Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

Journal of Psyeholinguistic Research, Vol. 4, No.

2, 1975

Studies in Dialogue and Discourse: II. Types of Discourse Initiated by and Sustained Through Questioning ~
Elliot G. Mishler 2

Received May 2, 19 74

A sociolinguistic approach to the study of language is described. Three types of discourse are defined empirically, each o f them initiated and sustained by questions. It is prooosed that these types are linguistic expressions Or "realizations" o f authority relationships between speakers. In the first, Chaining, a conversation is extended through successive questions by the initial questioner: in the second, Arching, it is extended by the respondent's questions; in the third, Embedding, there are two responses to the question. The general hypothesis about the relationship of the linguistic structure of discourse to differentials between speakers in power and authority is examined through an analysis of natural conversations in first-grade classrooms. The findings provide consistent support for this hypothesis. When adults initiate a conversation with a question, they retain control over its course by successive questioning, i.e., by Chaining," when children ask an adult a question, the adult regains control by responding with a question, i.e., by Arching. Children question each other through a more balanced use of Chaining and Arching that might be thought of as either more egalitarian or more competitive.

INTRODUCTION In this second report of findings from a study of natural conversations in first-grade classrooms, we describe three types of discourse and examine their differential use by adults and children. One general class of conversations is
This work has been supported by a grant from the National Science Foundation for research on "Social Context Effects on Language and C o m m u n i c a t i o n , " Grant Noo GS-3001. 1The research staff for the study included Lorraine Landy Hell as senior research assistant, Linda F. Butterworth and Christine R. Ehrlich as coders, and J o h n Rees and Edward DeVos as computer programmers. 2Harvard Medical School and Massachusetts Mental Health Center, Department of Psychiatry, Boston, Massachusetts. 99 (~)1975 Plenum Publishing Corporation, 22"7 West 17th Street, New York, N.Y. 10011. No park of this publication may be reproduced, stored in a retrieva! system, or transmitted~ in any form or by any means, electronic, mechanical, photocopying. microfilming, recording, or otherwise, without written permission of the publisher.

I O0

Mishler

the focus of research, namely, an exchange between speakers that both begins with and is extended through questioning. The various ways in which a conversation may be extended, or sustained, through the use of questions provide the basis for a typology of discourse. We propose that the different subtypes of question-sustained discourse reflect role relationships between speakers, particularly along a dimension of authority and power. This proposition is examined empirically by comparing such conversations between different questioner-responder pairs-adult to child, child to adult, and child to child. The ways in which social relationships are "realized" in. the structure of language is the central topic of the larger study from which these data are drawn; our approach to the analysis and interpretation of findings is guided by our concern with this general problem. In our work, the basic unit of analysis is referred to as dialogue. In question-initiated conversations, the only type of conversation with which we are concerned, the dialogue unit consists of three successive utterances: the Question, a Response-utterance from a second speaker, and a Confirmationutterance from the initial questioner. This question-initiated sequence of three utterances is called an Interrogative Unit, or IU; it is one class of dialogue. This conception of a three-utterance unit of conversation is not usual in linguistic studies of questioning, where the question-answer exchange is treated as the basic unit of communication (e.g., see Robinson and Rackstraw, 1972). However, although not usual, the conception is not unique. For example, Bellack et al. (1966) report that in their study of high school classes the "fundamental pedagogical pattern of discourse consisted of a teacher's solicitation followed by a pupil's response; this sequence was frequently followed by a teacher's reaction" (p. 55). Most, though not all, solicitations were questions; the similarity is evident between their Solicitation/Response/ Reaction and our Question/Response/Confirmation units. From another perspective, discussing basic "rules" of two-party conversation, Sacks (1972) proposes a normative basis for a three-utterance unit that begins with a question, namely, that the questioner has a "reserved right to talk again" (p. 343). Thus, despite the general emphasis in the field on the question-answer exchange, it is possible to find some empirical and analytical support for the choice of a three-utterance unit. What is of more importance, however, is that this approach allows us to formulate discourse in terms of dialogue: that. is, connected units of dialogue constitute discourse. There are three primary ways in which questions may serve to connect dialogue units and thus produce a type of discourse that is both question-initiated and question-sustained. These are referred t o as Chaining, where the Confirmation-utterance contains a

Studies in Dialogue and Discourse


question; Arching, where the

101 R e s p o n s e - u t t e r a n c e c o n t a i n s a q u e s t i o n ; and

E m b e d d i n g , w h e r e t h e r e are t w o r e s p o n s e s t o the initial q u e s t i o n . E x a m p l e s o f these t h r e e " ' m o d e s " o f c o n n e c t i o n are given b e l o w 3 : l. Chaining Q SI: Did you really finish all your work before everybody else? $2: Yeah. C/Q SI: What kind of work was it? $2: Well. We jus had a piece of paper and it had all the names on it. And we had to see if it was odd or even.

SI: Do you know all the odd numbers and all tile even
numbers? $2: Uh huh. SI: How far up can you count?

2. Arching S l: Whaddya mean? He is (S) He's R/Q ~ " , ~ ~ C C/R $2: (S) Is he under water? SI: He is. See now here's a scuba suit on him. $ 2 : 1 think he's dead.

