Currentresearch into talk-in-interaction focuses on the turn-taking system andextracts interest from the way in which participants observe the course of aturn in progress and predict the end of that turn.
This paper presents a ConversationAnalytic study of the distribution of pauses and gaps throughout theturn-taking system, instances of overlap and the behaviour of participants inspeech exchange systems. This article will review literature on the basicturn-taking mechanism and research that focuses on the implications of CA workfor how we think about language processing. It will analyse both the process ofcomprehension and production and the constant shift between the two inconversation. The micro level of analysis this paper will provide allows for amore fine-grained understanding of turn projection. Finally, this analysis willdemonstrate how this notion of turn projection is a key component in theorganisation of turn-taking.
Keywords:turn-taking, 1.IntroductionAccordingto CA, the projectability of TCUs plays a fundamental role in the ways in whichconversation is structured and in the process of listening to talk in progress(Liddicoat 2004). Listeners must process the turn in progress in advance inorder to be aware of speaker change and to achieve a smooth floor transfer, twoessential properties of a successful speech exchange. In termsof temporality, participants do not process a previous utterance once it hasoccurred and wait until it finishes prepare their own. Conversation would bemuch more drawn out with substantial gaps after every turn. Rather, the modelof turn-taking proposed by Sacks et al. (1974) involves considerable projectionof what the other interlocutor might utter and when their turn may end. Participant-orientedevidence exists in speech exchange systems for predictive languagecomprehension and for gaps in language production.
As Pomerantz (2012)describes participation in a conversation and assessment of that talk areclosely related processes. One way to account for inter-speaker gaps oftenbeing shorter than intra-speaker latencies is the process of overlapping inconversation. Participants demonstrate an assessment of this occurrence bykeeping any competing turns they may produce brief with one participant oftendropping out.
For thepurpose of this paper, it is important to differentiate between pauses, gapsand lapses and to define between-overlaps in comparison to within-overlaps. Silencescan be distinguished according to their length, the preceding turn and the turnwhich follows the silence. Pauses, therefore, refer to silences within turnsand on average last around 600ms and higher according to Levinson and Torreira(2015). Turns with such pauses usually consist of a grammatically incompleteturn or an “um” demonstrating place-holding behaviour (Mushin and Gardner2009). Gaps, however only average around 200ms and refer to those that requirea floor transfer between participants occurring at transition-relevance places(TRPs) or between turns.
Finally, extended silences between turns are known aslapses (Sacks et al. 1974) which often occur when no-one has selected to speak.With regards to overlaps, we can differentiate between two kinds according towhen they occur in conversation. Between-overlaps can be defined as floortransfers that occur without any gap between participants and within-overlapsare those with overlapping inter-pausal units that don’t cause floor transfer(Heldner and Edlund 2010). The following sections will present examples of suchfeatures extracted from a corpus of data and demonstrate how they aresignificant features of talk-in-interaction. 2.Turn-taking as a systemSacks etal. (1974) provide the foundational work on turn-taking and define it as a formof organization for conversation demonstrating this system using a model withvarious rules and observations.
The model consists of TCUs which can bedescribed as potentially complete turns (Schegloff 1996) which end with a spaceknown as a TRP which makes relevant a speaker change but does not require it. InCA, projectability is seen as being fundamental to the turn-taking model.According to the model, a complete TCU must be understood as such by another interlocutorwithin a given sequential context implying therefore that the other participantmust be able to project the action of the preceding turn and when that turn mayend (Liddicoat 2004).The modelclaims that social interactions are regulated by social norms which suggestthat one party talks at a time however do allow for open participation. Thisrefers to the fact that talk-in-interaction is highly flexible and occurs in amoment-by-moment fashion and therefore overlap may occur. According to therules of the model, overlap is expectable and contributes to the vast majorityof transitions (Sacks et al. 1974).
One such rule of the model claims that twoparticipants may self-select themselves as the next speaker resulting inoverlap. This may also be the case if a participant tries to demonstrate anunderstanding of the current turn where exactly the same words feature in theoverlap as in (1). (1)McLaughlin – First Dates 1.
Sarah: >thatsgood that they were there for ya< .hhh >cause like lads can be< (0.6)?like ither (0.3)-2.Phil: the $?MA(h)d jo(h)kesºorº- 3.Sarah: >?Yeah the like<
Phil: yeah its:5. Phil: (0.4) ?I think it helps that (0.6scratches brow) (h h h-) In (1),the word “yeah” feature twice in an overlapping nature between lines 3 and 4.
In this case, the two participants seem to agree on a single understandingwithout saying much about it. Phil has just explained how his friends werethere for him after the loss of his mum when Sarah states that this was a goodthing by implying something about “lads” however she doesn’t actually make anystatement. Phil seems to make a guess at what she might be referring to atwhich point the two of them form a kind of agreement or mutual understandingwith the overlapping “yeah”s.
