You are on page 1of 7

Journal of Pragmatics 42 (2010) 26202626

Contents lists available at ScienceDirect

Journal of Pragmatics
journal homepage: www.elsevier.com/locate/pragma

A coding scheme for questionresponse sequences in conversation


Tanya Stivers a,*, N.J. Eneld a,b,c
a

Max Planck Institute for Psycholinguistics, Wundtlaan 1, 6525XD Nijmegen, The Netherlands Radboud University Nijmegen, PB 9102, 6500 HC, The Netherlands c Donders Institute for Brain, Cognition, and Behaviour, PB 9102, 6500 HC, The Netherlands
b

A R T I C L E I N F O

Article history: Received 29 March 2010 Accepted 1 April 2010

1. Introduction In this document, we outline the coding scheme that was developed and used in the 10-language comparative project on questionresponse sequences in ordinary conversation, carried out from 2007 in the Multimodal Interaction Project at the Max Planck Institute for Psycholinguistics. The introduction to this special issue (Eneld, Stivers, and Levinson, this volume) provides the background and motivations for the study, and describes the data supplied by the researchers who applied this coding scheme. Each researcher was asked to code at least 350 questions taken from ordinary conversation among a range of dyadic and multi-participant interactions. The ten single-authored articles contained in this special issue give detailed information concerning the results of the coding scheme for each language. Our reasons for publishing the coding scheme and instructions information here are two: (1) we want to give readers of those articles some insight into how the data were collected and what the ndings mean, including how particular coding categories were operationalized, and (2) we want to provide researchers with a tool to investigate this domain in other languages, for future comparisons with our ndings. This coding scheme is part of a methodology that we think provides an optimal solution to the problem of balancing conicting motivations of ecological validity, on the one hand, and quantitative control and evaluability, on the other. We insist that natural data are indispensable to an account of what people do in social interaction (rather than, say, using questionnaires that are at risk of tapping into ideology about practices rather than actual practices). At the same time, we assume that quantitative data are useful methodologically. This coding scheme was developed through two cycles of pilot coding and evaluation, involving the application of draft coding categories to data from different languages, and collaborative discussion of conceptual and analytic issues that arose. This means that the coding scheme supplied here not only embodies the work of a large number of people (including all contributors to this special issue), it has been developed through iterative consideration in light of language-specic facts and the cumulative experience of analysts in applying the scheme. We therefore feel that the scheme is empirically wellgrounded and analytically well-motivated, and stands a good chance of usefully handling the kinds of distinctions in this domain that are likely to be relevant for any language in any cultural setting. This is not to say that the scheme is perfect. For instance, we found that it was difcult to attain validity or reliability in the coding of some aspects of sequential position and as a result we were unable to include information about sequential position in our comparative analysis. We hope that if

* Corresponding author. Permanent address: University of California, Los Angeles, 264 Haines Hall, 375 Portola Plaza, Los Angeles, CA 90095-1551, United States. E-mail addresses: Tanya.Stivers@mpi.nl (T. Stivers), Nick.Eneld@mpi.nl (N.J. Eneld). 0378-2166/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.pragma.2010.04.002

T. Stivers, N.J. Eneld / Journal of Pragmatics 42 (2010) 26202626

2621

further researchers work with the categories we put forward here, new facts from new languages and speech settings may help to rene and further improve this initial attempt. 2. The coding scheme 2.1. Inclusion criteria In order for a questionresponse sequence to be coded, the following criteria had to be met: (a) A question had to be either (or both) a formal question (i.e., it had to rely on lexico-morpho-syntactic or prosodic interrogative marking) or a functional question (i.e., it had to effectively seek to elicit information, conrmation or agreement whether or not they made use of an interrogative sentence type). Newsmarks such as Really?, Is it? or Yeah? were coded as functional questions. Under our broad categorization of question, as in a), newsmarks qualify because they are routinely treated as seeking conrmation. Questions seeking acknowledgment in, for example, the middle of a story telling the teller solicits specic acknowledgement (e.g., And it was a Weight Watchers recipe right?) were not coded as questions precisely because they sought not neither conrmation nor afrmation. Questions offered in reported speech (e.g., And then he said Arent you gonna come over? and Im like No way.) were not coded as questions. Requests for immediate physical action (e.g., Will you hand me a pencil; Can you open the door?) were not coded if it was a physical action that was the relevant next response. Balancing a desire for a broad conceptualization with a desire to constrain our scope to some extent, we viewed practical actions (e.g., the transfer of objects) as sufciently different from other symbolic vocal and gestural responses such as Mm hm, head nods or points, as to exclude these sequences from our collection.

