Werry, 1996). Like most spoken encounters it is immediate and co-constructed, yet it is textual not oral. Further, unlike spoken interactions, where the utterance is mutually available as it is being produced, in the kind of IM interactions being studied here, the message is composed and typed privately and is available to the co-interactant only after it has been completed and posted 2 (Garcia & Jacobs, 1999). It is these aspects of the IM context, specifically that participants privately compose textual messages not for deferred consumption but for an immediate interlocutor, that make the study of repair in IM interesting and relevant.
This article focuses on what has been identified in oral conversations to be the most common type of repair: the same turn construction unit (TCU) self-repair (Schegloff et al., 1977). In this type of repair, speakers interrupt their ongoing talk, before it reaches possible completion, to fix troubles in speaking, hearing or understanding. Such repairs are also found in IM interactions where participants repair and correct their ongoing messages in order to construct messages that are best suited for their sequential environment and for the interactional work that they are designed to perform (Drew, Walker, & Ogden, 2013, p. 75). However, before going into the details of self-repair in IM, I want to briefly discuss the issue of whether it is legitimate to use conversation analysis (CA) for studying interactions like IM.
There has been much debate in the field of computer-mediated communication (CMC) with regard to the applicability of conversation analytic apparatus for the study of online interactions like IM (see, for instance, Beißwenger, 2008; Garcia & Jacobs, 1998, 1999; Gonzalez-Lloret, 2011; Herring, 2011; Markman, 2006). In IM, unlike oral conversations, both participants can type and contribute messages simultaneously to the interaction (Beißwenger, 2008; Garcia & Jacobs, 1998, 1999; Markman, 2006) and therefore they do not have to compete for the floor (Beißwenger, 2008; Gonzalez-Lloret, 2011; Markman, 2006). Further, unlike oral conversations the participants do not have access to the messages as they are being produced, the message is available only when the writer has typed it in its entirety and sent it to the common window and therefore the concepts of turn construction unit and transition relevance place are not relevant to IM interactions (Garcia & Jacobs, 1999; Gonzalez-Lloret, 2011). Despite this CA has been very gainfully employed in much CMC research because for all their variance from oral conversations, computer-mediated discourse is interactive and co-constructed and CA provides useful tools to look at such discourse.
Attempts to study repair in chats using a conversation analytic approach have been made previously (Garcia & Jacobs, 1999; Meredith & Stokoe, 2014; Schönfeldt & Golato, 2003). Garcia and Jacobs (1999) are concerned with describing the turn-taking system in quasi-synchronous CMC and comparing this to the practices of turn-taking in oral conversations. They make only a passing reference to same turn self-repair and suggest that since the repair is unavailable to the recipient it is ‘uninteresting’ (p. 351). They, however, make a significant observation related to repair caused by sequencing issues where they observe that in quasi-synchronous CMC, an incoming message from a co-participant may cause another participant to consider his/her message-in-progress as a repairable and may repair this in order to respond to the newly arrived message. The current dataset also suggests that this sort of repair is characteristic of the IM environment and I discuss this in the analysis.
Schönfeldt and Golato’s (2003) research observes repair in webchats—where multiple people can chat with each other—often anonymously. They compare repair practices in oral conversations with those in webchats and find that webchat participants adapt repair practices of oral conversations to the specificities of webchats. They observe that like oral conversations, in webchats too there is a preference for self-repair. With regard to the positions from which participants initiate repair, their data did not afford them access to the same turn self-repair, and they note that the first possible position for repair in chats is the next turn or message.
Meredith and Stokoe’s (2014) study reports two kinds of repair in chats: message construction repair (which is also the focus of this research) and visible repair. The authors define message construction repair as the repair and correction work undertaken by the participants during message composition without any intervention from the recipient (p. 6). Visible repair, on the other hand, is identified as the repair work that is undertaken in the common window and which is available to both the participants (p. 2). With regard to visible repair, the authors find that almost all of these are geared towards correcting production-related errors, whereas repairs undertaken during message construction were found to be related to turn design, action construction, stance and pronunciation.
