Vous êtes sur la page 1sur 44

WHATS THE USE OF TRIADIC DIALOGUE?

:
AN INVESTIGATION OF TEACHER-STUDENT INTERACTION

Hossein Nassaji and Gordon Wells


Ontario Institute for Studies in Education
University of Toronto

Abstract1

The fact that the spoken texts of classroom interaction - particularly those involving teacher with
whole class - are co-constructed relatively smoothly, despite the number of participants involved,
suggests that they are organized in terms of standard strategies, embodied in typical forms of
discourse that have evolved for responding to recurring types of rhetorical situation (Miller, 1984;
Kamberelis, 1995). That is to say that, like written texts, they can be thought of as being constructed
according to one of a set of educational genre specifications. One such rhetorical structure, the
ubiquitous triadic dialogue (Lemke, 1990) (also known as the IRE or IRF sequence (Mehan,
1979; Sinclair & Coulthard, 1975), has attracted considerable attention in recent years, and has
variously been seen as, on the one hand, essential for the co-construction of cultural knowledge
(Heap, 1985; Newman et al., 1989) and, on the other, as antithetical to the educational goal of
encouraging students' intellectual-discursive initiative and creativity (Lemke, 1990; Wood, 1992).
Drawing on episodes of teacher-whole-class interaction collected during a collaborative action
research project, this paper will show, however, that the same basic IRF structure can take a variety
of forms and be recruited by teachers for a wide variety of functions, depending on the goal of the
activity that the discourse serves to mediate and, in particular, on the use that is made of the followup move.
________________________________________

In his seminal writings on the dialogic nature of utterance, Bakhtin (1986) pointed out that all
utterances both respond to what has preceded and anticipate a further response. While this is true,
it is also the case that, in speech, many utterances tend to be more oriented either to what preceded
or to what will follow, as is the case with the relationship that holds between the members of an
adjacency pair, such as question-answer (Sacks, Schegloff, & Jefferson, 1974). On the other
hand, not all conversational exchanges are limited to two moves, and many are much longer.
There are two main reasons for this.

First, as Halliday (1984) has argued, there are two basic exchange-types: a) Demand - Give-inresponse and b) Give (unsolicited) - Accept. However, a third equally basic type is frequently
created by the combination of the first two: c) Demand - Give-in-Response - Accept. Where
information is the commodity exchanged, this gives rise to the three-move exchange structure:
Question - Answer - Acknowledgement (Halliday, 1984). This exchange structure occurs quite
frequently in everyday conversation, as for example:

A: Which way should I go to get to the station?

D: Information

B: Take the first street on the left and then the second right.

G: Information

A: Thanks very much.

A: Thanks

Although this structure certainly occurs in classrooms, it is not nearly so prevalent as a rather
different three-move exchange type, in which the third move does not acknowledge the given
information, but rather makes a substantive reaction to it. The following is a stereotype example:

T: Which way did the Wolf go to Red Riding Hoods


Grannys cottage?

D: Information

S: He took a short cut through the forest.

G: Information

T: Thats right.

A: Evaluate

In this structure, however, only one participant typically initiates the exchange - the teacher; and the
teacher always has the right to provide the third move, often, as above, by evaluating the students
contribution for its conformity to what he or she considers to be a correct or acceptable response.

These differential rights to moves in the exchange have often been discussed in terms of the power
differential between teachers and students (e.g. Lemke, 1990), and there is no doubt that, if this is
not the primary reason for the participants unequal behavior, its perpetuation is certainly a likely
consequence. However, an alternative explanation has been proposed by Berry (1981) that does
not appeal to power, as such, and that is applicable in other settings than classrooms. In explaining
the different types of three-move exchanges, Berry makes a critical distinction between the
primary and the secondary knower and, on that basis, is able to provide a discursively
principled explanation of the difference between the two examples above.

In both these examples, two important discourse roles are involved: the initiator of the exchange and
the primary knower with respect to the information at issue. When the two roles do not coincide, as
in the first example (where a stranger seeks information from a supposed well-informed local), the
primary knowers critical contribution is made in the second, responding move, and the
questioners acceptance in the third move adds nothing to the information that is being exchanged.
In the second, classroom example, by contrast, the teacher is both the initiator and the primary
knower and, as a consequence, it is only when she confirms or disconfirms the students response
that the exchange of information can be treated as complete. To establish this argument, Berry
contrasts two exchanges from a hypothetical tv quiz show. In the first, the quizmaster confirms the
correctness of the contestants answer and, in the second, merely says oh in the third move. As
she correctly points out, in this case both the contestant and the studio audience would be fully
justified in objecting to the quizmasters improper realization of the third move.

Berrys argument also explains the second reason for a continuation of many exchanges beyond
two moves. If, for whatever reason, the exchange is not completed to the participants satisfaction,
they can use a variety of bound exchanges to rectify the problem. For this reason, Wells (1996)
argues for recognizing a larger unit, the sequence, as the basic unit of conversation. A sequence,
according to his definition, consists of a nuclear exchange and as many bound exchanges as are
judged necessary by the participants to complete what was initiated in the nuclear exchange. Bound
exchanges of three kinds regularly occur. Preparatory exchanges are used to establish
communication or to select a designated speaker; embedded exchanges are used to confirm
uptake or to repair various types of breakdown (e.g. clarification); while dependent exchanges
are used, for example, to give or seek additional information (comment) or justification for the
information already supplied (justification). In principle, in casual conversation either participant
can initiate a bound exchange at any point and, as a result, sequences can extend over many
exchanges (Eggins & Slade, 1997).