(b) (c)

(d) (e)

2.2. The coding dimensions Each coder was supplied with an Excel template for which each question below was allocated a column and each question that met inclusion criteria was given an identication number and a row. (1) Is the utterance formally a question? 0=No 1=Yes See inclusion criteria above. Examples of utterances that are formally questions included any question with lexical, morphological, syntactic or prosodic marking appropriate to the language. Thus, rhetorical questions used to express a stance such as Can you imagine going tuh school here? and declarative questions with, for instance, sharply rising intonation, were included. (2) Does the question have lexical, morphological or syntactic marking? 0=No 1=Yes Lexical marking included, e.g., Which or Who questions; morphological marking included question particles or clitics; syntactic marking included inversion as in the Germanic languages. For a number of the languages too little was known about language specic prosodic marking to code this separately. However, for languages where coders felt condent, this could be derived from codes 12. (3) What is the logical semantic structure of the question? 0=Polar Question A polar question is any question that makes relevant afrmation/conrmation or disconrmation. It contains a proposition with two possible answers in semantic terms: true/the case versus not true/not the case. The question might involve a question particle, inversion, or a tag. It did not necessarily involve formal interrogative marking (as in a declarative question). It could be positive or negative. 1=Content (Q-word) question A content or Q-word question (or WH question) is where part of a proposition is presupposed, and the utterance seeks the identity of one element of the proposition. Thus, in Who stole my newspaper it is presupposed that Someone stole my newspaper, and the purpose of the question (at least nominally) is to ascertain the identity of the person corresponding to this someone. Variation in a language in the syntactic position of the Q-Word is not relevant to whether it is coded as such (cf. Where do you work. or You work where?).

2622

T. Stivers, N.J. Eneld / Journal of Pragmatics 42 (2010) 26202626

2=Alternative question Alternative questions included the proposal of a restricted set of alternative answers in their formulation (e.g. Were you drunk or were you sober. or Do you want corn or our tortillas.). Note that just having or in the question does not automatically make it an alternative question. Do you want coffee or, for instance, was not coded as an alternative question because (1) the prosodic contour of these questions is recognizeable as a discrete way of asking a question, and (2) they are routinely treated as a practice for asking a polar question as evidenced by m, 1996). regularly receiving answers (cf. Lindstro (4) Is the question a through-produced multi-question? 0=No (Just one question in the turn) 1=Yes (2 or more questions in a turn) Two or more questions delivered as a single query were additionally coded as through produced multiplequestion questions. These did not include cases in which a single question is followed by lack of uptake or a quizzical facial expression, and the questioner pursues with a second question. An example of a throughproduced multi question is So howd you get them down there how many cars. This code allows for the exclusion of these questions from particular analyses which was necessary since some multi-questions combine polar and content or often the answer is to only one of the questions, so these types of questions must be considered separately. For polar questions: (5) Is the polar question marked with a turn-nal element? 0=No 1=Yes 9=N/A (non-polar questions) These are declaratively formatted turns that assert a proposition and add a turn-nal element that marks questionhood; these turn-nal elements include question particles (e.g., Japanese ka), lexical items (e.g., Right or Yeah?) or tag type clauses (e.g., Dont ya think? or Did she?). (6) Is the polar question negatively marked? 0=No 1=Yes 9=N/A (non-polar questions) Negative marking was coded for negative interrogative questions where the negative is brought up to the front (in English Isnt one of his jobs ve days a week? or Havent you been to Germany,). But, it also included negative declaratives such as You dont have to go downtown do you?. If negation was present in either the proposition or the tag, the question was considered negatively marked. Negative polarity items (e.g., any or never) were not, in themselves, considered negative marking. (7) Is the polar question dubitative (Maybe) marked? 0=No 1=Yes 9=N/A (non-polar question) Among polar questions, if the question had a marker of doubt/uncertainty in it (e.g., I wonder if) then it was coded as dubitative. (This appears to be a grammaticalized way to do polar questions in some languages). (8) Is the polar question a declarative question? 0=No 1=Yes 9=N/A (non-polar question) If a polar question was judged to be a question but was lacking interrogative morphology or syntax and/or was marked as a declarative (e.g., with ge in ), then it was coded as a declarative question; for instance But youre going. or You paid them already? Intonation was not criterial. For content (Q-word) questions: (9) Each content question was coded for which type of Q-word was used but this was done on semantic grounds. If the question was about naming the person whether that was done as Who? or Which person, the question was coded as a who-question