"~The separate functions of each utterance as Question, Response, and Confirmation are marked as Q, R. C. Multiple functions of utterances in different but connected Interrogative Units are indicated as C/Q or R/Q. In the Chaining example, the third utterance functions as a Confirmation in the first IU and as a Question in the second IU and is marked as C/Q. In the Arching example, the second utterance is a Response i n the first IU and a Question in the second, thus R/Q; the third utterance is a o Confirmation in the first IU and a Response in the second and is noted as C/R. In' Embedding, one Response is treated as primary and marked as R, the Embedded" Response as ER. The notations of (S) and underlining indicate utterances or parts of utterances that occur at the same time: that is, they are instances of simultaneous: speech. Fragments and false starts, such as "He is" in the Arching example, arc included without punctuation.

102 3. Embedding Q SI: Ohhh. What's he got on here? $2: (S) Skiis. $3: (S) Skiis. SI: Woooow. I would love to ski.

Mishler

ER

An Interrogative Unit (IU) may be connected to another successive unit by any one of these modes but only by one of them. That is, any single pair of connected Interrogative Units may be tied together by either Chaining, Arching, or Embedding. However, one unit may be tied to two or more successive units through any combination of these modes. For example, if both the Response- and Confirmation-utterances include questions, then the unit may be linked to one unit through Arching and to a second through Chaining. To illustrate this, a more complex series is presente.d below4:

~
~Ad: The discourse, connected successive

Ad: BI: Ad: BI: Ad: B1:

Whaddyawant? Is this the same? Yuh. What d'ya wanta know2 (S) Do you have a question? (S) Is that the same? Is that the same as this? Yeahis that tile same as that? Uhhhh, d'you think it is? No~ OK. You're right.

first paper in this series (Mishler, 1975) focused on the length of that is, on the frequencies of series with different numbers of tUs; attention was also given to variations in length and in rates of questioning as a function of differences between questioners and

41n this discourse series between an adult and a boy, there is a sequence of five successive questions, from the third through the seventh utterance. The fifth and sixth utterances, questions from the adult and the boy, respectively, serve different functions in three different Interrogative Units. That is, each of these Question-utterances is a Question in one IU, a Response in a second 1U, and a Confirmation in a third IU~

Studies in Dialogue and Discourse

t03

responders in terms of age and sex. We found that the rate of successive questions in connected discourse of this type could be approximated by a simple exponential fuilction. Briefly, the probability of a "next" question, following an exchange that contains a question, appears to remain constant in that it is not affected by the number of previous questions. In other words, the probability of a question seems to be relatively independent of the length of question-sustained discourse. The value of the constant, in the exponential formula, was found to vary with the identities of speakers: adult-initiated conversations had higher values than those initiated by children; that is, the constant probability of a question was higher in the former than the latter series. These findings suggested that certain features of discourse could be represented as a stochastic process, namely as a finite Markov chain. In these earlier analyses, no distinctions were made between series that were tied together in different ways. That is, we were interested only in the length of a series and not in how it was constructed. In this paper, we shali examine the use of the three different modes of connection and will be concerned primarily with the following questions. What is the relative frequency of use of Chaining, Arching, and Embedding? Is length of discourse related to the type(s) of connection used? Are there systematic relationships between types of connection and the age and sex characteristics of questioners and responders? Relationships between our approach and other analytical and empirical work On questioning and discourse will be discussed in the concluding section of the paper. METHODS The source of data is a collection of conversations tape-recorded in four first-grade classrooms. Portable cassette tape recorders were carried by selected children as they went about their usual business in class; a lead-in microphone was taped to a chest harness strap. Sixteen different children were equipped at different times with a tape recorder~ although never more than two during a class session. In addition to the nine boys and seven girls who were the "participant-recorders," many other children in the classroom as well as the teachers and adult observers are included in the taped conversations. The four classrooms were in three different elementary schools; although all schools are in the same suburban community, the pupil populations represent a range of social class variation from upper working class to upper middle class. Over a 7-month period from December to June in the 1971-1972 school year~ 81 separate tape recordings were collected each consisting of a 45-rain cassette. The subsample of typescripts used in these analyses consists of material drawn

104

Mishler

from 26 recordings of the original corpus, distributed over the school year and across target children. The tape recordings were processed in several stages. First, conversational episodes were selected 1;hat met the criterion of a continuing exchange between speakers, one of whom included the target child (i.e., the child with the tape recorder), which would generate at least one page of triple-spaced typescript. Only these selected portions of the recording were typed. Instructions for typing emphasized accuracy: no grammatical alteration was permitted; fragments of speech, nonlexical expressions, and interjections were all noted, The typescripts were then checked and corrected by a second listening. Within these episodes, all Question-utterances were located and marked. Dialogue units, consisting of successive Question-Response-Confirmationutterances were then "mapped" on the typescripts, as were the ways in which these units were connected together into different types of discourse (see previous examples of "mapped" units). These mapped units of dialogue and discourse were then coded; codes were transferred from optical scanning sheets to IBM cards and computer tapes for analysis, s

FINDINGS

Rates of Chaining, Arching, and Embedding


The frequencies of Chaining, Arching, and Embedding in all connected discourse series, of any length, are presented in Table I. Since a series consisting of one pair of IUs can be connected by only one of these modes, but longer series may have combinations, single and multiple pairs are separated in the table. Chaining is clearly the most frequently used way of extending a conversation through questioning; only about one-third of all series do not include Chaining, Embedding is absent in about two-thirds of the series, and the rate of Arching falls between them. The rates of Chaining are higher for both single and multiple pairs. Embedding and Arching are relatively equal in rates of single pairs, but Arching shows considerably more multiple pairing. On the whole, the probability of a dialogue unit being connected to another through Chaining is about twice as high as through Arching, and about two and one-half times as high as through Embedding. This means that it is much more likely for the initial questioner to ask 5The complete coding manual, a lO0-page document that includes mapping instructions and detailed codes, is available from the author. There is a charge of $5.00 to cover costs of duplicating and mailing.