This article focuses on self-repair in IM interactions. As discussed above, the message-in-progress is not available to the recipients, and consequently neither the repairs nor corrections undertaken by participants. This makes self-repair in IM fundamentally different from that in oral conversations. Unlike the latter, the trouble source, the operation performed to accomplish repair and the final repair solution are all private to the speaker—the one composing the message (see also Garcia & Jacobs, 1999; Meredith & Stokoe, 2013; Schönfeldt & Golato, 2003). In the analysis that follows, I observe participants’ self-repair practices in their posts-in-progress to identify the operations that they perform (Schegloff, 2013), the technology that they use to do so, and the action that the repair brings about (Drew et al., 2013).
Data and Method
The data collected for this research comprise 15 IM interactions between 30 participants. IM, as described above, is a one-one, textual interaction in which two people who are in each other’s contact list, therefore known to each other, can chat in real time. The participants compose their messages privately in the message construction window and then post it to the common window for the recipient to read (see Figure 1).
Amongst all the messengers that provide this service, Yahoo Mail was chosen for this research, for two reasons: first, like Gmail, it automatically saves the chats which can easily be retrieved later. Additionally, it also prefixes each message with a time stamp to indicate the time when it was posted and this information is important for a conversation analytic study.
The repairs that IM participants may perform explicitly in the common window are available through the chat logs but recording the repairs and corrections during message construction requires some intervention by the researcher. Meredith and Stokoe (2014) used the screen capture method for the purpose. In this research, I used a key logging software. The software when installed and running on a computer captures every keystroke of the user. Using this keystroke report it was possible to record the repair and corrections undertaken by the participant for a particular message.
While this software greatly aided the collection of required data, it also had implications on the research design. In order to record IM interactions using this software, participants had to chat using the computer installed with this software and not through their own computer in the privacy and comfort of their own environs. To ask them to conduct a casual conversation in these artificial settings would have been unsettling for them. Therefore, a fake experiment was designed in which the participants were told that the researcher was studying ‘consensus building in online interactions’ and as a part of this they were asked to debate and discuss any one of the 10 topics that were given to them. It must be noted that though this data is not natural, yet as the participants’ attention was focused on the debate and as they were not aware that their repairs were being studied, to that extent even this dataset gives us access to the natural practices of repair in IM.
Fifteen students from the institute I work in interacted with a friend from outside the institute, resulting in a corpus of 15 one-one chats between 30 participants. The software was installed only at the terminal in my institute and therefore key logging reports are available for the messages at one end only. The analysis of self-repair is based on the messages and key-log reports of the 15 students from my institute who produced a corpus of 671 messages across the 15 chats. All the participants were informed of the recording and written consent was sought and all names have been changed to protect participants’ identity. The total corpus of 671 messages was analyzed using the comparative method as discussed below (see Drew et al., 2013).
The following message (see Table 1) is from a chat where the participants were debating whether Google’s growing power is good for the world. In the message below, the participant referring to Google is saying how the availability of such a platform makes the exchange of ideas easy. The ‘final post’ is the post that appears in the common window and one that is available to the recipient, whereas the ‘initial post’ is a record of all the keystrokes that were typed by the user to compose this message. 3 A close reading of the initial post, gives us access to all the repairs and corrections that the participant made before posting the message. 4
[S12-2-10]
In this case, for instance, we see that participant makes one correction on the spelling of ‘availability’ and two repairs—one where she changes ‘is’ to ‘makes’ and second at the end of the message where she deletes ‘and’. In the analysis to follow, I will be using this comparative method to discuss the various kinds of repair operations (Schegloff, 2013) that participants engage in to construct particular preferred actions and to sequence their messages in the conversation.