T. Stivers, N.J. Eneld / Journal of Pragmatics 42 (2010) 26202626

2623

0=Who (note, it can be done as which person) 1=What (note, it can be done as which thing) 2=Where (note, it can be done as which place) 3=When (note, it can be done as how late or what time) 4=Why (note, it can be done as for what reason) 5=How (note, it can be done as in what manner) 6=How many/much 9=N/A (not Content Q) General repair initiators such as Huh? in English or H? in Dutch were coded as What questions because they are functionally equivalent to, or elliptical for, larger statements such as What did you say?. (10) Is the question functionally a question? 0=No 1=Yes See above under Inclusion criteria. (11) What social action is the question doing? 0=Request for information (real question) Questions were coded as requesting information only if it seemed that there was no other primary action to be coded. Something ambiguous might be Are you busy tonight? where the action is arguably to nd out whether the real invitation/request can be made (a pre-invitation/request). Thus, in such cases, coders were asked to code Other and list the action as pre-invitation. Other cases were subsequently tabulated and if there was a sufcient number a new category was created. 1=Other initiation of repair (OIR) Questions including open class repair initiators (Huh? or What?) as well as partial repeats (He went where?) were coded as other initiations of repair. If it seemed that the repair was more a challenge than an initiation of repair, Other was coded and the action was listed as challenge. 2=Request for conrmation (non-OIR) Questions (usually declarative although this was not criterial) that asserted a proposition for conrmation such as So youre coming tomorrow night were coded as requests for conrmation. 3=Assessment (stating evaluation; seeking agreement) Evaluations that were formatted to seek agreement such as Isnt it beautiful out today or Shes such a pretty girl isnt she were coded as performing an assessment. 4=Suggestion/Offer/Request Questions that suggest, propose, or offer something to another as well as questions that request something from another were coded in a single category (e.g., Did you want some? [about a breakfast cereal]). This was because there were insufcient numbers to warrant several discrete categories but these actions seemed to cohere in various ways. 5=Rhetorical question Questions that may seek a response but do not seek an answer. For instance, questions that assert an opinion as in Everything comes out in the wash doesnt it? said by a husband to his wife after he has spilled something on the table cloth, were coded as rhetorical questions. 6=Outloud Questions delivered to no one in particular often with lower volume and do not appear to be designed to secure a response (e.g., Now where are my keys. while looking in a bag) were coded as outlouds. 7=Other If the action did not t into the other categories well, then contributors were asked to code Other and list, as specically as possible, the social action that the utterance was being used for. (12) At the point that the question is asked is the interaction dyadic? 0=No 1=Yes 8=Cant be determined from the video