Studies in Dialogue and Discourse Table 1. Rates

105

of Embedding, Arching, and Chaining in Discourse Series: (Frequencies and Percentages of Paired Units) a Number of discourse series with paired Units connected by Chaining Arching N 139 56 53 248 % 56.0 22.6 21.4 100.0 N 164 68 16 248 Embedding % "66.1 27.4 6.5 100.0

Number of paired I Us None One Two or more Total


i

N 78 l O0 70 248
i

% 31.5 40.3 28.2 100.0

aOnly discourse series that include two or more Interrogative Units are included in these tabulations, since by definition isolated IUs cannot he connected. The number of isolated IUs, including both incomplete and complete IUs, is 206, or 43% of all discourse series that have the minimal pair of questioner and responder. There are 248 discourse series that are two or more IUs in length. Tile table may be read as follows: in 3!.5% of such series there are no Chained pairs, in 56.0~. no Arched pairs, and in 66.1%, no Embedded pairs. In 40.3% there is one Chained pair. etc. another question, within the same dialogue exchange, than it is for the responder. In this study, we have been concerned primarily with certain features of conversation as "realizations" of social relationships and social functions. That is, we are interested in the ways in which social reality is constructed t h r o u ~ language. We have assumed that questioning is one o f the ways through which one speaker attempts to exert control over another. For us. it is a realization or an expression of authority relationships. Through the act o f questioning, one speaker defines the way in which the other is to continue with the conversation and thus defines their relationship to each other along a dimension of power and authority. This control function of questioning may be obscured by the assumed equivalence of the two terms " t o ask" and " t o question." Although often considered synonyms, they differ in their respective ranges o f connotative meanings. For example, Webster's New International Dictiona~ (2nd edition) offers the following synonyms for "ask": need, entreat, beseech, petition, implore. All of these suggest that the questioner is in a subordinate or subservient position vis-a-vis the answerer. However, the synonyms for "question" include challenge, demand, dispute, call into question, examine, charge, accuse, and doubt~ Here the questioner is clearly in a superordinate or dominant position vis-h-vis the answerer and a different complex of meaning is involved.

106

Mishler

These ambiguities reflect the social "meaning" or functions of questioning. A question, that is, a sentence defined as a question on linguistic grounds, may serve to either "ask" or "question." Throughout this work, we have tacitly assumed that we have been studying "questioning" rather than "asking." In accord with this view, we excluded certain types of questions from analysis since they appeared to ,"ask." In particular, we excluded permission-seeking questions such as "May I leave the room?" and questionresponses to summons or calls such as "What?" or "Huh?" to "John" or "Come here." More precisely, these types of questions could not initiate a discourse series, but they were retained in the analysis if they appeared within a series. By this partial elimination of this class of "asking" questions, we hope to have reduced some of the ambiguity and to be in a better position to treat the remainder as "questioning" questions. Following this line of argument, we hypothesized that the mechanism of Chaining would represent a higher degree of control and authority on the part of the questioner than would Arching, since the former involves an attempt to continue control into the next cycle of the conversation. In Arching, on the other hand, the respondent is attempting to gain control over the conversation. An extended discourse series that was tied together through Chaining would therefore indicate a high degree of control on the plart of the questioner; it would be a "realization" of an authority differential between speakers. The overall findings, in Table I, on relative rates of Chaining and Arching are consistent with this general framework of interpretation in which questioning is viewed as the exercise of control. That is, if we ~ a n t that the asking of an initial question represents an act of control then it appears that this gives the questioner an "edge" in that he/she is more likely to control the beginning of the second cycle of the conversation than is the respondero There is a concordance between these statistical findings and Sacks' norm, but we are interpreting the questioner's "reserved right to talk again" as an exercise of social power. To ask a question in response to a question !s an act of countercontrol and may require either that there be a "'true" differential in social power where the responder has more real authority or, at a minimum, that there be equality between the speakers. The problem of Embedding is more complicated since it involves responses from two speakers to the same questioner. This might suggest a particularly high degree of power on the part of the questioner. However~ about 80% of all instances of Embedding occur in discourse series that contain only one pair of IUs; its function may be clarified in the more detailed analyses which follow.