However, in the classroom, the dominant mode of interaction is not casual conversation, since
most talk between teacher and students has a pedagogical purpose. In teacher-whole-class
interaction, in particular, it is almost always teachers who initiate sequences; they also tend to initiate
most bound exchanges (although students do occasionally ask for clarification) (Cazden, 1988).
This, we suspect, is because, in the classroom, in addition to the role of primary knower, there is
another important role to be filled, which is independent of the particular information under
discussion. In a group of thirty or more people, it is necessary for somebody to ensure that the
discussion proceeds in an orderly manner and that, as far as possible, all participants contribute to,
and benefit from, the co-construction of knowledge that is the purpose of the discourse (Dewey,
1938; Peters, 1966; Rogoff, 1994). Whatever this role is called - manager or facilitator - it is
part of the teachers responsibility to ensure that it is enacted, and most frequently teachers take it
upon themselves to do so. Whether or not they are the primary knowers, then, they almost always
assume the role of manager and it is often in this role that they ask for clarifications and

justifications. In the same role, they may also extend sequences by offering meta-comments of
various kinds on the quality or organization of the discourse. It is also noticeable that, even when
students initiate a sequence, the teacher very often provides a response that, in function, is similar to
the third, follow-up move of the three move exchange or 'triadic dialogue' (Lemke, 1990).2

For example, in the following sequence taken from a grade 4 class discussion following the reading
of a chapter from Mrs Frisby and the Rats of Nimh (DZ2; see table 1 for details), it is a student
who asks the initiating question and another student who answers, assuming the role of primary
knower on this issue; but it is the teacher who, in a follow-up move, comments positively on the way
the students have been carrying on the discussion:3

Student 4: How d'you know they'd be SMART ENOUGH to bring back the wire?
Student 2: The smart ones could have taught them ...
Teacher:

Mmm .. I really like the way that you're talking back and forth to each other .. that's
great! In a discussion it's best if you can try to respond to what the person who just
talked said . I know it's not always possible but it's great when you're going back and
forth

In much of the discussion of triadic dialogue, it has been assumed that the prototypical function of the
follow-up move is to evaluate the student response that immediately precedes. Indeed in Mehans
(1979) study, this three-part structure was labeled Initiate-Respond-Evaluate (IRE) and the label has
been perpetuated in most subsequent North American research on classroom discourse (e.g. Cazden,
1988). However, as the preceding discussion of three-move exchanges, and of ways in which they can
be extended, makes clear, there is a much wider range of options available to teachers in the third
move. For this reason, Sinclair and Coulthard (1975) dubbed the third move Feedback, subsequently
changed to Follow-Up, and proposed three categories of act that can occur in this slot. These are:

accept (including reject), evaluate, and comment, with the latter category expanded to include the
more delicate sub-categories of exemplify, expand and justify.

In a fourth grade unit on structures, the teacher was introducing the topic of bridges and the various
structural forms and methods used in their construction (GD1). She was making use of a book on the
subject and had just read a paragraph about the use of reinforced concrete, which led to mention of
the Oakland Bridge.
Move

Function

K1

Init

Question

K2

Resp

Answer

Dep

K1

Init

Question

Emb

K2

Init

Req.Conf.

Emb

K1

ReInit

Prompt

K1

F-up

Comment

K1

Init

Question

8 S2: Um- beaches

K2

Resp

Answer

9 T:

K1

F-up

Accept

K1

Init

Question

K2

Resp

Answer

1 T:

Where's Oakland?

Exch.

Status as

Type

Knower

Nuc

2 S1: Oakland? It's in San Francisco, I think

3 T:

What um state?

4 S1: What state?


(no further response)

5 T:

Anybody know? ....


(no response)

6 T:

It's in California

7 T:

What do they have a lot of in California?

10 T:

Dep

Well, they do have that!

But something that would affect bridges

11 S3: Sun?

Dep

7
12 T:

That doesn't affect the bridge

13 S4: Tornadoes?
14 T:

&Dep4

No

15 S5: Earthquakes
16 T: Right, Brian .. bridges have some

&Dep

K1

F-up

(Reject)/ Comment

K2

Resp

Answer

K1

F-up

Reject

K2

Resp

Answer

K1

F-up

Evaluate

problems with earth tremors and earthquakes

Comment: Expand

Key. K1 = Primary Knower; K2 = Secondary Knower; Nuc = Nuclear; Dep = Dependent; Emb = Embedded;
Init = Initiate; Resp = Response; F-up = Follow-up

In the above example5, where the teacher is the primary knower, we see her providing a variety of
follow-up moves. Moves 9,14, and 16 all involve a form of evaluation; in the first two the teacher
accepts or rejects a student answer, and in 16 she provides an explicit evaluation. In 6 and 16 she
gives a comment in which she supplies or expands the answer and in 12 she makes a comment that
implicitly rejects the preceding answer and simultaneously justifies that rejection. However, we can
also see the occurrence of another option. Instead of negatively evaluating a student response or
providing the required information in a comment, the teacher can equally ask a further question to the
previous speaker, or any other student, in order to obtain a more adequate answer (move10). When
this is proffered, the teacher once again has the same range of options available for making a further
follow-up move. In this way, the initiating question of a nuclear exchange can give rise to a number
of dependent exchanges that ultimately lead to a satisfactory completion of the sequence. Mehan
referred to such extended sequences as topically related sets (1979, p.65).

However, there is a still further possibility. As we saw earlier, where the teacher has the role of
primary knower, as in the preceding example, there is a strong expectation that he or she will give the

stamp of approval to the contribution that the secondary knower makes in the second move. On the
other hand, if it is the student who is the primary knower, or if no participant lays claim to this role,
there is no requirement for the teacher to perform an evaluating function in the third move. Instead,
she or he can add a comment that extends the discussion or ask a question that invites a student to do
so. The effect of adopting this latter strategy, in particular, is to cast the responder in the role of
primary knower and thereby to create a more equal mode of participation. At the same time, by
posing questions that solicit the students opinions and conjectures, the teacher can encourage a more
dialogic and exploratory stance to the topic under consideration (Mercer, 1995; Nystrand, 1997;
Rogoff, 1994; Wegerif & Mercer, 1997).

These, then, are some of the possible ways in which teachers can use the follow-up move in triadic
dialogue in order to achieve a variety of different pedagogical purposes. In the study to be reported
here we first describe the frequency with which these possibilities are taken up in a corpus of
episodes of teacher-whole-class interaction and then discuss their effects on student participation.

