LA CONFIGURACIÓN DEL ENTE

Having introduced the data and the theoretical framework underlying this study in chapters 3.1 and 3.2, this chapter outlines how the data was prepared for the analysis. As a first step,

I compiled collections of different types and scenarios of speaker change to allow further investigation. Starting from collections of instances, which are then analysed in greater detail, is classic CA methodology: Collecting how one specific interactional practice is realised in concrete conversations allows scholars to detect recurrent “default pattern[s and] […] also reveals the operation of underlying preferences or principles [as well as] […] the identification of ‘deviant’ instances” (Sidnell 2009a: 16). Therefore, I first identified all places of possible speaker change in the conversations, i.e. I marked all TRPs – thereby taking into account the features and characteristics explained in chapter 3.2.1. As described above, TRPs are central for turn-taking, they are the ‘places of action’ in the conversation. That is, they play an essential role for the interactants – both current speaker and potential next speaker(s) – who orient themselves towards these places, e.g. by deciding whether to claim, hold or yield a turn, or to encourage the turn occupant to continue talking. Scholars can then piggyback on these members’ analyses. As Schegloff puts it, the “[m]anagement of the interface or conjunction of action [undertaken] and implementing utterance is a key task of the parties and a key topic for disciplinary analysis of talk-in-interaction” (1996: 58f).

As a next step, I analysed each TRP in greater detail, focusing on three aspects: (1) the type of turn allocation which took place,

(2) the scenario it triggered, and – if present –

(3) the strategies speakers used to claim or hold a turn. Each of these notions is discussed in the subsequent paragraphs.

In all interactions analysed, speaker change basically follows the patterns described in Sacks et al.’s 1974 paper. That is, in order to achieve minimal gaps and overlaps in a (natural, i.e. non-institutional) conversation, three types of speaker change can occur (cf. Sacks et al. 1974: 704):

• (type 1a) The current speaker selects a next speaker – either directly, e.g. via address terms or gaze, or indirectly, e.g. via topic choice;

• (type 1b) no next speaker has been selected by the current speaker and another speaker self-selects; or

• (type 1c) no next speaker has been selected and the current speaker continues talking.

As a pilot study showed, these three types of transition can be further sub-divided into three scenarios each. After a next speaker has been selected, transition can occur immediately (scenario (1a:1)) or after a small gap (scenario (1a:2)). Gaps are defined as “silence[s] after

a possible completion point”, i.e. a TRP, by Sacks et al. (1974: 715, annot. 26), and thus differ from pauses, which occur within the TCU. However, even though Sacks et al. originally speculated that the majority of speaker changes takes place without any gap or overlap (1974: 700), they later modified their statement. In fact, so-called ‘unmarked next position onsets’, i.e. transitions which involve an inter-turn silence of approximately 0.2s, showed to be more common (Jefferson 1984a: 18f; Heldner & Edlund 2010: 563f). This duration is also the threshold when it comes to whether speakers actually notice a gap in everyday conversation (cf. Walker & Trimboli 1982). I therefore decided to treat only transitions with offset times of more than 0.2s as (1a:2)-scenarios. Apart from that, a number of next speaker-selections could neither be classified as (1a:1) nor as (1a:2), for instance, because they involve overlaps or trigger repair sequences. These cases were tagged as (1a:3) and are discussed in greater detail in chapter 4.1.1 below.

Similar scenarios can be observed when no next speaker has been selected and a participant self-selects at the TRP. If transition takes place after the current speaker has abandoned their turn, i.e. without overlap, this was coded as (1b:1), if it does not involve a silence of more than 0.2s and as (1b:2), if the silence is longer. Self-selecting speakers can also actively claim a turn, using the strategies described above (1b:3) (cf. chapter 3.2.2). As this scenario turned out to be the most frequent case of self-selections, it was further classified according to the strategy type(s) employed. An overview of the tag symbols used for codification is given in table 3.5 below.

Finally, current speakers who continue talking might do so only after a lapse (1c:1). Lapses are defined as “extended silences at transition-relevance places” by Sacks et al. (1974: 715, annot. 26), but the authors do not specify when a gap can be classified as “extended”. In fact, many studies do not even distinguish between lapses and gaps at all, because both involve silences at a TRP (e.g. Heldner & Edlund 2010: 556). However, lapses differ from gaps in a number of aspects: First, whereas gaps are short but noticeable silences and can occur in combination with any type of speaker change, lapses can only arise if no next speaker has been selected, i.e. if rule (1a) does not apply. They are the consequence of “a series of rounds of possible self-selection by others and self-selection by current to continue […] in none of which are options to talk exercised” (Sacks et al. 1974: 715). In other words, lapses are due to the current speaker trying to yield their turn and none of the other interactants starting up. Furthermore, while lapses are a typical by-product of task- based activities, they are noticeable to participants of everyday conversations and often mark the ending of an interactional sequence (Hoey 2018: 329f; cf. also Schegloff 2007:

194). That is, lapses fulfil a number of specific functions in interaction, which is why it makes sense to differentiate them from unmarked gaps. In my analysis, I follow Hoey (2018) and define a lapse as a period of inter-TCU silence which is 0.5s or longer. That this duration is perceived as extended by interactants, is suggested by studies with Dutch or English data, which show that current speakers regularly resume their turn and other speakers are more likely to initiate repair after this period (Kendrick 2015; ten Bosch et al. 2005). Apart from continuations involving lapses, current speakers might also actively hold their turn, using a number of strategies (1c:2). As with turn-claimings, I grouped these strategies and coded them (cf. table 3.5 below and chapter 3.2.2 for a more detailed overview). And finally, other speakers might yield their turn and encourage the current speaker to continue talking, for instance by providing continuers (1c:3).