Studies in Dialogue and Discourse Differences in Modes of Connection by Questioner-Responder Pairs

107

The social meanings of these different modes of connection may be examined in a more precise way by taking into account the "true" reality differentials in power and authority among our speakers. Our conversations take place in first-grade classrooms, and the authority differential between teachers and pupils is an obvious part of this setting. On the basis of our earlier discussion, we would expect Chaining to be a relatively more frequent method of extending a conversation when the teacher is the initial questioner and Arching to be more frequent when the child is the initial questioner. We would also expect Embed.ding to be more frequent in response to an adult questioner and Arching to be relatively frequent when an adult responds to a child's question. Rates of Chaining, Arching, and Embedding controlled by the identities of questioner and responder in the first pair of IUs in the series are presented in Table II. The left half of the table distinguishes between children and adults (usually but not always the teacher); the right half adds the further distinction of the sex of the child. The findings bear out our expectations quite consistently. Chaining is present in 84% of all Discourse series initiated by adults; it is present in 59% of child-initiated series where a child i s t h e first respondent and in 67% of these series where an adult is the first respondent. The pattern reverses for Arching, which is found in 30% of adult-initiated series and in 46% of child-child series. The child-adult series show the highest rate of Arching, with 78%. Thus, although Chaining is used more than Arching as a general mode of connection, the disproportion in use is considerably more marked for adults. Only within the child-adult series do we find Arching used with more frequency than Chaining. This is of particular interest, since Arching involves a "questioning" of the questioner and therefore represents an effort at countercontrol. This finding, that the adult respondent ~ the childquestioner at such a high rate, lends further support to the underlying idea that "questioning" is a mode of communication t h r o u ~ which authority relationships are realized. Figures in the right half of Table II show rates of Chaining, Arching, and Embedding when the sex of child is taken into account, both as questioner and as responder. In adult-initiated series, Chaining occurs at a higher rate if the respondent is a boy rather than a girl. The difference is relatively slight when it is a matter of simple presence or absence of Chaining; it appears in 86 and 81% of adult-boy and adult-girl series, respectively.

108

Mish~r

Table I1. Questioner and Responder by Frequency of Pairs of Interrogative Units Connected by Chaining, Arching, and Embedding (Percentages of Paired Units in Discourse Series with Two or More Interrogative Units) a
.1~ i II II I I[ III

(N) Chaining Adult-child Child-adult Child-child

None One

Two+

(N)

None One

Two+

(87) 16.1 32.2 (27) 33.3 37.0 (134) 41.0 46.3

51.7 29.6 12.7

Adult-boy Adult-girl Boy-adult Girl-adult Boy-boy Boy-girl Girl-boy Girl-girl Adult-boy Adult-girl Boy-adult Girl-adult Boy-boy Boy-girl Girl-boy Girl-girl Adult-boy Adult-girl Boy-adult Girl-adult Boy-boy Boy-girl Girl-boy Girl-girl

(57) (22) (10) (15) (41) (12) (11) (41) (57) (22) (10) (15) (41) (12) (11) (41) (57) (22) (10) (15) (41) (12) (11) (41)

1 4 . 0 28.1 57.9 22.7 3 1 . 8 45.5 60.0 20.0 34.1 50.0 54.5 39.0 68.4 68.2 30.0 40.0 53.7 50.0 36.4 39.0 1.8 4.5 10.0 40.0 12.2 0.0 9.t 22.0 29.8 27.3

Ar~:hing Adult-child Child-adult Child-child

(87) 70.1 (27) 22.2 (134) 53.7

2.3 44.4

27.6 33.3

30.0 60.0 10.0 1 3 . 3 3 3 . 3 53.3 48.8 58.3 54.5 46.3 63.2 50.0 80.0 80.0 75.6 50.0 63.6 70.7 34.1 17.1 16.7 25.0 36.4 9.1 3 6 . 6 17.1 28.1 36.4 20.0 6.7 22.0 41.7 36.4 29.3 8,8 13.6 0.0 13.3 2.4 8.3 0.0 0.0
I

31.3 1 4 . 9

Embedding Adult-child Child-adult Child-child

(87) 58.6

28.7 1 2 . 6 7.4 2.2

(27) 81o5 11.1 (134) 67.9 29.9

aWhen the sex of the child could not be identified, the speaker was coded as C in the typescripts~ These instances are excluded from the right half of the table and account for the reductions in N from the original totals for each pair of speakers. However, the difference is especially marked in tonger series. For example, in series that include three or more Chained pairs (not presented separately in Table II) almost all instances involve a boy as the r e s p o n d e n t - o f the total of 27 adult-initiated series with three or more Chained pairs, 24 begin with an adult questioning a boy. This suggests that adults engage in longer conversations with boys through a mode of control, namely, questioning~ The