Background to the Study

The source of the data for this study was a collaborative action research project conducted over the
years 1991 - 1997, involving nine elementary and middle school teachers and three university
researchers in Toronto and its satellite towns. In all, it involved classrooms in four school districts
and one independent school. The project had two aims: first, through action research, to explore ways
of adopting an inquiry approach to learning and teaching; and second, to investigate the role of
spoken and written discourse in the activities that took place in the participating teachers classrooms.
In the first phase of the project (1991-1994), the focus was on inquiry in science and only lessons in
science were observed. In the second phase (1994-1997), the study was extended to all areas of the
curriculum and written as well as spoken discourse was investigated. In the first year (1991-1992),
only three teachers were involved; they were all in one school, which was selected on the

recommendation of a school district official. Two of these teachers withdrew at the end of the year.
In subsequent years, new members who joined did so entirely of their own volition. In phase two, all
the participating teachers were self-selected volunteers.

As already mentioned, the action research component of the study was set up to explore ways of
enabling students to take a more active role in negotiating the curricular topics to be studied and the
means used in investigating them. Each teacher selected an issue related to these objectives that he or
she wished to investigate (e.g. How can the childrens questions be made more central to the way in
which a science topic is taught?, How can written dialogue between students be given a central role
in the study of history?) and one of the university researchers provided support by making
videorecorded observations of relevant events and discussing them with the teacher concerned. Once
a month, the whole group met together at the university after school, where members took it in turns
to present their inquiries.

At the same time, both in these meetings and via email, the group discussed readings from
sociocultural theory (e.g. Moll, 1990; Vygotsky, 1978; Wertsch, 1991) and began to try to formulate
the principles underlying what we have referred to as an inquiry approach to curriculum (Wells (Ed.)
in press). In this way, over the six years of the project, a theoretical framework was developed that
allowed systematic comparisons to be made between the individual teachers inquiries and provided a
common basis for individual and group presentations of our work at conferences and in publications
(e.g. Donoahue, 1998; Hume, 1998; Shechter, 1998; Galbraith et al., 1997; Wells, 1999).

This collaborative action research can be seen as a first level within which the present study of
triadic dialogue was nested at a second level. Since each of the teachers' inquiries generated videotaped observations of classroom events, almost all of them involving group and whole-class
interaction, we had amassed a considerable corpus of classroom discourse by the time the action
research project reached the end of the period of funding. Quite a number of these observations

10

had already been transcribed as part of the teachers inquiries and it is this corpus of transcribed
observations (well over 130 hours) that provided the data to be analyzed in the present investigation.

Throughout the action research project, both teachers and university participants had noted informally
that the episodes of whole class discussion seemed to be critical to the development of an ethos of
inquiry; this was further confirmed when the second author investigated a number of episodes in
some detail (Wells, 1993a, 1996). From this work, it appeared that a) a very large proportion of
teacher-whole-class discussions made substantial use of triadic dialogue', and b) it was the choice of
follow-up move that largely determined how the discourse developed. On this basis, we decided to
carry out a systematic descriptive investigation of the nature of triadic dialogue in our corpus and, in
particular, of the use made of follow-up moves. The question to be investigated was: What are the
various forms and functions of triadic dialogue in teacher-whole-class episodes of interaction in the
corpus as a whole and how does the choice of follow-up move affect the nature of the students'
participation?

Development of the Scheme of Analysis

In parallel with the second phase of the action research project, the university researchers worked on
the development of a coding scheme with which to analyze the discourse data from the classroom
events that were being recorded. In keeping with the sociocultural orientation of the project as a
whole, the scheme sought to integrate the approach to discourse analysis developed within systemic
functional linguistics (Eggins and Slade, 1997; Halliday, 1984; Lemke, 1990; Martin, 1992) with
work in activity theory (Engestrm, 1990, 1991; Leont'ev, 1981). The key to our approach was the
recognition that spoken discourse always occurs as mediator of some purpose within a larger
structure of joint activity; in activity theoretic terms, discourse is one of the operational means selected
to achieve the goal of the current activity. Sometimes, as in practical work, the discourse is ancillary
to non-verbal behavior, which is the primary operation; in other contexts, such as whole-class

11

discussion, the discourse is constitutive of the activity while non-verbal behavior plays an ancillary
role (Martin, 1992). In either situation, however, the discourse can only be fully interpreted in relation
to the purpose of the activity as a whole (see Wells, 1993a, 1996, for a fuller exposition).

One of the chief advantages of this approach is that it provides a principled basis on which to segment
the stream of speech into units for analysis. The largest unit is an Episode, which consists of all the
talk produced in carrying out a single activity or one of its constituent tasks (e.g. deciding what to do
with an injured chrysalis (AJ9, grade 6), or creating a web from the childrens responsive comments
noted by parents in a previous shared story reading (VM2, grades 1 and 2)). Episodes are made up of
Sequences that, individually and cumulatively, contribute to the achievement of the activity or task
goal. Each sequence consists of an obligatory Nuclear exchange and any Bound exchanges,
(Preparatory, Embedded or Dependent) associated with it.6 Finally, whether nuclear or bound, each
exchange consists of obligatory Initiating and Responding moves and may also contain a Follow-up
move. (This rank-scale of constituent elements can be shown diagrammatically, as in Appendix.)

It has to be recognized, however, that boundaries are not always clear-cut, even for the participants. In
some cases, more than one exchange may be proceeding in parallel, for example when different
students respond at the same time; in others, a move by one speaker may be interpreted differently by
those who follow, resulting in different, and competing, sequential implications. Not surprisingly,
these problems are greater when the teacher does not control the discussion. In addition to
considering the speakers presumed intention in coding each move, therefore, it is also necessary to
take account of how the move is taken up in the ensuing discourse (Sinclair & Coulthard, 1975).

In our analysis, the sequence is treated as the focal unit. Each sequence within an episode is coded
for the activity/task in which it occurs and for the role it plays as an operation in advancing the activity
goal. Such Episode Activity Orientations (EAOs) range from directing or planning action, through
problem solving, reporting previous activity, and monitoring what has been learned, to constructing

12

interpretations of experience and information, and reflecting on the processes and outcomes of
activities completed or in progress (see Appendix, Category 1, for a complete list). The internal
organization of sequences is next analyzed in terms of constituent exchanges (Appendix, Cat. 2), then
each of these is coded for the constituent moves and for the speaker of each move. Moves are also
coded for their degree of Prospectiveness (i.e. the extent to which they determine the following
move(s): Demand> Give> Acknowledge:7 see discussion in opening section; also Appendix, Cat. 3)
and for their Function with respect to the commodity that is exchanged (goods-and-services or
information (Halliday, 1984); also Appendix, Cat. 4). Finally, moves are also coded for their length
and complexity: less than a complete clause (e.g. a polarity response or a single noun phrase) (1);
main clause with or without one dependent clause (2); three or more ranking clauses (3). From these
codings, the Mean Response Length/Complexity (MRL) of each episode was determined by
calculating the mean score assigned to all student responses.