Analysis
IM participants use the message construction space to correct and repair many kinds of ‘production’ and ‘interactional’ errors (Jefferson, 1974). In this analysis I focus on repair operations undertaken by participants to design messages that are best suited for the particular sequential environment and to do the interactional work they are designed to perform (Drew et al., 2013, pp. 74–75). Meredith and Stokoe (2014), who similarly studied message construction repair, observed how IM participants repair their messages to attend to issues of person reference, stance, pronunciation and intonation (p. 12). In this article I draw on the comparative method of studying self-repair proposed by Drew et al. (2013) to observe the repair operations that IM participants perform on trouble sources. The initial version of their messages (available through the key logging report) is compared with the final version (available through the chat log) to identify the repair operations (Schegloff, 2013) performed, the technology used for effecting the repair and the interactional action achieved. Schegloff (2013) identified 10 operations that speakers perform in same TCU self-repair, including replacing, inserting, deleting, searching, paranthesizing, aborting, sequence-jumping, recycling, reformatting and reordering. It is noteworthy that in a medium such as IM, which is considerably different from oral conversations, participants are found performing many of the same operations to alter the action. The analysis presented is presented in two parts. In the first part I follow Schegloff’s organization (2013) and discuss the repair operations identified in the current data set. In the second part, the focus is on the repair undertaken by participants to orient to issues of sequencing. As we know IM allows both participants to produce messages simultaneously and this, as we will see, is often a source of trouble.
Replacing
Schegloff (2013, p. 43) coins ‘replacing’ to refer to an operation in which a speaker substitutes a different element for a wholly or partially articulated element of a TCU-in-progress. The following instance is from a chat where the participants are debating whether social media like Facebook and Twitter should be banned in corporate spaces. Mayank types the following post as a part of this debate:
Reading the initial version of this post (see the second row of Table 2), we see that, the participant, Mayank, produces the initial part of the message ‘like you know we as humans have the tendency to do or try out exactly what we are not su-’ without any trouble. There is a cut off at ‘su-’ followed by a retracing of the post suggesting that the participant detected trouble in the turn design. The trouble source “what we’re are not su-” is replaced with “what is banned”. One thing that is clear is that the initial formulation had a negation whereas the solution does not. Second, given the context and the partially articulated post, it is likely that Mayank would have completed his post with something like ‘what we’re not su<pposed to>’. If this were the case, we see that the trouble source (what we are not supposed to) and the repair solution (what is banned) are co-class members—both refer to a set of actions or activities that are questionable. Yet the initial formulation ‘what we are not su<pposed to>’ puts the moral responsibility on the users of social media including the speaker and his co-participant, whereas the repaired solution ‘what is banned’ reformulates it as a policy issue or a regulation that has to be adhered to. While this analysis is contingent on the aborted ‘su’ being completed as ‘supposed to’—what is clear is that the IM participant did identify trouble in his post-in-progress, specifically on how he was constructing the referent and replaced this with a preferred option.
[A3-4-7]
Couple of further observations: First, as we will see, the location at which participants identify trouble and initiate repair is not insignificant. In this instance, the participant identifies trouble during the production of the repairable. Other locations from which IM participants have been found to initiate repair are just before producing the trouble source, just after it, or much later—after the whole message has been typed.
Second, unlike oral conversations, where the public nature of the message-in-progress gives the hearer access to the trouble source or the partially articulated trouble source, and the repair solution, and thereby allows him/her access to the speakers preferred ways of speaking, this is not the case in IM. The recipient only has access to ‘what is banned’ and not the first version.
Third, because the trouble source is perceived privately and the repair is also executed privately, we see that Mayank does not use any elaborate mechanisms to signal the initiation of repair, he simply erases the trouble source and replaces it with the preferred version.
In the following instance (see Table 3) the participant, Anaya, performs a repair on a trouble source to replace a less specific formulation with a more specific formulation. The participants are just about to begin the debate when Anaya posts the following message:
[A1-7-2]
Reading the initial version of the post, we see that in the first instance Anaya types out ‘listen, we need to’ which most likely is on its way to be completed by stating some action—say X that they need to jointly perform. But just before she chooses such a completion, the participant notices trouble in the potential formulation. She aborts her ongoing post, erases it, retaining only ‘listen’ and replaces it with the repair solution ‘one needs to be for and one has to be against the topic’ (see also Lerner & Kitzinger, 2007). Here we see two simultaneous repair operations—replacing of ‘we need to do X’ stated as a joint activity that both the participants need to perform with ‘one has to be “for” and one has to be “against” the topic’ where the action for each interlocutor is stated separately. While the action X in the initial formulation cannot be known, what is known is that the collective ‘we’ is replaced with a more specific formulation of what each one of them has to do. The other repair operation is that of reformatting where ‘we need to’ is reformatted as ‘one needs to ...’. In an analysis of a very similar instance, Schegloff (2013, pp. 63–64) identified replacing as the first-order repair operation and reformatting as the second-order operation.