Studies in Dialogue and Discourse

109

relationship between the length of a series and the modes of connection used will be taken up in more detail in later analysis. When we turn to series initiated by a child's question and responded to by an adult, we find that the presence of both Chaining and Arching is higher for girl-questioners than for boys. Girls alSO show higher frequencies of multiple occurrences of both of these modes. Whether or not these differences are associated with different lengths of discourse is again an important question. However, since the pattern of differential use (particularly for Chaining) reverses in the comparison of adult-child and child-adult discourse, then one of two possibilities exists. Either adulbboy conversations are longer than adult-girl ones and, at the same time, girl-adult conversations are longer than boy-adult conversations or these types of conversations do not differ in length but there is a difference in the ways in which boys and girls enter into conversations with adults. We have already suggested that adults may exert a higher degree of control over their conversations with boys than with girls, using the evidence of the relatively greater amount of Chaining in the former. To the degree that higher amounts of both Chaining and Arching are found when girls initiate a conversation, this would indicate that they make more active efforts to control their interaction with adults than boys do. Intersex comparisons within the child-child series show slight differences between same-sex and cross-sex pairs. Both Chaining and Arching show higher rates when boys question boys and girls question girls than when either boys or girls question a child of the opposite sex. Embedding, however, is more frequent for cross-sex pairs. We have been arguing that Arching and Chaining represent two different modes of relationship, that is, the former indicating more equality or at least competition for control, the latter indicating a difference in authority. When both go in the same direction, as in these and the findings on girl-adult vs. boy-adult series, some doubt is cast on this line of interpretation. An analysis of discourse length as related to these modes of connection will throw more light on this problem. Before proceeding with the analysis of discourse length, it will be useful to our understanding of the processes involved to have some concrete examples of these relatively abstract statistical trends. These examples were selected to represent the differences in types of discourse associated with different questioner-responder pairs; they do not constitute a random sample but may be thought of as exemplars, We have noted the prominence of Chaining in adult-child series~ The following excerpt is typical of such a pattern; the adult asks a series of seven successive questions, each serving to tie together the successive pairs of dialogue units:

I10

Mishler

Ad:

Carolyn tile words you wrote on the board were they words you didn't know (S) or words that you were happy you had learned? (St Yeah. Which one? The words I didn't know. OK. I'll show you a little later, alright? They're the words you taught me. What? The words you taught me. The words 1 taught you this morning? Yeah. 1 know, it, but everybody forgets. You know we hafta say em over again. We'll work on em. You tell me when you're ready after art. OK? We'll do it again. Alright. Will that be a good time?

GI: Ad:

In the following child-adult series, we find the adult responding to a question with a question, thus c o n n e c t i n g the first two units through Arching~ The adult then takes control o f the conversation through questions that serve to Chain the succeeding units together. This pattern underscores the p r o b l e m of interpreting the findings in Table II, which are organized in terms o f the initiating pair in a series. The relatively high rate o f Chaining in child-adult series does n o t signify a greater degree of power for the child b u t may, as in this example, reflect the power of the adult who regains control after the first exchange:

Studies in Dialogue and Discourse

111

Ad:

Everybody else finished their other papers for me (S) an 1 wish you would.

s l P e h whee ylite bin? did__


Ad: : ~ ld What little thing? The thing you cut out? No. What little thing? My gween little bug.

in here an watched it. Alan-where d'your papers usually go? In the paper box. OK.

The conversation b e t w e e n two girls in the n e x t excerpt includes both Chaining and Arching. In the c o n n e c t e d series o f five IUs, there are four successive questions, two from each o f the girls:

GI:

Debbie, do you think our house is ugly? No. Well I think it is. Huh? It is. We still hafta paint the rest of it cuz we didn't paint all of it. You know how much it's gonna cost? How much? All together? How much? Five million dollars.

G2: GI:

~
~ ~ G I :

G2: GI: G2:

~GI:

0ooooo.

I 12

Mishler

The following two excerpts are particularly instructive as a contrasting pair. The same boy is involved (BI), the conversations take place during the same period of observation, and the topic is much the same. In the first series, we find a mix of Chaining and Arching; b o t h children ask each other questions. The second series is tied together solely by Chaining, through the adult's questions. This comparison expresses quite clearly the different functions o f Arching and Chaining that we have been proposing: CI: BI: CI: Y'know what? What? Ya know what? If you got yer snack you could eat it. If you got hun,,ry. What do you ~_otfor a snack? These. Will you give me one? What will ya give me? A cracker.

BI:

~ ~

What kind2 C1 :
BI:

These. These. N'I don't-No, I don'l want one a - I Steve, a~e those good cookies? Um-hum (YES). Did you help make them? Noooo. (S) These is chips. (S) Did your mommy make them? Noooo. We bought 'era. Chips Ahoy cookies, Oh, sure. Chips Alloy cookies.

Ad: BI:
Ad:

BI:
Ad:

BI: Ad:

Relationships Between Length of Discourse and Modes of Connection Data are presented in Table III on length of discourse in relation to the identities of questioner and responder, It is evident that conversations which

Studies in Dialogue and Discourse

!t3

include adults, whether they initiate them by a question or are the first respondent to a child's question, are longer than conversations between children. About .twice as many of child-child conversations stop with one exchange-than do adult-child conversations; abdut half of all child-child conversations are only one dialogue unit in length. The pattern reverses for long conversations, that is, those that are at least three units in length; adult-child exchanges of this length are about twice as frequent as those between children. At this point, it is important to emphasize again that we are focusing on one specific type of discourse, namely, a connected exchange between speakers that is both initiated and sustained by questioning. To report, as we have above, that child-child conversations are shorter than those between adults and children refers only to the use of questioning to sustain conversations. A conversation may be tied together through its content, t h r o u ~ ritualistic devices, through nonverbal acts, and through other linguistic ways of establishing "cohesion" in a text. The findings presented above should be interpreted specifically as indicating that, in conversations between children and adults, questions are more likely t o be used as a way of constructing extended discourse than they are in conversations between children. The length of such conversations between adults and children is influenced to some degree by the sex of the child. Adult-initiated conversations with boys are longer than those with girls; those with boys show about 10% fewer one-unit series and about 10% more of the longer conversations with three or more units. This sex difference reverses when we turn to conversations with adults initiated by children. Those initiated by girls are longer than those initiated by boys; very few of the girl-adult series stop with one dialogue unit, whereas about 40% of the boy-adult series are this short. More than twice as many girl-adult than boy-adult series are three or more units in length. These long series constitute about two-thirds of all girl-adult conversations, and this is a higher rate than for any other combination of speakers. Within child-child series, similarity or dissimilarity in sex of speakers makes more of a difference than whether the questioners and/or responders are boys or girls. There are shorter conversations in dissimilar sex pairs than in similar sex pairs; this is particularly evident in the comparisons of longer conversations. The rate of short conversations, of one unit length, is lower in girl-girl series than in any other combination of child-child speakers. These findings lend support to the earlier suggestion that differentials in rates of Chaining, Arching, and Embedding may be associated with length of discourse. Clearly, modes of constructing discourse cannot be independent of length of discourse; some degree of relationship is built into the definition of

