3.     Coherence and cohesion in written conversation

Here is an example of interaction in a stretch of discourse text from the data source outlined above:

Example 2

Five different participants (six including OrtiK?s personal recorder) take nine turns, some of which are in the third person. Participants enter and exit.. Seemingly unrelated turns are juxtaposed with one another. A number of conversations run in parallel It is these interaction patterns ? the interleaved threads of conversation ? which are considered broadly as a feature of cohesion. Cohesion contributes to the coherence of the discourse by being the ?actual forms of linguistic linkage? (Quirk, Greenbaum, Leech and Svartvik, 1985:1425), that is to say, the linguistic manifestation of coherence.

This section explains why cohesion in SCMC operates in different ways to cohesion in spoken discourse. In subsequent sections conversational threads as they relate to certain types of conversational floor are considered more relevant to an SCMC account of cohesion and coherence in the discourse.

The issue of coherence is central to the study of discourse. Cook?s (1989:4) definition of coherence is: ?? the property of being unified and meaningful.? Discourse analysis itself is defined by Cook (ibid.:14) as: ?? the search for the answer to the problem of what gives stretches of language unity and meaning.? This is to say, discourse analysis is dedicated to discovering what makes language in use coherent for those who use it.

In this section coherence and cohesion are discussed with reference to SCMC discourse. It is suggested that both for participation in and analysis of SCMC discourse, cohesion is not dependent on the coordination of transfer in turn-taking, as it would be in spoken discourse. Rather, broader and looser constructs such as the conversational floor, as described in following sections, are the cohesive ?glue? that contribute towards participants? ascribing coherence (unity, meaning) to the discourse. In examples 1 and 2 above, for instance, coherence was not achieved through an adherence to the same principles of conversational interaction as apply to much spoken discourse.

The search for coherence, in whatever discourse type, and whether by the participants or by analysts, is an interpretive process. It is said that coherence is ?in the eye of the beholder? (Bublitz and Lenk, 1999). In SCMC discourse we certainly appeal to linguistic form at clause level (lexis and syntax) to aid us in the process of ascertaining what gives language in use coherence. On a broader inter-turn level, as with spoken discourse, the formal surface connective linkages of text ? lexical and referential cohesion (Halliday and Hasan, 1976) ? go some way to explaining how a text is coherent. In one-way SCMC discourse of the type described here, such inter-turn cohesion as it exists in spoken discourse is often not readily apparent. However, it is participants who ultimately accord meaning and unity to the text in the discourse process, and they evidently manage to do so. The discourse of SCMC is coherent for its participants; if it were not, it would not be so popular, as noted by Herring (1999).

A token of the lack of cohesion in turn taking ? what Herring (1999) calls the lack of sequential coherence ? is the resultant disrupted turn adjacency (Herring, ibid.). The example below is a typical instance of disrupted turn adjacency. Here and in subsequent examples, turns have been numbered for ease of reference.

Example 3

In comparison with spoken conversation, written conversation displays a reduced sensitivity to coordination of transfer in turn-taking. This can be viewed as a lack of fine tuning.

A number of commentators on linguistic features of SCMC note the dissimilarity of turn-taking patterns in SCMC and in spoken discourse. Cherny (1999) and Herring (1999) discuss these differences in detail. Chun?s (1994:26) remark is illustrative of the view that turn-taking in SCMC is entirely unlike that in spoken discourse: ?In terms of discourse management during a discussion, turn-taking as done in spoken conversation is not a factor in CACD [computer-assisted class discussion].? Kitade (2000:149) notes that there is ?no turn-taking competition? in SCMC.

The lack of fine-tuning in SCMC turn-taking is the responsibility of two fundamental facts of this type of discourse: 1) turns cannot be seen until they are sent; and 2) the visual and auditory (paralinguistic and prosodic) cues which in spoken discourse underpin the turn-taking system are missing. The consequence is disrupted turn adjacency. Herring describes disrupted turn adjacency in SCMC (1999:3): ?? a message may be separated in linear order from a previous message it is responding to, if another message or messages happen to have been sent in the meantime.? And in an early study of SCMC, Murray (1988) notes that: ?? the sender may make a second move before receiving a response to the first and a message may interrupt a turn.?

In the stretch of text in example 3 above, turns 1 and 2 follow the pattern of an adjacency pair (Schegloff and Sacks, 1973). In an adjacency pair the relationship between the first and second pair parts is one of conditional relevance (Schegloff, [1968] 1972). Put simply, the presence of the first pair part is said to open a slot in conversation for an expected, or conditionally relevant, second pair part. MichaelC?s first pair part (turn 1) is followed by the second pair part (2) from Ying-Lan. This response, ?not so good?, is a dispreferred response (Heritage, 1984:265-269; Nofsinger, 1991:71-2). That is to say, although the response is expected, or conditionally relevant, it is not as expected (or preferred) as a response such as ?I?m fine thanks?. Following a tendency noted in dispreferred second pair parts, the response is followed by an elaboration in turn 3. But MichaelC?s next turn (4) seems to be in response to Ying-Lan?s turn 2 rather than turn 3. This is a case of disrupted turn adjacency.

The disrupted turn adjacency in this extract may well be a result of reduced coordination of transfer in that MichaelC was typing turn 4 at the same time as Ying-Lan was typing her elaboration following her dispreferred response (turn 3); it happened that they sent their turns at about the same time, but Ying-Lan sent her turn fractionally before MichaelC sent his. Thus it appears in the log of the chat, and appeared at the time on the screen, that Ying-Lan answers MichaelC?s question before he asks it.

An alternative, and rather problematic, possibility is that the turns appeared in different orders on different computer screens. When considering the technology of SCMC we should recall that not all aspects of the discourse setting are shared. Individual computers have varying levels of processing power. Internet connections have different speeds. Turns thus potentially arrive on different screens in different orders. Evidence of this phenomenon ? system-related disrupted turn adjacency, or lag (Herring, 1999) ? can be seen in the following stretch of SCMC discourse text recorded on ICQ:

Example 4

ying poses the question (turn 3) ?Who is Gerald Ford?? before, it seems, any mention of Gerald Ford has entered the conversation. We can assume that Vance and ying?s turns appeared in the logical sequence on their own screens but, because of a system delay, appear in the logs in reverse order.

We cannot be certain how common a system-related disrupted turn adjacency is. Unless there is some explicit reference to a later turn, the explanation for disrupted turn adjacency might equally be the first posited above: that turns were being written at the same time, and one was posted fractionally before the other (thus appears out of place).

On a broader level, these observations on disrupted turn adjacency tend to support the view that applying models of turn-taking in spoken conversation directly to SCMC discourse is not profitable. For the remainder of this paper, we turn our attention to an alternative perception of cohesion: the conversational floor.