Inserting
Schegloff (2013) coins ‘inserting’ to refer to a repair in which a ‘speaker inserts one or more new elements into the turn-so-far, recognizable as other than what was on tap to be said next’ (Schegloff, 2013, p. 45). Wilkinson and Weatherall (2011) define it as a practice in which the speaker halts their turn-in-progress to go back and add something else into their turn before resuming. Unlike oral conversations where a cut off on an ongoing utterance or a frame is used to set off the repair from preceding talk, in IM, because the perceived trouble and the speaker’s act of producing a repair are both private, IM participants do not use such technology. Therefore, the fact that some elements of the message were inserted later is something that is not accessible to the recipient. From the analyst’s perspective, however, a peek into (and analysis of) the occasions in which IM participants perform repair through inserting additional material is insightful.
On some occasions, as is the case below, participants may insert a phrase, in order to make the message and its referent more specific and unambiguous (see also Wilkinson & Weatherall, 2011). The following is an extract (see extract 1) from a chat where the participants are debating whether private MBA colleges should reserve seats for candidates belonging to disadvantaged sections of society. Kedar has taken a position against reservations, while Sanchita is arguing for them.
Extract 1: [S8-3]
01
8:37:07
Kedar:
If we will give them reservations, it means we are
02
creating a division in the society.
03
8:37:58
Sanchita:
its like a trade off, by providing reservations, you
04
are just trying to level the playing field.
Reading the initial version (see Table 4), we see that Sanchita in the first instance had typed ‘its like a trade off, you are’ (ignoring error corrections and repairs on ‘t’ and ‘u’) but retraces ‘you are’ to insert ‘by proving reservations’ and then continues with ‘you are just trying to level the playing field’. Through this repair we see that the participant prefers the design ‘its like a trade off, by providing reservations, you are just trying to level the playing field’ to ‘its like a trade off, you are just trying to level the playing field’. Inserting ‘by providing reservations’ has added specificity to what is going to level the playing field.
[S8-3-20]
The following is a message from a chat in which the participants are debating whether it is ethical to sell fairness creams. Prior to this message Megha has argued that fairness creams don’t really work, to which her co-participant Shekhar responded by saying that the creams would not sell if they didn’t work. Megha in the current post is making a claim about why users of fairness creams buy the product.
Reading the initial version (see Table 5), we see that in the first formulation, Megha types ‘but people buy the product’ which is revised to ‘i think they buy it only for ...’ (emphasis added). By inserting ‘I think’ Megha hedges her claim about why other people buy fairness creams and thus shows sensitivity to issues of epistemic authority (Drew et al., 2013; Stivers, Mondada, & Steensig, 2011). Megha recognizes the need to repair her formulation and make it sound less certain much later—the repair is initiated after ‘product’. The repair is effected by erasing the first formulation (using backspace) and replacing it with the revised formulation. It must be noted that while the rest of the post is similar in the first and the second versions, there are slight differences, for instance, the objection raising ‘but’ is missing from the revised formulation and ‘people’ is replaced with ‘they’.
[S10-7-8]
Deleting
Schegloff (2013, p. 47) identifies deleting as an operation in which a ‘speaker deletes one or more elements already articulated in part or fully in the turn-so-far’. Below, is a particularly interesting instance of this operation in IM which is quite in contrast to the case we just saw above. Here, as we will see, rather than hedging the claim, the participant deletes some segment of talk, to make a more universal claim.
The instance is from a chat where the participants are talking about sweatshops in India and China that use cheap labour to manufacture goods.