114

Mishler

i.

0.)

~.-o

..= .?
"0 "0 0 .'~

oJ

r',.. r,-

.__." A
J

~..,~

Studies in Dialogue and Discourse

115

connected discourse since it may be constructed and extended only through these particular modes of connection. However, the data presented in Table II, on differential rates of Chaining, Arching, and Embedding as related to speaker identities, use relatively crude measures that would allow for a fair amount of variation between length and number of connectors. More importantly, length may be extended through any of the three mechanisms, and differences associated with different questioner-responder pairs are neither implied nor required by the overall relationship between length and numbers of connecting questions. The longer discourse series found for adult-boy than for adult-girl exhanges, in Table III, would fit with the differentials in rates of Chaining between these pairs. However, these differences in length would be inconsistent with the higher rates of Embedding found for the adult-girl series and would be independent of Arching, on which these questioner-responder pairs are relatively equal. The differences found in rates of Chaining and Arching would be consistent with the longer series of girl-adult than boy-adult and of similar-sex than cross-sex pairs. It is important to disentangle the length of a series from the mode of construction used since one of the central issues in the study is the question of how different social relationshipships are realized in the structure of dialogue and discourse. The length of a connected conversation bears on these differences, but the ways in which a conversation is constructed seem to us to be a more sensitive and a more interesting topic for examination. Although the data presented to this point d o not permit a clear distinction between length and mode of connection, there is one finding in Table III that suggests a way of examining the relationship between speaker identities and mode of connection that is independent of length. The frequency of two-unit discourse series, that is, of paired IUs, is relatively equal for all primary questionerresponder pairs. About 28% of all discourse series are two units in length for adult-child, child-adult, and child-child pairs; more variation is present for child-child subpairs. A pair of dialogue units can be connected by one and only one type of connection. If the relative rates of Chaining, Arching, and Embedding are analyzed separately for two-unit and longer series, we would be able to determine more precisely whether there is a relationship between speaker identities and the use of these different modes independently of the length of discourse series. These findings are presented in Tables IV and V. The data in Table IV, on two-unit discourse series, are clear and straightforward. For adult-child series, Chaining is the primary way through which paired units are tied together; Arching is used rarely and Embedding at a moderate rate. For child-adult series, Arching is the most prominent mode. Among child-child

116

Mishler

Table IV. Rates of Chaining, Arching, \ and Embedding for Discourse Series Consisting of Two Dialogue Units (Freqtiencies and Percentages by QuestionerResponder Pair)
ii ] II i iii iii1~

Mode of connection Number of two-unit series 34 8 81 21 6 9 23


i

Chaining N 20 1 35 10 3 3 10
i

Arching At 2 6 25 7 1 3 9 (%) (6) (75) (31) (33) (17) (33) (39)

Embedding N 12 1 21 4 2 3 4 (%) (35) (12.5) (26) (19) (33) (33) (17)


i

Questionerresponder Adult-child Child-adult Child-child -Boy-boy -Boy-girl -Girt-boy -Girl-girl


i

(%) (59) (12.5) (43) (48) (50) (33) (44)

series, although Chaining has the highest frequency there is a relatively even distribution of all three modes; Arching is considerably more frequent than in adult-child series. Among subt}pes within the eNid-child series, differences appear to be relatively slight. These findings tend support to our general theoretical view o f questioning as a form through which differentials in authority and control are expressed. The relative power of adults vis.a-vis children is realized in the Table V. Rates of Chaining, Arching, and Embedding for Discourse Series of Three or More Dialogue Units (Frequencies and Percentages by Questioner~ Responder Pair)
i

Questionerresponder Adult-child Child-adult Child-child -Boy-boy -Boy-girt -Girl-boy -Girl-girl

Number of three+ series 53 17 53 20 6 2 18


i

Number with Chained pairs N 53 16 44 17 3 2 15 (%) (100) (94.1) (83.0) (85.0) (50~ (lO0.) (83~ '

Number with Arched pairs N 24 15 37 14 4 2 13


i

Number with Embedded pairs N 24 4 22 6 4 1 8 (%) (45.3) (23.5) (41.5) (30.0) (67.0) (50o0) (44.4)

(%) (45.3) (88~2) (69~ (70.0) (67.0) (100.) (72.2)