It is an axiom of systemic functional linguistics that coding schemes need to be developed in relation
to the particular questions addressed in the research. Thus, although the scheme just described can, in
principle, be tailored to address a wide variety of questions, the version described here was designed
specifically to address teachers use of follow-up moves in triadic dialogue and the consequences of
their choices for the extent of student participation. Accordingly, particular attention was given to the
coding of follow-up moves and to the contexts in which they occurred.

Follow-up moves were first coded for their degree of prospectiveness and for their function. Six
categories of function were recognized: Evaluation, Justification, Comment, Clarification, Action, and
Metatalk, and further subcategorizations were made within them (see Appendix, Cat. 5). These
function categories were only coded if the follow-up move was of a greater degree of prospectiveness
than a simple Acknowledgement of the preceding move. Give moves were always coded for function;
however, they were treated as alternatives to acknowledging moves with respect to their role in
sequential organization (i.e. as simply a follow-up move). Demanding follow-up moves (Appendix,

13

Cat. 6), on the other hand, were always treated as initiating a bound exchange, either embedded or
dependent, depending on the function in question. By coding in this way, it was possible to
distinguish between sequence-initiating questions and questions that followed up on the nuclear
exchange; it also enabled us to calculate the Mean Sequence Length for each episode, by dividing the
total number of exchanges by the number of sequences in which they occurred as constituents.

Initiating moves were also coded in some detail. First a distinction was made with respect to
prospectiveness: whether the initiating move was a Demand or a Give. When the commodity
exchanged was information, further sub-categorizations were made. Starting from Berrys
(1981) distinction between primary and secondary knower, and Labovs somewhat similar
distinction between A and B events (Labov & Fanshel, 1977), we distinguished three main
categories of information (Appendix, Cat. 4(b)-(d)): Assumed Known Information (where one
party, almost always the teacher, already knows the answer and is concerned to discover whether
students can supply it, e.g. "Who was the king of France? Let's see who remembers this"
(KM1)); Personal Information (where the information is known only to the person addressed, e.g.
"What did other people think when they were watching that experiment? Did it surprise you the
way that the water mixed or didn't mix?" (DZ8)); and Negotiatory Information (where the
answer is to be reached through open-ended discussion between teacher and students together,
e.g. "Neil has said that there are not enough troops What are you saying in response to that?
(HK2); "Do you agree with Nir? Give us a reason" (AJ8)). In this latter category, no-one is
considered to be the primary knower and the information refers to neither an A nor a B event; thus
all participants contributions are assumed - at least in principle - to be of equal significance in
working through exploratory talk towards a consensual conclusion (cf. Barnes, 1976; Wegerif
and Mercer, 1997).

14

We were particularly interested in the relationship between initiating moves that demanded
information of these different kinds and the types of follow-up move that were selected in each case.
Where the demand concerned information that was assumed to be known (henceforth referred to as
Known Information), it could reasonably be expected that the teacher would adopt the role of primary
knower and would therefore be likely to provide an evaluation in the follow-up move. Where the
information was personal, on the other hand, it would be the responder who was the primary
knower and so an evaluation would be out of place; if a substantive follow-up did occur, it would be
more likely to take the form of a comment or a request for further information from the giver of the
information.

However, it was cases where the initiating demand requested information for negotiation that
interested us the most. If such questions genuinely asked for students opinions, explanations and
conjectures, would teachers still evaluate the students responses, or would they select options that
offered their own opinions on an equal footing, or options that invited further student contributions?8
We anticipated that, if the teachers were creating communities of inquiry in their classrooms, as was
their avowed intention, this would be manifested, in part, by a more equal, dialogic mode of discourse,
with a greater proportion of questions inviting information for further exploration and negotiation
and, in the follow-up move, a tendency to choose the evaluative option less frequently and options
soliciting further student contributions more frequently.9

The follow-up move frequently contained several constituents, however. For example, a teacher might
first acknowledge or evaluate the student response, then give one or more comments, and finally
demand either a confirmation or a further substantive response, thereby initiating an embedded or
dependent exchange, respectively. When calculating frequencies in order to explore the relationship
between type of initiating question and teacher choice of follow-up, as in the present study, this poses
some problems, since there may be more than one follow-up for a given initiating question. For the
analyses reported below, we chose to tally the evaluative response, if one occurred, and to ignore other

15

forms of follow-up; if there was no evaluation but both a give and a demand, the demand was tallied
as, being higher on the scale of prospectiveness, it was this move that determined how the sequence
proceeded. A Give follow-up was only tallied, therefore, if it occurred alone.

Collection and Analysis of the Data

As already explained, the corpus of discourse data for this investigation consisted of all the episodes
recorded in the preceding action research project that had already been transcribed. Of these, 44
involved teacher-whole-class interaction. All nine teacher members of the project contributed, but
over differing time spans, as not all of them participated throughout the full six years of the project.
However, each of the six years as well as each of the nine teachers was represented in the database
that was analyzed, with no teacher contributing less than two episodes. It is also of interest that the
database included episodes from grades 1 through 8 and concerned predominantly curricular
activities in science and literature, although there was also a small number of episodes from history.
The distribution of the 44 episodes is shown in table 1, with the length of each episode in number of
sequences shown in parentheses. The same table shows the Episode Activity Orientation of each
episode.