2624

T. Stivers, N.J. Eneld / Journal of Pragmatics 42 (2010) 26202626

To be dyadic there could have been only 2 people present, or if there were multiple people present, only 2 were engaged in the conversation at the moment of this question, and the other people were not attending to this. (13) Is a specic next speaker (or party) clearly selected? 0=No 1=Yes 8=Cant determine Clearly selected could have been through gaze (e.g., the speaker was looking at one of the three individuals present), by epistemic domain (e.g., asking a question about chemistry when only one of the individuals present knows about chemistry), with an address term/name, register (e.g., pitch that indicates it is for a child), volume (e.g., loud or soft might differentiate between a recipient near or far). If only two people were present, all cases were coded as selecting a next speaker unless the question was delivered in a way so as not to require response (e.g., with low volume or looking out into space or out a window). (14) If a next speaker was selected, is the selection done with gaze? 0=No 1=Yes 9=N/A (no next speaker selected) If the speaker gazed towards a particular recipient at some point in the question and this could be seen and understood by the recipient as helping to select him/her to respond, then this was considered to be a case in which gaze served as a form of selection. (15) If a next speaker was selected, is selection done with address term? 0=No 1=Yes 9=N/A (no next speaker selected) If the speaker made use of a name or a term of endearment (e.g., Honey) in the question then this was considered as an address term used for selection of next speaker. (16) If a next speaker was selected, is selection done by domain of epistemic authority? 0=No 1=Yes 8=Cant tell 9=N/A (no next speaker selected) If the requested information was about something within the addressees particular domain of expertise (e.g., his house, homeland, wife, child, mother, profession, schedule, rice elds, animals) then this would be considered to be a case of using domain of epistemic authority as a form of next speaker selection. Coders were asked to code this conservatively such that simply being able to answer did not equate with having particular epistemic authority over the topic. (17) What sort of response is given? 0=None (=>9 for all further cells to end of coding scheme) Non-response was coded if the person did nothing in response, directed his/her attention to another competing activity, or initiated a wholly unrelated sequence (i.e., not something such as a repair initiation that would allow an answer to ultimately be provided). 1=Non-answer response The recipient of the question gave a verbal or visible response that failed to directly answer the question as put. This included laughter, I dont know, initiation of repair (e.g., What?) or other inserted sequences, gestural responses such as shrugs that do not answer the question. Other sorts of non-answer responses included Maybe, Possibly or responses that deal with the question indirectly (e.g., A: Do you see Jack much? B: He moved). 2=Answer Answers directly dealt with the question as put. Answers could be visible (e.g., a head nod or shake) or vocal (Uh huh, Yeah or longer more involved answer including partial repeats of the question to conrm or disconrm). 8=Cant determine (cant hear/see participants, etc.)

T. Stivers, N.J. Eneld / Journal of Pragmatics 42 (2010) 26202626

2625

(18) If a next speaker was selected (see Q13), is the selected next speaker the (rst) one to respond? 0=No 1=Yes 8=Cant determine/Dont know 9=N/A (no next speaker selected) No was coded in cases where Person A selected Person B to respond to a question but Person C answers the question, whether rst (Person B subsequently answers) or as the only one to respond. If two people responded simultaneously, who started rst was coded. In cases of precisely simultaneous starts the selected speaker was coded. (i.e., =1). (19) If there was a response, what is the offset of the response? #=a number in milliseconds (rounded to 20 ms increments) Response offset was measured as the amount of time between the end of the question and the beginning of the response. Response time was measured in milliseconds and rounded to 20 ms increments. All timing was done by observation in ELAN (http://www.lat-mpi.eu/tools/elan). Negative milliseconds indicate overlapping speech whereas positive milliseconds indicate gaps of silence. Response onset was said to have begun if a sound or an inbreath were heard that could be understood as the response. A visible response was coded as having started if the preparatory phase could be understood as the beginning of a response (e.g., the raising of a hand after a Where question would be understandable as the beginning of a pointing gesture). (20) If there was a response, what was the timing of the response relative to the pace of the surrounding talk? 0=No overlap or delay 1=Overlap 2=Delay 9=N/A (e.g., no response) As a second measure of timing, investigators were asked to judge no overlap or delay versus delay using a modied version of Gail Jeffersons transcription procedures (Atkinson and Heritage, 1984). Thus, investigators were asked to judge whether the response began in keeping with the rhythm and speed of the preceding talk. If, relative to the pacing of the talk, a response was late (regardless of whether that would be judged 0.2 or 0.9 s in Jeffersonian timing), then it was coded as delayed. If, relative to the pacing of the talk, there was no discernable delay then the response was coded as neither overlapped nor delayed. (21) If there was a response, is all/part of the response a visible action (e.g., nod or eye brow ash)? 0=No 1=Yes 8=Cant evaluate (e.g., if the investigator could not see) 9=N/A (e.g., no response) Responses that included gestures were not automatically counted. Only if the gesture or bodily action was concerned with providing either an answer or a non-answer response (most typically, head shakes, nods, shrugs or pointing gestures) was it coded here. (22) If there was a response, is all/part of the response a vocal action? 0=No 1=Yes 8=Cant evaluate (e.g., if the investigator could not hear) 9=N/A (no response) Responses were coded as having a vocal component if there was a lexical item present that provided all or part of an answer or non-answer response. This included mm and the like if it was clearly in the service of giving one of these two response types. (23) If the question was a polar question which received an answer, does the answer conrm or disconrm? 0=conrming answer (regardless of form - - yes; no; partial/full repeat) 1=disconrming answer (regardless of form) 9=N/A (not a polar question; did not receive answer)