To sum up, the three general types of speaker change can be sub-divided into a total of nine different scenarios. Two of them, active turn-claimings and turn-holdings were further classified, as they involve the use of additional resources surrounding the TRP (cf. chapter 3.2.2 above for an introduction and overview). In most cases, speakers use a combination of different strategies to hold or claim their turn (Clayman 2013: 153). Cut-offs, for example, are typically accompanied by glottal stops, i.e. a syntactic resource is combined with a phonetic one. Apart from that, syntactic turn-holdings are often combined with lexical planners, such as er or erm (Clayman 2013: 153). (3.10) illustrates a typical instance of how a prospective next speaker uses several strategies to claim a turn:

Example (3.10): George W. Bush (ICE-JA, S1A-010)

01 N.N.: [(I=wasn’t) (0.2) >(spo[ken to)]=[ ( ) < ]= 02 Sue: =[I=↑mean ↑BUSH]= 03 Mar: =and=Bu[sh is a very (.) ↑rational]

04 Sue: [>Bush- some ↑of- some ↑of-] some of Bush’s< ↑statements [are scary in that respe]ct!

Two speakers, Sue and Marie, are claiming the floor from the unidentified speaker in this extract. Both are combining different strategies to achieve their goal: Sue starts up at a TRP in the current speaker’s turn and continues talking despite the overlap, raising her volume and announcing her intention to grasp the floor: I mean BUSH (l. 02). She restarts several times and persistently repeats her utterance in fast succession, until the other speakers drop out (l. 04). Marie begins in latch position (l. 03) and syntactically links her TCU to the previous turns. She makes sure that she is the first starter after the turn occupant stops talking and continues well into Sue’s overlap.

In order to investigate which of the strategies frequently co-occur and whether the speaker groups differ in that respect, I coded the larger resource categories, i.e. I distinguished lexical from phonetic strategies etc. Table 3.5 gives an overview of the coding schema used for the analysis.

Table 3.5: Overview of tag symbols used for the analysis

Type of speaker change Scenario of speaker change Turn-holding/-claiming strategy (if used)

Tag symbol Explanation Tag symbol Explanation Tag symbol Explanation31

1a Current speaker selects next 1 Speaker change without gap 2 Speaker change with gap 3 Other32 1b Next speaker self-selects 1 Current speaker abandons turn (no gap) 2 Current speaker abandons turn (gap) 3 Turn-claiming by next speaker I Interruption La Latching L Lexical ‘warning’ signals O Overlap P Phonetic strategies S Syntactic strategies 1c Current speaker continues 1 Continuation with gap 2 Turn-holding by current speaker La Latching L Lexical planning strategies O Overlap P Phonetic planning strategies S Syntactic planning strategies 3 Continuation with active turn-yielding by second speaker(s)

31 _{For a more detailed explanation, please see chapter 3.2.2 above.}

On the one hand, this tag system facilitates the qualitative analysis of my data, particularly, as I am working with several hours of multi-party interactions, leading to very long and detailed transcripts. The coding system resulted from an in-depth qualitative analysis and also helps to ensure the reliability of my results, as being able to systematically search for specific features makes it easier to double-check my findings and enables further comparison between different practices (cf. Hopper 1989: 57). As mentioned above (cf. chapter 3.2), this approach is not contradictory to CA methodology but is directly derived from them, provided that coding is not imposed on the data from above but developed out of it. According to Stivers (2015), combining CA with formal coding comes with three limitations: hard boundaries, freezing the analytic frame, and the potential of an inappropriate use of mixed-methods. I am aware of these caveats and make sure to take them into account in my investigation. For instance, coding is necessarily reductive to a certain extent. It requires the researcher to resort to clear categories with hard boundaries, which obviously do not replicate the complexity of real behaviours. When looking at how interactants who have been selected as the next speaker react, for example, coding seems to be unproblematic at first glance: Speakers can start up immediately (1a:1) or after a small gap (1a:2). (3.11) and (3.12) illustrate these two scenarios:

Example (3.11): Telephone number (ICE-JAM, S1A-006) 01 Luk: you ↑have (.) >the number for the man there?< 02 Bob: yeah man

Example (3.12): Daughter (ACE, VN_LE_con_pho restaurant) 01 Thu: your ↑first?