Extract 2: [V15-10]
01
10:13:20
Vijay:
Sweatshops like Sivakasi in Tamil Nadu are good
02
examples of corporate chauvinism
03
10:13:31
Jayesh:
Exactly my point.....
04
10:13:50
Vijay:
One must look at where China was just decades
05
ago.
06
10:13:56
Jayesh:
These are sweatshops that serve Indians staying in
07
India
If we look at the initial version of the posts that appear in lines 6 and 7 (see Table 6) we see that Jayesh in his first attempt has typed ‘Most of these’—where ‘these’ refers to the sweatshops in Sivakasi mentioned by Vijay in line 1. But he initiates repair on this formulation after ‘these’ and replaces this with ‘These are sweat shops ...’. The quantifier ‘most’ hedged the claim about the number of sweatshops being referred to, but Jayesh deletes ‘most’ to make a more universal claim about all the sweatshops in Tamil Nadu.
[V15-10-18]
Adding
IM participants were found to add more interactional material to their message than what they had originally planned. The following is an extract from a conversation where the participants are debating whether private MBA colleges should reserve some seats for candidates belonging to disadvantaged sections of society. Pallavi has taken a stand ‘for’ reservations, whereas Hemant is arguing against reservations:
Extract 3: [S9-3]
01
10:34:43
Hemant:
see if the student is deserving he will get through this ....u
02
can give a good head start by providing provision for
03
preparing for the entrance instead of guaranting a seat in
04
the college
05
10:35:18
Hemant:
that the only point ever body is using caste as a tool to take
06
benefit
07
10:35:34
Hemant:
no one is really concerned about the needy
08
10:35:42
Pallavi:
But then is not the implementation of that really difficult
09
10:35:56
Pallavi:
how do we gauge it
10
10:36:11
Pallavi:
and determine whether it will be really beneficial
We see that in lines 8, 9 and 10, Pallavi is questioning Hemant about the proposal he has just made in lines 1 to 4—where he suggests that rather than reserving seats, candidates belonging to disadvantaged sections of society should be helped with their preparation for the entrance examination. The initial version of lines 9 and 10 (see Table 7) suggests that Pallavi in her initial formulation was asking a single question ‘how do we gauge it’—she posts the sentence final punctuation in the subsequent post (see initial version of line 10). However, she repairs this post, by erasing the question mark and adding more to her preceding question namely ‘and determine whether it will be really beneficial’. The use of the conjunction ‘and’ shows that she has designed this question as a continuation of that in line 9. If we compare the trouble source which in this case is the question ‘how do we gauge it?’ with the repair solution ‘how do we gauge it and determine whether it will be really beneficial’ we see that repairing the initial message and adding more elements to it has helped Pallavi to ask a more pointed question.
[S9-3-29-30]
Searching
During an ongoing turn, speakers may search for certain words or other interactional material to make their description more specific, to address a specific referent, or to speak of a particularly delicate matter (Schegloff, 2013, pp. 49–50). IM participants were also found to engage in searching, particularly for the right way in which to refer or construct the topic of discussion. The following post is from a chat where the participants are debating reservations in private MBA colleges:
Observing the initial version (see Table 8) we see the participant produces the initial part of the turn without any trouble: ‘I feel that reservations should continue till we have a proper’—it is after this that we see the participant searching for the right referent. And as a part of this search, in which ‘primary education structure’ is finally accepted as a good candidate, the participant makes multiple repairs: first ‘basic’ is replaced with ‘primary’ so that the referent reads ‘primary structure’ to which ‘education’ is inserted to read ‘primary education structure’. Thus, we see in this case how replacing and inserting were part of a larger operation of searching.
[S8-3-28]
Aborting
Aborting is a type of repair in which the speaker abandons a turn prior to its completion. Schegloff (2013, pp. 52–53) identifies two cases in which a speaker may abandon an ongoing utterance, one, in which the speaker abandons altogether what is being said, and second in which the speaker abandons the way in which something is being said, to accomplish the same undertaking in another way.