Coding was carried out by a team of graduate students, who were trained by the principal investigator.
Codes were entered directly into a custom designed program, using FileMaker Pro 3.0. When all the
data had been entered, the coding was checked by the principal investigator and quantitative analyses
were carried out by both authors.10

16

Table 1. Provenance of Episodes Analyzed


Science
Practical
Grade 1/2
Grade 2

VM1 Prob-Solv (14)

Grade 3
Grade 3/4

Science
Discussion

Literature
Discussion
VM 2 Launch (42)

History
Discussion

VM 2 Prob-Solv (38)
VM 3 Review (62)
VM 1 Launch (14)
WG 1 Construct (40)
WG 2 Construct (23)
GD 1 Launch (25)
GD 2 Construct (30)

Grade 4

Grade 4/5

WG 3 Report (5)

Grade 6

AJ 1 Plan (13)
AJ 8 Generate (36)

WG 4 Construct (18)
DZ 6 Generate (50)
DZ 8 Report (16)
DZ 10 Organize (14)
DZ 12 Construct (60)
AJ 2 Construct (16)
AJ 4 Monitor (9)
AJ 6 Construct (28)
BJ 1 Prob-Solv (6)
BJ 3 Construct (10)
BJ 5 Monitor (5)

DZ 1 Generate (18)
DZ 2 Construct (30)
DZ 3 Construct (40)
DZ 4 Generate (21)
DZ 5 Review (28)
WG 5 Construct (7)
DZ 7 Plan (4)
DZ 9 Generate (26)
DZ 11 Plan (18)
AJ 3 Review (8)
AJ 5 Generate (25)
AJ 7 Construct (22)
BJ 2 Generate (11)
BJ 4 Generate (20)
BJ 6 Plan (30)

Grade 6/7
Grade 7

Grade 8

KM 1 Launch (41)
KM 2 Review (52)
NS 1 Monitor (12)
HK 1 Monitor (74)

NS 2 Construct (11)
HK 2 Formulate (8)

Results

In this section, we present the results of analyses carried out to answer the question: What are the
various forms and functions of triadic dialogue in teacher-whole-class episodes of interaction in the
corpus as a whole and how does the choice of follow-up move affect the nature of the students'
participation? Subsequently, in the light of a preliminary analysis, this question was refined to
consider episodes in relation to the contexts in which they occurred.

17

As a first step, a data matrix was constructed in which each episode was tallied with respect to
Episode Activity Orientation, number of constituent sequences, Frequency of Student Initiations,
frequency of Teacher Sequence Initiating Questions, whether demanding Known Information,
Information for Negotiation, or Personal, and frequency of Evaluative Follow-up moves in relation
to the type of initiating question.11 Because of the low frequencies of occurrence of many of the
subcategories of question and follow-up, it was not feasible to make a finer-grained analysis.
Furthermore, since episodes differed substantially in length (Range 4-74 sequences; Mean 24.5), it
was judged inappropriate to make comparisons based on raw frequencies.

Accordingly, a number of proportional measures were derived from the raw data for each episode.
The first was Proportion of Sequences Initiated by a Teacher Question (T.Seq.Init.Q). Since a
teacher sequence initiating question is a prerequisite for a sequence to be treated as triadic dialogue,
this measure provided the base from which proportions of the different types of question and the
overall proportion of evaluative follow-ups were calculated. The other measures were: Proportion of
Known Information Questions (%Known Inf Q, calculated as % of T.Seq.Init.Q); Proportion of
Negotiatory Questions (%Negot Q, calculated as % of T.Seq.Init.Q); Proportion of Teacher
Evaluative Follow-ups (%Total Eval, calculated as % of T.Seq.Init.Q); Proportion of Responses to
Known Information Questions followed up by Evaluation (%Eval KIQ; calculated as Eval.F-up to
Known Inf Qs as % of Known Inf Qs); Proportion of Responses to Negotiatory Questions
followed up by Evaluation (%Eval NQ; calculated as Eval.F-up to Negot.Qs as % of Negot.Qs). In
addition, as described in the first part of this paper, two measures of length and complexity were
calculated: a measure of the number of exchanges per sequence, Mean Sequence Length (MSL);
and a measure of complexity of student responses, Mean Response Length (MRL). Finally, the
proportion of Student-Initiated Sequences (%S-Init.) was calculated for each episode.

In the first analysis, using the raw frequencies, all the episodes in the database were divided into two
groups, which were labeled Science and Arts (the latter combining activities in literature and

18

history), as it seemed likely that triadic dialogue might take different forms in these different
curricular contexts. The two sets were then compared. The first and most salient result of this
comparison was the difference in number and types of Episode Activity Orientations used in
Science as compared with Arts. Of the total range of different types of EAO (see Appendix, Cat. 1),
9 were used in Science, while only 5 were used in Arts. Second, there were some related differences
in the relative frequencies of Known Information and Negotiatory Questions between the two
domains. For example, of the 265 sequences in the episodes of Constructing in Science, 69 (26%)
were initiated by Known Information Questions (KIQs) and 124 (47%) by Negotiatory Questions
(NQs); by comparison, in Constructing in the Arts, where there were 70 sequences, only 5 (7%)
were initiated by KIQs while 44 (63%) were initated by NQs. In Review episodes the relative
frequencies were even more different: in Science, where there were 70 sequences, 33 (47%) were
initated by KIQs and 23 (33%) by NQs, while in the 80 sequences in Review episodes in Arts the
comparable frequencies were 24 (30%) KIQs and 42 (53%) NQs. Relative frequencies of
Evaluative Follow-ups also differed substantially between the two domains.

The above within and across Science and Arts (literature + history) comparisons suggested that
there was a tendency for science and arts activities to be carried out by means of episodes with
different activity orientations to achieve their goals. In addition, there appeared to be relationships
among EAOs, the types of questioning moves that initiated their constituent sequences, and the
incidence of evaluative follow-ups. However, since the Arts episodes were distributed over a
smaller range of Episode Activity Orientations than the Science episodes, it was not possible to
make comparisons between Arts and Science for each of the EAOs taken one at a time. At a higher
level of abstraction, however, both domains included episodes realizing two broad groupings of
orientations with respect to the larger activities in which they occurred. These we characterized as
Managing and Exploring.

19

As the next step, therefore, within each of the domains, we combined all the Episode Activity
Orientations (EAOs) into the two groups, Manage and Explore. The Manage category included
those EAOs concerned with the management of tasks and with checking on their completion:
Organization, Monitoring, Planning, Problem-solving, and Launching. The Explore category
included those EAOs concerned with the interpretation and evaluation of experience and
information: Generating, Constructing, Formulating, Reporting, and Reviewing. First, we examined
the raw frequencies obtained by pooling all the episodes within each of the four macro groupings
(see table 2).