02 (0.3)

03 Tem: ↑yeah (0.2) DAUGH↑TE:R

In (3.11), the current speaker Luke directly addresses Bob and asks him for a telephone number. Questions are classic FPPs of adjacency pairs; they require another speaker to provide a relevant SPP to be complete. In the excerpt, Bob’s response follows immediately, i.e. with an offset time of less than 0.2s (l. 02). Similarly, (3.12) also depicts a Q&A- adjacency pair. This time, however, the selected speaker, Tembam, only answers after a short gap of 0.3s (l. 02 and 03). Both extracts are unproblematic with respect to formal coding, as they constitute relatively clear cases – (3.11) would be classified as (1a:1), (3.12) as (1a:2). Categorisation becomes less clear, however, when looking at an example like (3.13):

Example (3.13): Psychology (ACE, SG_ED_con_4)

01 An: ok[ay ]

02 Zhi: FPPbase [but] you still have strong (ins) in psychology,

right?

03 Wan: FPPins I still have what?

04 (0.4)

05 Zhi: SPPins er: s- ins- s- strong interests in psycholog[y? ]

06 Wan: SPPbase [yah],

>of course< (.) my master’s was in: clinical psy[chology], yea[h]

As before, the current speaker, Zhi, opens up an adjacency pair by asking a question addressed to another conversationalist, Wan (l. 02). Wan starts up immediately, but she does not provide the answer SPP. Rather, she points at a trouble source in Zhi’s talk, the unclear pronunciation of ins ‘interests’: I still have what? (l. 03). Through her question, Wan is other-initiating repair; she opens up a second adjacency pair within the overarching base Q&A-sequence. In order to be able to complete the base adjacency pair, the problem in Zhi’s question has to be solved, and the interactants do this by means of a so-called post- first insert expansion sequence (l. 03 and 05), which breaks the contiguity of the base pair (l. 02 and 06) (cf. Schegloff 2007: 100ff). Wan’s inserted FPP (FPPins) is thus addressed to the original question (the FPPb) and establishes an immediate relevance for Zhi to provide repair of the trouble source. As soon as Zhi repeats her utterance and corrects her pronunciation (strong interests, l.05), Wan completes the base adjacency pair. Contrary to the previous examples (3.11) and (3.12), situations like (3.13) pose a problem for formal coding. Wan’s utterance in line 03 fits neither of the two classical scenarios observed for speaker change after next speaker-selection. One way to solve this problem would be adding more scenarios to the coding system, e.g. a code for next speaker-selection triggering repair. However, this still does not capture the rich variety of situations – just think of Wan’s answer in line 06, which is neither repair-initiating nor appears ‘in the clear’ but overlaps with the end of Zhi’s question. I cannot solve this problem when creating my coding scheme, as no formal system can be detailed enough to depict the complexity of real life. This is why the qualitative analysis of the data remains so central, particularly when dealing with cases like (3.13).

Furthermore, only close qualitative analysis of the data can prevent what Stivers calls “freezing the analytic frame” (2015: 14f), i.e. the danger of overlooking unusual or previously undetected cases. In my analysis, for instance, the occurrence of alveolar clicks in turn-initial or turn-medial position in the Southeast Asian data group would have been missed without a close bottom-up analysis. And thirdly, while a quantitative analysis can depict broad tendencies, it cannot capture the intricate makeup of human interactions. That

is, while I am able to comment on general types of turn-taking or on larger groups of similar strategies (such as the group of ‘phonetic strategies’ etc.), the complex interplay of these variables can only be grasped by close qualitative analysis. In fact, “quantifying CA practices is not always appropriate, nor is it always analytically productive” (Stivers 2015: 15). One of the interactions I analyse, for instance, takes place in a lively restaurant and features seven participants (ACE, VN_LE_con_pho restaurant). Multi-party conversations of more than three participants are likely to split into several conversations, as “any pair of parties not getting or taking a turn over some sequence of turns can find their mutual accessibility for getting into a second conversation” (Sacks et al. 1974: 713). Such schisms are speakers’ reactions to larger groups of conversationalists – the one-party-at-a-time-rule is preserved, while more participants get a chance to speak (Sacks 1992 [1968]: 34). In the seven-party-conversation, speakers often split into smaller groups, which leads to several simultaneous conversations. While this makes the interaction particularly interesting for a qualitative analysis, formal coding cannot yield usable results for a quantitative investigation and only leads to a large amount of unclear cases. I therefore decided to exclude conversation VN_LE_con_pho estaurant (ACE) from the quantitative analysis, which leaves approximately two hours of formally coded conversation for each data group.

Nevertheless, as mentioned above, CA should not refrain from employing quantitative methods. As Stivers puts it, “[a]s long as the practices under study include an investigation of interactant orientation already, formal coding can provide a second story of the analytic house, thus improving the view and reach of CA research” (2015: 16). The second part of my analysis (chapter 5) is therefore concerned with examining the relationship between the findings from the qualitative analysis (chapter 4) and the speakers’ cultural backgrounds.