The following instance is from a chat in which the participants are debating whether Google’s growing power is good for the world. The participant who is speaking in favour of Google, posts the following message:
Reading the initial version of the post (see Table 9) and comparing it to what was finally posted (ignoring other errors and repairs such as correcting the spelling of ‘availability’ and replacing of ‘is’ with ‘makes’) we see that the participant, Sneha, in the first formulation types ‘this platform availability to all of us makes the exchange of idea easy and’ but deletes ‘and’. The presence of ‘and’ suggests that in the initial design Sneha had plans to say something more but by deleting ‘and’ she aborts her message and alters the turn design such that it lists only one benefit. Thus we see that deleting ‘and’ is not akin to merely deleting a word, it means forgoing everything that was to come after it. The repair is effected on the trouble source ‘and’ immediately after it has been typed and brought about by simply erasing it. As we will see below, there could be interactional consequences of such repairs. Since this is a debate, everything that one participant says is open to scrutiny and contention by the co-participant. So soon after this post we see Salil, Sneha’s co-participant, taking up the point she made about ‘easy exchange of ideas’ and discussing the demerits of the same:
[S12-2-10]
Extract 4: [S12-2]
01
11:25:54
Sneha:
i think it is very evident that Google is one of the biggest
02
blessing of technology which has given us all the common
03
platform
04
11:26:24
Sneha:
this platform availablity to all of us makes the exchange of
05
idea easy...
06
11:27:58
Salil:
i agree with the fact its the biggest blessing.
07
11:28:01
Sneha:
sure
08
11:28:01
Salil:
but problem is.
09
11:28:07
Salil:
after one point
10
11:28:45
Salil:
or in other words.
11
11:29:00
Salil:
such easy medium to transfer ideas
12
11:29:07
Salil:
can be used negatively.
13
11:29:33
Sneha:
in what context are u sayin dis
14
11:29:42
Sneha:
the negative use?
15
11:29:50
Salil:
videos on tutorials for triggering a bomb from 500 rupees
mobile can be seen on youtube.
Reading the discussion above, we see that the point Sneha makes about how Google makes the exchange of ideas easy (lines 4, 5) is contested by Salil (lines 8–12 and 15). However, we know that she aborted her utterance in lines 4 and 5 to state only one benefit. Thus, we see that what she left out, gets left out from mutual knowledge as well as the from the possibility of discussing it. Aborting—leaving things unsaid—may thus have significant interactional consequences.
In the following instance, the speaker aborts an ongoing message in order to adjust the way in which the action was being brought about. The posts are from a chat in which the participants are debating seat reservations in private MBA colleges for candidates belonging to disadvantaged sections of society. Sancita, who is ‘for’ such reservations, posts the following:
Reading the initial version of the post (see Table 10), we see that Sanchita produces the initial part of the message, ‘they have not had the same facilities as you and me while growing up’—without trouble, it is the subsequent ‘so help<ing>’ that she considers to be a source of trouble and repairs this formulation, by aborting it at ‘help’, erasing the partially articulated ‘so help’ and replacing it with ‘this has hampered their development.’ In the first formulation, the fact that such candidates did not have the same facilities while growing up is perhaps being offered as a reason for us to help them (as can be seen from the use of ‘so’) but in the repaired formulation, in which ‘so help’ is deleted, the same is being offered as a cause for their lack of development. Thus we see how initiating a repair at ‘so help’ and aborting the ongoing post is used to alter the design of the post.
[S8-3-10]
Sequence Jumping
Schegloff (2013, p. 56) identifies sequence jumping as a type of repair operation in which speakers abandon their TCU-so-far and turn to something sharply unrelated. Like all self-repair in IM, this repair too is not available to the recipient, yet the fact that the co-participant has suddenly switched to a different matter, than the one underway is perceptible to the recipient. The following extract is from a chat in which the participants are debating whether it is ethical to sell fairness creams. Kamla and Rohit are discussing the brand Dove which according to Kamla promotes real beauty:
Extract 5: [K5-7]
01
6:37:10
Kamla:
they promote real beauty
02
6:37:18
Kamla:
which lies within
03
6:37:37
Kamla:
and their value proposition is mainly freshness moisturising
04
6:37:45
Rohit:
aren’t they going to loose their sales in near future
05
6:38:15
Rohit:
because as per the consumer behaviuor we always want to be
06
like our ideal-self
07
6:38:20
Rohit:
from actual-self
08
6:38:28
Rohit:
which is not there in this value proposition
09
6:38:34
Kamla:
no. because our generation is modern and respects the fact
10
that beauty is inherent and does not come from skin colour
11
6:38:46
Rohit:
are u sure?