Table 2. Raw Frequencies of Sequences, Teacher Initiating and Evaluative Follow-up Moves in
Manage v. Explore in Science v. Arts

SCIENCE
Manage
Explore
Total
ARTS
Manage
Explore
Total
TOTAL

Total
Sequ

Student
Init. Seq.

T.Seq.
Init Q

Known
Inf. Q

Pers Q

Negot.
Q

Total
Eval.

Eval
KIQ

Eval
NQ

Mean
MSL

Mean
MRL

260
523
783

18
68
86

208
407
615

82
111
193

2
23
25

109
264
373

91
151
242

35
44
79

49
104
153

1.58
1.89

1.61
1.72

56
246
302
1085

2
48
50
136

47
188
235
850

24
30
54
247

1
13
14
39

19
128
147
520

18
51
69
311

9
21
30
109

13
28
41
194

1.95
1.51

1.60
2.22

Table 2 shows the distribution, in Manage and Explore in Science and Arts, of the different types of
move in relation to the total number of sequences and to the total number of sequences initiated by a
teacher question (raw frequencies). From this table it can be calculated that, in all four macro
groupings, between 75% and 85% of sequences were initiated by a teacher question, of which the
majority asked for known information or negotiatory information. (The remaining teacher initiated
sequences involved either a Teacher Give move, a Demand for Action, or a question asking about
Personal Experience, see note10.) However, there were also differences between the groupings.
While the ratio of negotiatory questions to known information questions was considerably more
than two to one in Explore in both Science and Arts, the two types of initiating questions (i.e.

20

negotiatory and known information) were more closely balanced in Manage in Science (109 vs 82),
and in Manage in Arts there was actually a higher frequency of known information than negotiatory
questions (24 vs 19). There were also differences in the relative frequencies with which students
initiated sequences and with which their responses to teacher questions received an evaluative
follow-up. Student initiations were more frequent in Arts than in Science and, for both domains, in
Explore compared with Manage. With respect to evaluative follow-ups to student responses, in
Manage the proportion of sequences initated by a teacher question that involved evaluation was
44% in Science and 38% in Arts; in Explore, by comparison, the respective proportions were lower:
37% in Science and 27% in Arts.

However, for the reasons already explained, we considered it preferable to explore these
relationships by means of a reanalysis using the indices of proportional frequency
described above: Teacher Sequence Initiating Questions (%T.Seq.Init. Q), Known Information
Questions (%KIQ), Negotiatory Questions (%NQ), Total Teacher Evaluations (%Total Eval.),
Evaluation of Responses to Known Information Questions (%Eval KIQ), Evaluation of Responses
to Negotiatory Questions (%Eval NQ). In this analysis, we represented each variable by the mean
of the values for each of the episodes contributing to the cell.

The results of this further analysis (see figure 1) were corroborative of those obtained from the
analysis of raw frequencies. In all four macro-categories, we found a high frequency of teacher
initiating questions, suggesting that, irrespective of the curriculum domain and the kind of EAO, the
teachers in these episodes had a strong tendency to initiate sequences with a nuclear exchange in the
triadic dialogue format. When the type of initiating question (Known Information or Negotiatory)
was examined, however, there were found to be substantial differences. In Science, Negotiatory
Questions outnumbered the Known Information Questions in both Manage and Explore. In Arts,
on the other hand, Known Information Questions outnumbered Negotiatory Questions in the
Manage block, while the reverse was the case in Explore, with KIQs occurring far less frequently

21

than NQs. In general, however, the results show that the teachers involved in this study had an
overall preference for questions that initiated sequences of information-negotiation and coconstruction of knowledge rather than questions that elicited information assumed to be known (see
table 5 below).

Figure 1. Proportional Frequencies of Indices in Manage v Explore in Science v Arts


100

80

60
T. Seq. Init. Q
know n Inf. Q
40

Mean

Eval. Know n Inf. Q


Ne got. Q

20

Eval. Ne got. Q
0

Total Eval.
Manag e/Sci

Manage/Arts
Explore/Sci

Explore/Arts

Manage vs Explore

One of the most common findings in research on classroom interaction is that teachers not only ask
questions, but they also evaluate the students responses (Cazden, 1988; Mehan, 1979). However,

22

we hypothesized that the incidence of evaluation would vary with the type of initiating question and
that, more specifically, teacher evaluative follow-up moves would be more likely to occur in
sequences initiated by Known Information Questions than in those initiated by Negotiatory
Questions, since it is in the former case that there is most likely to be a "right answer". To evaluate
this hypothesis, we examined the relationship between the two types of initiating question and the
kind of follow-up that student responses received in each of the four macro-categories already
discussed.

As can be seen from figure 1, the overall frequency of Negotiatory questions far exceeded that of
Known Information questions. At the same time, it was responses to Negotiatory questions that
tended, proportionally, to be evaluated more frequently than those to Known Information questions,
The only exception to this clear trend occurred in Manage in Arts, where there were more Known
Information than Negotiatory questions and where, proportionally, responses to Known
Information questions were more likely to receive evaluative follow-ups than responses to
Negotiatory questions. These findings were rather unexpected, given our initial hypothesis, and they
will be further discussed below. Overall, however, these results suggest that triadic dialogue was
being used rather differently in the four macro-categories under investigation. T-tests were therefore
performed to establish whether any of these differences were statistically significant.

As table 3 shows, there were significant differences in Arts between Manage and Explore in the
proportions of both Known Information and Negotiatory questions. A much greater proportion of
Known Information Questions was used in Manage than in Explore episodes (p = .004), whereas
with Negotiatory Questions, the picture was reversed (p = 041). However, no significant
differences were found in Science, nor were there any with respect to frequencies of evaluation in
either domain.