This may be due to the fact that Sarah hasn’tmade any direct observations herself and neglects to do so in line 3 where Philmakes an attempt to show that he understands self-selecting himself to take thefloor, perhaps before Sarah was finished. Overlapmay also occur if a participant misinterprets the end of a turn. This is commonin conversation however the rules also predict that latencies during a turnwill be longer and more frequent than those gaps between turns as participantscan actually often predict when speaker-change may occur, the main focus ofthis article. This is shown in example (2), suggesting that participants do havea good understanding of when a turn may end, that is, they are able to predictthe end and can therefore prepare and swiftly produce their own response butwith no overlap (Levinson 2015). (2) is an example of latching in conversationwhere one turn occurs immediately after the previous turn without any pause.
(2) McAllister– First Dates6. Stu:
Theexample demonstrates that prediction plays a key role in the process oflistening as Jen’s turn does not overlap with Stu’s in line 5 but responds withno gap suggesting that she may have thought he had finished his turn as he hadcompleted an action sufficiently, that is, of providing her with an explanationfor his loss of sight. In turn, line 7 suggests that Stu may have alsopredicted that “oh wow” concluded Jen’s response and her turn and thereforeselected himself for the next turn, again with no gap or overlap. This idea ofno-gap-no-overlap was prominent in early work of the projection theory howevermore recent studies show that transitions with slight gap are actually morefrequent than those with none (Heldner and Edlund 2010) and it is more accurateto state that universally, the target in turn transition is to minimise gapsand instances of overlap between turns. Othercomponents of the turn-taking model include basic observations such as theoccurrence or quite often recurrence of speaker-change, a varied turn order andsize allowing for longer units when mutually arranged and an indeterminatenumber of participants (Sacks et al.
1974). 3.Distribution of gaps I have previouslydefined the silences in conversation differentiating between differentinstances and when they occur, suggesting that lapses are often longer thanpauses, however, pauses are often longer than gaps. Of course, this is notalways the case.
Gaps which occur between question-answer sequences for examplecan often present longer gaps especially if a wh-question occurs. Naturally, amore complex question specifies a longer answer than those expected of polarquestions, therefore more complex responses which require more preparation willresult in a more substantial gap between turns as in (3). On the other hand,question-answer sequences are one of the few instances which make a floortransfer relevant where at least one participant is waiting on a response. Inmost other cases, floor transfer is optional and it is normal for participantsto consider whether or not they want to contribute any more to theconversation. (3)McAllister – First Dates 9. Jen: (0.8) what sort of women do you go for (0.
7)10. Stu: i’vehad a run with a few (.) ?gingersI guess they would call themselvesginger Line 8 demonstrates an example of a pause at the endof a syntactic unit but within a single turn: “ (0.
8)” This pause constitutes a TRP. This suggests that Jen has madean attempting to leave the floor open for Stu to self-select for the next turnhowever when he doesn’t do so, she continues, producing a wh-question and thusexplicitly selecting Stu to respond. While the purpose of this example is tohighlight a substantial gap between turns, note that the focus of this paper isstill evident in that the first pause is still slightly longer than theinter-speaker gap, supporting . Referring now to the effect of the wh-questionand addressing the question of how such action was accomplished, we can observethat quite a long gap was produced implying that the response in line 9required some preparation, more so, than that for a simple yes/no question. 3.
1 The role of silences in dispreferred responsesSacks etal. (1974) suggest that a preference exists for no gap between turns inconversation however it is evident that silences do occur in this way in anumber of instances. With questions, for example, if an interlocutor asks aquestion they will expect some kind of response and as participants are verysensitive to timing, extensive gaps in this TRP can be seen as problematic. Inother words, silences of 700ms or longer after a question constitute asdispreferred responses (Kendrick and Torreira 2015).It is alsoimportant to note that gaps between turns do not only represent participantsstill in the progress of preparing what to say.
Other reasons exist for thesesilences, for example, as a politeness strategy. In other words, gaps may occurbetween turns in an interaction if one participant wants to avoid anyimposition, confrontation or embarrassment (Brown and Levinson 1987; Nakane2007). The context of extract (4) presents a typically uncomfortable settingwhere both participants of the date are asked whether they’d like to see each otheragain by a third interlocutor.
It is important to note that Ben has alreadyanswered the question with a yes and the extract below demonstrates Tamzin’sturn to answer. (4) Reid –First Dates11. Ben: $I thinklet’s face it (1.
3) its more (h) about (h) = =whether (h) she’s gonna? See me? $12. Tamzin: (h)13. Tamzin: (h) (0.
6) 14. Ben: She’s the one with the (.) fussychecklist.15.
Ben: (h) 16. Tamzin: (h)17. Tamzin: It’s not a no. (h)18. Ben: It’s not ano? 19. Tamzin: (h)20.
Ben: Is it a yesthough? (1.0)21. Tamzin: That’s not what I’m saying.