12
6:38:49
Rohit:
13
6:40:03
Kamla:
ok. now enough discussion is done
14
6:40:09
Kamla:
we can close it now
Reading the initial version of the post (see Table 11), we see that in the first formulation, Kamla was continuing with the discussion. And she makes multiple attempts to say something related to the ongoing discussion, but finally chooses to abort and abandon all of this and decides to bring the debate to a conclusion.
[K5-7-39]
The observations presented above show that like speakers of oral conversations, IM participants too repair their messages using the same operations of replacing, inserting, deleting, searching, aborting, reformatting, sequence jumping and adding—an operation that Schegloff (2013) does not mention. In the current dataset I did find not instances of paranthesizing, reordering and recycling. While these repairs are not mutually available, we observe that these repair operations on perceived trouble sources indeed perform much interactional work. In the next section, I discuss how the affordances of IM cause particular sequencing problems for participants and discuss the repairs participants undertake to address this.
Responding to Incoming Message
Observing message construction repair in IM reveals a new source of trouble in these interactions. In IM as both the participants can compose and post messages simultaneously, it is possible that as one participant (say A) is typing a response to some preceding part of the discussion, his/her co-participant (say B) may post a new message—thereby changing the interactional situation, at times quite significantly. Based on data from the key logging software (installed at terminal of participant A) and comparing the initial and final versions of the message being composed by participant A—it was observed that A may abort his/her ongoing action, prioritize the newly arrived message posted by B and respond to it (see also Garcia & Jacobs, 1999, p. 351). As we will see below, it is not a particular stretch of talk— neither the message being composed, nor the newly arrived message from the co-participant—that is a cause of trouble, but it is the new interactional situation engendered by the new message that causes the participant (A) to view his/her ongoing message as a repairable. The following example illustrates this.
The extract is from a conversation where the participants are debating whether Google’s growing power is good for the world, and Sneha and Salil are for and against the topic respectively:
Extract 6: [S12-2]
01
1:29:50
Salil:
videos on tutorials for triggering a bomb from 500
02
rupees mobile can be seen on youtube.
03
a1
11:30:28
Sneha:
But i think then it is upto us to monitor the data is
04
available for which set of audience
05
b1
11:30:32
Salil:
All the sites which should not be allowed to
06
children under certain age group, are easy
07
accessible through a single google search.
08
11:31:10
Sneha:
for all this we have parental locks and monitoring
09
has to be done by the parents
I want to draw the reader’s attention to Sneha’s post marked a1 (lines 3 and 4) and the message she was composing immediately after this marked a2 (see Table 12). If we combine what Sneha has posted on lines 3 and 4 with what she was typing immediately after this, her message reads as ‘But i think then it is upto us to monitor the data is available for which set of audience (a1) and try to limit on the data’ (a2). Her use of the conjunction ‘and’ at the beginning of a2 clearly shows that she was composing a message in continuation to something she had said earlier. However, as she is typing this, there is a new message from her co-participant—marked b1. If we go back to the initial version of the post (Table 12)—we see that Sneha deletes her original message ‘and try to limit on the data’ and types a response to the newly arrived message ‘for all this we have parental locks and monitoring has to be done by the parents’ which appears in lines 8 and 9 (see Extract 6). Thus, in IM interactions we see a new source of interactional trouble, engendered specifically because both parties can produce messages simultaneously. I discuss below a set of extracts that reflect the diverse instances found in the data.
[S12-2-18]
The following is an extract from a conversation where the participants have been debating whether it is ethical to sell fairness creams. This extract is almost from the end of their conversation. In line 1, Kamla makes an attempt to conclude the debate but Rohit does not take this offer up and instead asks a question (line 3). The ‘they’ in line 1 refers to fairness creams and that in line 3 refers to the company Dove.