23

Table 3. The Difference Between the Proportions of Known Inf. Q, Negot. Q, Eval. Known
Inf. Q and Eval. Negot. Q in Manage v Explore in Science and Arts

Manage v Explore

Mean

SD

Manage

30.46

31.76

Explore
Manage

32.09
59.19

35.11
31.72

Explore
Manage

58.22
13.51

36.91
15.12

Explore
Manage

14.29
27.80

21.00
23.62

Explore

20.69

20.73

Manage

58.92

22.72

Explore
Manage

9.28
27.14

14.59
38.38

Explore
Manage

79.76
21.07

25.29
5.55

Explore

5.65

12.44

Manage

18.57

26.26

Explore

14.92

9.38

t-value

Sig
(2-tailed)

-.136

.893

.078

.938

-.116

.908

.921

.364

3.96

.004

-2.44

.041

1.65

.137

.361

.728

Science
Known Inf. Q
Negot. Q
Eval. Known Inf. Q
Eval. Negot. Q
Arts

Known Inf. Q
Negot. Q
Eval. Known Inf. Q

Eval. Negot. Q

We also compared Mean Response Length (MRL) and Mean Sequence Length (MSL) in Manage
v Explore in Science and Arts (see figure 2). The results are reported in table 4. As this table shows,
there is a strong trend, in Arts, for Mean Response Length to be greater in Explore than in Manage.

24

In Science, on the other hand, there is little difference between Manage or Explore with respect to
either of these indices.

Figure 2. Mean Response Length & Mean Sequence Length in Manage v Explore in
Science v Arts

2.4

2.2

2.0

1.8

Mean

1.6
Mean Response Length

1.4

Mean Sequence Length


Manage/Sci

Manage/Arts
Explore/Sci

Explore/Arts

Manage vs Explore

Given the emphasis placed on inquiry in the action research project that generated these episodes of
interaction, we hypothesized that, irrespective of the kind of activity or Episode Activity Orientation,
if the teacher wished to encourage more equal participation in discussion, he or she would, in

25

general, be less willing to take the role of the primary knower and, instead, would pose questions
that solicited students opinions and conjectures or which invited a more exploratory stance to the
topic. To address this hypothesis, we examined the relative proportions of the two types of
question and of evaluative follow-ups to them in all episodes in our database combined.

Table 4. Mean Response Length and Mean Sequence Length in Science and Arts

Manage vs Explore

Mean

SD

Manage

16.07

2.97

Explore
Manage

17.17
15.81

3.79
6.46

Explore

18.93

6.97

Manage

15.97

2.54

Explore
Manage

22.16
19.48

3.99
1.67

Explore

15.08

6.54

t-value

Sig
(2-tailed)

-.881

.385

-1.306

.201

-2.03

.07

.904

.39

Science
Mean Response Length
Mean Sequence Length
Arts
Mean Response Length
Mean Sequence Length

As figure 3 shows, overall, the proportion of Negotiatory Questions was much greater than that of
Known Information Questions. The mean percentage of Known Information Questions was
28.68% while the mean percentage of Negotiatory Questions was 61.01%. A paired t-test showed
that this difference was statistically significant (table 5). With respect to the proportional frequency
of evaluative follow-ups to responses to these two types of question, on the other hand, the trend for
responses to Negotiatory questions to receive an evaluative follow-up more frequently than
responses to Known Information questions was not statistically significant either in the database as
a whole or in Science or Arts taken separately. Overall, therefore, these results suggest that - across
grade levels and across curricular domains - the teachers in these interactions did try to encourage

26

dialogic inquiry in their classrooms. That is, in general, they tended to pose questions to be
answered through negotiation and exploration of the topic more frequently than questions that
simply called for a display of known information.
Figure 3. Proportions of Known Information and Negotiatory Questions, and of
Evaluations of Responses to them in all Episodes

70

60

50

40

30

Mean

20

10

0
know n Inf. Q

Negot. Q
Eval. Know n Inf. Q

Eval. Negot. Q

Table 5. The Difference Between Known Inf. Q and Negot.Q, and Between Eval. Known
Inf. Q and Eval. Negot. Q

Variables

Mean

SD

Known Inf. Q

28.68

44

31.95

Negot. Q
Eval.Known Inf. Q

61.01
12.80

44
44

34.30
17.57

Eval. Negot. Q

21.65

44

20.16

Pair 1
Pair 2

t-value

Sig
(2-tailed)

-3.37

.002

-1.90

.064

27

In this connection, we were also interested in the relationship among the initiating questions, the
complexity of the responses they elicited, and the proportions of these responses that received
evaluative follow-ups. To investigate these relationships, we computed correlations across all
episodes among Known Information and Negotiatory Questions, their respective Evaluative Followups, and the students mean response length (MRL). The results of these analyses are reported in
table 6.

Table 6. Correlations Among Questions, Their Evaluation Follow-ups, and Mean


Response Length
Known Inf. Q
1.
2.
3.
4.
5.

Known Inf. Q
Eval. Known Inf. Q
Negot. Q
Eval. Negot. Q
MRL

.86***
-.84***
-.45*
-.31*

Eval. Known Inf. Q Negot. Q Eval. Negot. Q

-.80***
-.33
-.30*

.61***
.29*

- .25

*** Correlation is significant at the 0.001 level (2-tailed).


* Correlation is significant at the 0.05 level (2-tailed).
These results extend the previous findings in very interesting ways. First, there was a very
significant tendency for Negotiatory and Known Information Questions to be mutually exclusive
within episodes (r = -.84). At the same time, a significant positive relationship was found between
the proportional frequencies of both types of question and their evaluative follow-ups, although this
was much stronger in the case of Known Information Questions (r = .61 and r = .86, respectively).
In the case of Negotiatory Questions, it seems that, instead of evaluating student responses, the
teachers who asked more questions tended to provide follow-ups of other kinds, such as asking for
clarifications, explanations, or alternative opinions, or by offering comments or meta-comments of

28

their own. The proportional frequencies of these alternative types of follow-up vary too greatly
between episodes for statistical comparisons to be interpretable. However, the raw frequencies
presented in table 7 provide some evidence in support of this inference. In all episodes combined,
Acknowledge and Give occur proportionately more frequently following Negotiatory Questions
than following Known Information Questions, as do dependent exchanges initiated by a
Negotiatory Question.
Table 7. Raw Frequencies of Different Types of Follow-up to Known Information and
Negotiatory Questions in Manage v Explore in Science v Arts
Initiating Q