2626

T. Stivers, N.J. Eneld / Journal of Pragmatics 42 (2010) 26202626

Depending on the design of the question, a conrmation may be either positive or negative in formal terms. For instance, a negative question such as Hes not coming? in English could be conrmed with No. In Japanese, this same question would be conrmed with nn (yes; meaning Yes, thats correct, hes coming). Thus, the coding allows for this and is interested only in conrmation or disconrmation. (24) If the question was a polar question that received an answer, what form does the answer take? 0=repetitional answer 1=interjection answer 2=marked interjection answer 9=N/A (not a polar question; did not receive answer) Repetitional answers included full repeats of the question (e.g., A: Have you gone to the store? B: Ive gone to the store), partial repeats (e.g., B: I have) and modied repeats (e.g., I have) (e.g., Sorjonen, 1996; Raymond, 2003). Interjection answers included yes or no as well as variations such as mm hm and uh huh, head nods and head shakes. Marked interjections included such conrmations as Absolutely, Indeed or Totally. End of coding scheme.

3. Concluding remark With this coding scheme, we hope to provide researchers of social interaction with a well-grounded and well-motivated set of guidelines for the systematic and comparative investigation of questions and their responses from an interactional point of view, where the question is seen not merely as a kind of grammatical structure or illocutionary type, but primarily as a way of mobilizing a response from an interactant. There is much to be learned in this domain, both for individual languages and comparatively; and we offer this scheme as a place to start for further comparative research. The papers in this special issue and in related work (e.g., Stivers et al., 2009, 2010) suggest that this method can provide signicant leverage on the questions of universality and cultural specicity of conversational structures. Acknowledgements We would like to thank all of the contributors to this special issue for their contributions to the development of this coding scheme. We thank Steve Levinson for helpful comments on an earlier draft. References
Atkinson, J. Maxwell, Heritage, John, 1984. Structures of Social Action: Studies in Conversation Analysis. Cambridge University Press, Cambridge. m, Anna, 1996. Afliation and disafliation in Swedish conversation. Unpublished PhD dissertation. Department of Sociology, UCLA. Lindstro Raymond, Geoff, 2003. Grammar and social organization: Yes/No interrogatives and the structure of responding. American Sociological Review 68, 939967. Sorjonen, Marja-Leena, 1996. Repeats and responses in Finnish conversations. In: Ochs, E., Schegloff, E.A., Thompson, S.A. (Eds.), Interation and grammar. Cambridge University Press, Cambridge, pp. 277327. Stivers, Tanya, Eneld, N.J., Brown, Penelope, Englert, Christina, Hayashi, Makoto, Heinemann, Trine, Hoymann, Gertie, Rossano, Federico, de Ruiter, J.P., Yoon, Kyung-Eun, Levinson, Stephen C., 2009. Universality and cultural specicity in turn-taking in conversation. Proceedings of the National Academy of Science of the United States of America 106, 1058710592. Stivers, Tanya, Eneld, N.J., Levinson, Stephen C., 2010, February. Answering Polar Questions in Social Interaction, Workshop on Questions in Conversation (A Cross-Linguistic Project). University of York.

You might also like