Extract 8: [K5-7]
01
a1
6:36:17
Kamla:
so can we conclude that they are not ethical but not
02
completely unethical
03
b1
6:36:42
Rohit:
what is the value proposition are they offering
04
6:36:50
Rohit:
*remove are
05
6:37:50
Kamla:
they promote real beauty
In a1 Kamla has typed ‘so can we conclude that they are not ethical but not completely unethical’ and at a2 (see Table 13) she continues ‘as they do give confidence and happiness to a p’ but she deletes this (using Ctrl x) and responds to Rohit’s question b1. For Kamla it was the choice between posting the second half of her concluding statement and responding to Rohit’s question. Thus we see that she uses the message construction space to make a contribution that is appropriate for the changed semantic environment. Same turn self-repair provides us with strong evidence of what participants consider as more appropriate forms of actions in particular sequential contexts and this instance shows that Kamla considers it more appropriate to respond to the question than pursue her conclusion.
[K5-7-35]
Prioritizing the incoming message over one’s own ongoing turn has interactional consequences. The first is obviously that it makes the interaction more dialogic and immediate. Else, given the fact that IM participants can produce messages simultaneously, it is technically possible for them to pursue two completely independent monologues. Second it plays a role in what gets topicalized. For instance, in the example above, Dove’s value proposition becomes a topic of discussion for some time after which Kamla initiates closure again.
Discussion
This article has attempted to add to the limited literature on repair in computer-mediated conversations (Garcia & Jacobs, 1999; Meredith & Stokoe, 2014; Schönfeldt & Golato, 2003). By focusing exclusively on self-repair during message construction, the research has attempted to draw attention to the work done by IM participants towards maintaining inter-subjectivity in the interaction. In oral conversations, repair practices are known to address troubles of speaking, hearing and understanding. In IM, as the messages are textual rather than oral and further as they are typed, not handwritten, trouble rarely arises out of perceptual difficulties. Trouble in IM is caused by production errors such as typographic, spelling and grammar errors as well as due to interactional issues. In this article I focused on self-repair in the message construction space where participants alter the design of their messages so that these may best perform the intended action for a given sequential environment. Such self-repair practices in IM are fundamentally different from their oral counterparts, because in IM, the entire repair episode is inaccessible to the recipient. In oral conversations, a comparison of the fully or partially articulated trouble source and the repair solution allows the recipient access to speaker’s preferred orientations to the situation. Such a heuristic is not available to IM participants. Yet, as this article has hopefully been able to demonstrate studying self-repair practices in this medium adds to our understanding of the phenomenon of repair and also to our understanding of users communicative practices in this new medium which is simultaneously textual and interactive.
Observing participants’ self-repair practices revealed that the repair operations that IM participants perform are not very different from speakers in oral conversations which include replacing, inserting, adding, deleting, searching, aborting and sequence jumping. Further, as the repair execution is private, we see that participants do not adopt any elaborate mechanisms to initiate repair, they simply erase the trouble source and replace it with the preferred design. Also unlike oral conversations, repair in IM does not halt the progressivity of the interaction—it may slow down the interaction—but the recipient cannot discern this. The analysis showed how participants achieved many significant outcomes by altering the design of their turns including upgrading or downgrading the claims made, searching for the right referent, increasing the specificity of the action, limiting the scope of the message by aborting it prematurely amongst others.
Another significant role played by repair operations in IM is to enable participants to adapt to the changing sequential environment. As is known, in IM, both the parties can post messages simultaneously. Because of this affordance, at times, a newly arrived message can change the interactional situation considerably. It was observed that IM participants use the message construction space to abort and delete their ongoing message and respond to the newly arrived message—enabling them to make more contextually relevant contributions.
If future research records keystrokes at both ends we will get a fuller picture of the interaction. Also studying repair practices in a variety of mediated interactions will help us identify those of our repair practices that are universal and those that are specific to the medium of interaction.
Funding
The author gratefully acknowledges the funding she received from her home institution, MICA, to pursue this research.