The silence in this extract can be viewed as anegative politeness strategy where Tamzin appears to use silences as a form ofdistancing tactic (Nakane 2007) as she avoids providing a yes or no answer tothe question of seeing Ben again. The first gap in line 13 is 600ms almostmeeting the time proposed for a dispreferred response and in fact, resulting inBen self-selecting again. In line 20, Ben poses another question: “Is it a yesthough? (1.0)” explicitly selecting Tamzin to speaker next which is followed bya longer, quite extensive gap of 1 second. This implies that Tamzin may want toavoid answering with a definitive yes or no in attempt to avoid anyembarrassment. Extract (4) also suggest ordered rules within theturn-taking system.
The inter-speaker gaps already mentioned demonstrate howtypically it is expectable for Tamzin to have rights to the next turn unit especiallyas both turns beforehand are questions however as she does not speak at all,Ben continues with rights to the turn unit (Levinson and Torreira 2015).4. OverlapPreviousevidence claims that 80% of the transitions in face-to-face conversation are gapsand 20% are partial overlaps (Levinson and Torreira 2015). Overlaps are mostlikely to occur at turn transitions however they are brief and subject torepair in that when two participants speak at the same time, very often onewill drop out quite quickly thus repairing the trouble. Overlap can occur inmany situations, some of which are already discussed in section 2.
Another briefbut very common instance of overlap is a setting where a participant may entersome environment and several other participants greet them simultaneouslycreating an instance where a vast number of speaker utterances may overlap witheach other. Accordingto Sacks et al (1974), simultaneous starts may occur as an example of overlapwhen both participants self-select at some possible TRP, for example, inextract (5). (5) O’Hanlon– First Dates 22. Toby: Youalright, (.) nice to meet you23. Bree: yeahnice to meet you ((kissingsounds))24. Bree: ?hi25. Toby what’syour name sorry?Lines 24 and 25 illustrate an overlap where bothparticipants have self-selected at the same time.
Interestingly, Bree was thelast participant to speak before they kiss and she still self-selects tocontinue after they do so. In this case, both participants project possiblecompletion points of the other and one will drop out. In this instance, Breemay only have planned to say “hi” however there is also the possibility thatshe dropped out as she predicted that Toby’s turn was a question not close tocompletion. 5. Turn projectionTen Have(2007) suggests that conversational flow is responsible for turn projectionwhich he claims is vital for both language production and comprehension insitu. Oneinstance which claims that participants often predict the others’ utterance andwhen they may stop is demonstrated in bi-clausal structures where thecompletion of a turn may be carried out by another participant (Lerner 1991). 5.1 Turn-final cues Thesources that participants use for projectability depend on various properties knownas turn-final cues.
These sources include syntactic closure, pragmatic actionsor gestures and prosodic cues. With regards to prosody, certain intonational contourswhich are consistent with the ends of turns such as phrase-final syllablelengthening. 6. The Corpora The dataused for analysis in this research is of naturally occurring conversations sourcedfrom a British reality television show and compiled into a corpus ofrecordings. These recordings were collaboratively transcribed by a class ofstudents using CA practices. The transcriptions have thus provided evidence forthe claims made by this paper.
7. Experimental studies of turn-takingThemethods involved in CA involve acquiring data that is naturally occurring andmechanically recorded. This is important as hypothetical or reinventedinstances of talk are difficult to construct in the same way that they occur inactual conversation and therefore they may not be thought to be reasonable byan audience (Sacks et al.
1974). The next step is to then transcribe the dataand make observations such as those made in this paper. As CA is typicallyinductive, it is then these observations that help to provide a theory.
Inother words, conversation analysts often extract turns from theirtranscriptions of talk-in-interaction and observe fine details. While thecorpora for this work is somewhat restricted in the nature of videos from YouTube,data of this kind allow for the actual temporality of the conversation, includingpace and silences which can now be analysed. However, with regards to projectability and languageprocessing there have been many experimental studies carried out whicheliminate the constraints that this restricted corpus faces.
One example, fromDe Ruiter et al. (2006) employed the use of a button by participants when theydetected the end of a previous turn. Other studies involve picture naming tasksand eye movement tracking which suggest that planning is required forproduction and provides a way to time this process. Processing speed can beseen as the main difference between language production and languagecomprehension with the latter occurring at three or four times the speed ofproduction (Levinson and Torreira 2015).
While these methodologies are lessconstrained, they also lack free interaction and in contrast to CA approaches,the participant’s response is often of non-linguistic nature. As this paperaims to present a conversation analytical study in the way of examining datafrom actual conversational practices, it will view the listener as an activeparticipant in the interaction rather than a passive recipient of incomingspeech items (Liddicoat 2004). 8. ConclusionIntra-speakergaps are longer by c. 150 ms than inter-speaker gaps (ten Bosch et al.
, 2005),suggesting ordered rules (the rights to the next turn unit belong first to thenext speaker, and only if not exercised, to the current speaker).Accordingto Schegloff (2000), participants’ turns are often co-ordinated and theirresponse times often change to match their interlocutors’.