Studies in Dialogue and Discourse

!i7

disproportionate use of Chaining in adult-initiated discourse. Through this pattern of successive questioning, asking a "next" question when there has been a response to a first question, the adult who initiated the conversation maintains control over its continuation. The relatively high rate of Embedding in adult-child series may be viewed as another index of differential power since it reflects responses from two different children to one question from an adult; since these tabulations are restricted to paired dialogue units, neither of these responses is itself a question. Further support for the "control" function of questioning is found in the disproport!onately h i ~ rate of Arching in child-adult series; two-thirds of all children's questions to adults are responded to with a question which then begins a dialogue unit that terminates the conversation. Within child-child series, the differences among types of connection are not significantly different from each other and may therefore be viewed as exhibiting a relatively balanced use rather than a strong preference for any one of them. Rates of Chaining, Arching, and Embedding for discourse series containing three or more IUs are presented in Table V. Since combinations are possible in longer series and one mode does not exclude another as in the two-unit series, the measure used is the presence or absence of each type of connection within a series. Nevertheless, despite the less precise meaning of these rates, the findings are fully consistent with the findings on two-uiait series. Without exception, that is, in all series of this length, adult-initiated discourse includes Chaining as one of the modes of connection. This drops to 83% for child-child discourse and stays at this level for each of the subtypes of same- and cross-sex pairs. With regard to Arching, the pattern reverses as it did in two-unit series-the 70% rate of Arching for child-child discourse is considerably higher than the 45% rate for adult-child series. Although the sample size is quite small, the high rate of Arching in child-adult series is consistent with that found in the two-unit analyses. In sum, by disentangling the length of a discourse series from the ways in which the series is constructed, we have demonstrated that there are differences between questioner-responder pairs in how questions are used to sustain a conversation. The evidence provides strong support for the guiding hypothesis of the study, namely, that differentials in authority between speakers are "realized" linguistically through differences in how questioning functions to control a conversation. Adults "Chain" dialogue units together by successive questions afier children's responses; they also recapture control of a conversation from a child if the latter should initiate an exchange with a question. Children question each other in a more "balanced" way; this may be a display of either peer egalitarianism or competition for control.

118 DISCUSSION

Mishler

This paper is both an exploration of a particular method for studying the social functions of language and a presentation of findings. The emphasis throughout has been descriptive and empirical with only passing mention of some underlying concepts and theoretical concerns9 One reason for this emphasis is that the assumptions on which the method relies and which it explicates are widely current if not universal among sociolinguists (e.g., see such recent texts and collections as Gumperz and Hymes, 1972; Robinson, 1972; Sudnow, 1972). A second and related reason is that the sociolinguistic literature is still weighted toward programmatic statements; it is rich in hypotheses and assertions but relatively sparse in systematic methods and detailed findings that bear on these hypotheses. At this point, however, it may be useful to relate our approach and study to other current work. Our principal assumption, and it is perhaps basic to and distinctive of the field of sociolinguistics, is that specific features of language "carry" social meaning. Alternative current versions of this statement are that social reality is linguistic~illy encoded or that social functions are realized in the structure of language. Essentially, this asserts that "how" one talks reveals "who" one is, at least relative to intended hearers, and is the way through which social rules, norms, and codes are both expressed and maintained. The critical problem is that there is no isomorphism between linguistic structure and social function-it is evident that the "same" function may be carried by different features and the same feature may serve different functions. Therefore, the general assumption is always hedged with qualifications that refer usually to the context. That is, what one means depends on the context or situation within which one speaks~ There are several approaches to this problem. Hymes (1971) proposes the notion of "appropriateness" and argues that linguistic competence in the broad sociocultural sense requires an understanding on the part of the speaker of when it is appropriate to say what to whom and in what manner. This implies another level of competence or knowledge as to the na'ture of the situation so that the "appropriate" linguistic program can be selected from the 9range of alternative possibilities in each speaker's repertoire. An alternative view, associated with the ethnomethodological approach, argues that the criteria for judgment With regard to the situation and the rules of appropriate talk are not only vague but also do not function as a programmed guide to speaking. Rather, a speaker talks in ways that are roughly consistent with a few general rules which allow for correction and reformulation of his actual speech in accord with the specific ways in which the conversation develops. Some general rules for conversation are available to the speaker-for opening

Studies in Dialogue and Discourse

1t9

and closing conversations, for turn-taking, for story-telling, for punning-but the speaker is more of a "rule- or norm-user," as Churchill suggests (1970), than someone who is guided by a set of programmatic rules. (A sampling of ethnomethodological studies of language ma~/ be found in the papers of GarfinkeI, Sacks, and their students collected in Sudnow, 1972, and Turner, 1974.) In research on language, as in any other inquiry, we must begin somewhere. That is, we must take some features of the world as "given" for the purposes at hand. Either perspective on the rules that regulate how social meanings are realized in and through language, whether as guides or "for use." requires an examination of actual speaker practices. We nmst still choose particular problems for study, select particular ways to observe and collect samples of relevant speech, and define particular units and relationships for analysis. We began with some intuitive notions, such as the importance of collecting spontaneous, that is, "natural" or nonelicited, speech. We wanted a relatively standard setting that at the same time offered the possibility of contrast and variation among speakers-thus first-grade classrooms and portable tape recorders rather than the laboratory and testing procedures. In addition, and most importantly, we considered the classic and primary language unit of the linguist, namely, the sentence, as too -restricted for exploring the social functions of language use-hence the more inclusive units of utterance, dialogue, and discourse. (The limitations of the sentence for traditional linguistic inquiry have recently received attention; e.g., see Sherzer's review, 1973.) Associated with this was a view of the structure of language as organized into a hierarchy of levels such that larger units like discourse were composed of units at the next level down, that is, dialogue, and so on through the descending levels of utterance, sentence, phrase, and word. Further, the social functions of language are assumed to be carried at each level, although through different ways. (Our approach to language structure and function comes closest to the linguistic theory developed by Halliday; for a recent summary, see Halliday, 1973.) For example, we hypothesized that power and authority should be expressed, that is, realized or encoded, not only at the lexical level, as Brown and Gilman found (1960), but also in the syntax of sentences, in the forms of questioning, and in the ways that discourse is constructed. Our methods of collecting samples of talk, of segmenting the corpus into conversational episodes, of selecting questioninitiated exchanges, and of urtitizing and connecting units of dialogue and stretches of discourse-all of these procedures were intended to permit an analysis of language through which we could explore these intuitive and orienting assumptions~ Questioning was selected as a place to begin, in part because there are relatively standard and conventional criteria for locating questions in a text,

120

Mishler

although still many problematic issues, and in part because it is a topic that has received some attention from other investigators (e.g., Brown, 1968; Churchill, 1970; Holzman, 1972; Robinson and Rackstraw, 1972; Uttan et al., 1974). In addition, questioning appeared to represent one way through which social relationships along a dimension of authority and power are realized in language and this would allow fairly direct access to the basic problem of relating structure and function. Finally, the act of questioning implies a response and thus contains within it a unit of language that we believe is particularly appropriate for sociol!nguistic study, namely, the dialogue between speakers. This paper and the earlier report of the study focus on findings at the level of dialogue and discourse. A number of these findings serve to support the utility of the approach. For example, we have found that the length of question-initiated and question-sustained discourse is regulated in such a way that it may be described by a formal rule; that is, it follows an exponential function (Mishler, 1975). Data presented in this paper demonstrate that different ways of constructing discourse are used differentially by speakers with different role relationships to each other. Further, these relationships are consistent with the presumed control functions of particular linguistic strategies and with known authority differentials between speakers. Discourse initiated by adults has a different structure than that initiated by children, and the particular structure depends as well on the identity of the respondent. In brief, there has been consistent support in the findings for our genera/ hypothesis that different ways of extending a conversation through questioning express different authority relationships-adults Chain a conversation together through successive questioning and regain control of a conversation by responding to a child's question with a question, that is, by Arching. Among children, a pattern of relative equality or of competition for control is expressed through a more balanced use of different modes of questioning. Analyses now under way focus on other levels of language, on utterances and the forms of questions, responses, and confirmations. These findings will be presented in other papers in this series.

REFERENCES
Bellack, A. A., Kliebard, H. M., Hyman, R. To, and Smith, F. L., Jr. (1966). The Language of the Classroom, Teachers College Press, New York. Brown, R. (1968). The development of Wh questions in child speech. J. Verbal Learn. Behav. 7: 279-290. Brown, R., and Gilman, A. (1960). The pronouns of power and solidarity. In Sebeok, To A., (ed.)0 St vte in Language, MtT Press, Cambridge, Mass. Churchill, Lo (1970). Questioning as a social activity. Paper presented to meetings of the American Sociological Association, September 1970, mimeo, 1I pp.

Studies in Dialogue and Discourse

1.21

Gumperz, J. J., and Hymes, D. (eds.) (1972). Directions in Sociolinguistics: The Ethnography of Communication, Holt, Rinehart and Winston, New York. Halliday, M. A. K. (1973). Explorations in the Functions of Language, Edward Arnold, London. Holzman, M. (1972). The use of interrogative forms iri the verbal interaction of three mothers and their children../". Psycholing. Res. 1: 311-336. Hymes. D. (1971). Competence and performance in linguistic theory. In Huxley, R.+ and Ingram, E. (eds.), Language Acquisition: Model$ and Methods, Academic Press, New York. Mishler, E. G. (1975). Studies in dialogue and discourse: An exponential law of successive questioning. Language in Society 4: 31-52. -: Robinson, W. P. (t972). Language and Social Behavior, Penguin, Middlesex, England. Robinson, W. P., and Rackstraw, S. (1972). A Question of Answers, Vols. I and II, Routledge and Kegan Paul, London. Sacks, H. (1972). On the analysability of stories by children. In Gumperz, J. J., and Hymes, D. (eds.), Directions h7 Sociolhzguistics, Holt, Rinehart and Winston, New York, pp. 329-45. Sherzer, J. (1973). On linguistic semantics and linguistic subdisciplines: A review article. Language in Society 2: 269-289. Sudnow, D. (ed.) (1972). Studies in Social hTteraction, Free Press, New York. Turner, R. (ed.) (1974). Ethnomethodology, Penguin, Middlesex, England. Ultan, R., Moravcsik, E., and Baron, N. (eds.) (1974). Questions: Readings for Scientists of Language, in press.

You might also like