Ack

Eval

Give

Dep.KIQ

Dep.NQ

Emb.Q

KIQ
NQ

82
109

1
11

35
49

15
27

17
2

3
12

4
8

KIQ
NQ

111
264

31

44
104

26
45

5
9

5
35

23
40

KIQ
NQ

24
19

9
13

3
1

4
4

KIQ
NQ

30
128

21
28

3
20

52

2
6

1
22

KIQ
NQ

248
520

7
94

109
194

46
92

29
11

10
54

32
74

Science
Manage

Explore
Arts
Manage

Explore

All Episodes

A second set of interesting results has to do with the relationships between the questions and
follow-ups and the complexity of student responses. As table 6 shows, there is a significant
positive relationship between Negotiatory Questions and Mean Response Length (r = .29) and a
negative though non-significant relationship between MRL and evaluative follow-ups to Negot. Qs
(r = -.25). Conversely, there are negative relationships between Known Information Questions and
MRL (r = .31) and between evaluative follow-ups to Known Inf. Qs and MRL (r = .30), both of
them being significant. These correlations suggest that, as hypothesized, asking negotiatory
questions leads to longer and more complex student responses, while asking known information
questions has the opposite effect; also that where the teacher gives a greater proportion of evaluative

29

responses to either type of question there is a tendency for students to give shorter and less
complex responses.
In the light of these results, the finding, reported in table 5, that the difference in proportional
frequency of evaluative responses to Negotiatory and Known Information Questions was not
statistically significant was initially surprising. This prompted us to look more closely at the
transcripts and then at some of the videorecordings of the episodes in which high frequencies of
evaluations occurred. What we noticed was that the choice of evaluative option tended to differ
between the two types of questions. Whereas evaluative follow-ups to responses to Known
Information Questions tended to evaluate whether the proffered information was correct or
incorrect, evaluative follow-ups to responses following Negotiatory Questions often took the form
of praise for the relevance of the contribution, or repeated the response to ensure that everybody
could hear, or prefaced some form of elaboration. The following are two examples of such
evaluations. In the first (DZ9), the teacher is eliciting suggestions about where gears might be used.
She repeats Andrew's suggestion (a form of evaluation) prior to adding it to the list of suggestions
already offered.

Teacher:

Any other ideas? Andrew?

Andrew:

Cranes and tools

Teacher:

Cranes- great- ok and tools (writes)

The second example is again taken from a science lesson (HK1) in which students are being asked
for their suggested explanations; here the follow-up is one of praise.

Teacher:

If you know how light behaves, how do we explain the stretched or


squashed mirror image? Adam?

Adam:

There are lots of reasons . for example it's angled 90 degrees here

Sheila:

Like feet . like the light stretches it . it can go there

30

Teacher:

Good . OK . terrific

In these and similar sequences, although the teacher's follow-up move has the form of an evaluation,
the function it is performing is more one of encouragement than of judgement, as is readily
apparent from the intonation and non-verbal signals.12 Further support for this interpretation can
be found in the fact that, in episodes in which teachers made such encouraging evaluations, there
was no tendency for students to curtail the complexity of their responses and, as in the immediately
preceding example, they quite frequently built on each other's responses.

The final example (table 8) is an extended sequence that illustrates the range of follow-up moves
chosen as alternatives to evaluation. The episode in which it occurred (AJ8) took place during a
biology unit in a grade six class, in which the students had been observing the development of
painted lady caterpillars.13 Most of the caterpillars had reached the stage at which they had attached
themselves to the gauze covering of the plastic cups in which they were kept and had spun the
cocoons within which they would metamorphose into butterflies. How, then, should the state of the
one or two cocoons that remained on the bottom of the cups be interpreted?

Several features are worthy of note. First, the sequence starts with a student initiation, posing a
substantive problem for solution. In her role as manager of the discussion, the teacher does not
answer this question herself but, in a lengthy follow-up move (move 2), she first amplifies the
student's question so that the problem is clear for all to consider, then she invites suggestions.
Having drawn an inference from the first suggestion, she invites further discussion by asking a
student who agrees with the inference to justify his agreement (move 11). Since the proffered
justification includes a critical assumption, the teacher calls on him to further justify his reason for
making it. Finally, she summarizes the arguments by both students that led to the conclusion that
the cocoon is dead (move 15).

31

Table 8. Investigating the Metamorphosis of Caterpillars in Grade 6


Text

Exch

R: OK, what do we do if the


caterpillars are on the floor and
they're- they're xx?
T: One of them is hanging and one is
on the floor in its cocoon stage
Any ideas?
- not hanging but it's in its cocoon
stage .. it's in a cocoon already ..
What should we do?
Any suggestions?

T: Michael?
M: I think that if it dropped . I think
it's dead now because it has to be
hanging .. and if it dropped there's
quite a way to drop so if they
drop they're probably damaged
SS: [some agree others disagree in
low voices]
T: Excuse me . you'll have a chance
to talk . let Michael talk.
M: Well, if it was hanging on the side
then it must have not had enough
energy to get right to the top
T: OK, so you're - you're um
concluding that it's dead
T: How many people agree with
Michael?
N: [signals that he agrees]
T: OK, why do you agree with
Michael, Nir?

Initiator
R

Prospect
D

Function
Sugg

Function

Ampl

D
G

Sugg
Ampl

Sugg

SS

Acc/
Rej

Ack
G

Ampl

Ack
D

Just

Conj

Conj

Bid
(NV)

Just

N: Because if it was hanging on the


side and it fell it probably got dead
like cause it should stay in the
same spot . without moving
T: Had that (text) said that anywhere
um <that it- that it can-> that it
can . die <if it's in the same->. Um
if it's moved in the cocoon stage?
D

Prospect

Opin
N

F-up

Act

D
Cont
.

Function

Nom
M

N: No, but like if someone is moving


you a lot like so you'll fall after all
you can get- ...
T: OK, so you agree with Michael that
it died because you think it's
moved and that because it fell it's
damaged somehow and it died, OK

Responder

Just

Conf/
Just

Sum

Just
N

Just

32

In this sequence, there are thus several points at which the teacher chooses not to foreclose the
discussion with her own answer to the problem, given in the role of primary knower. Furthermore,
by demanding justifications rather than evaluating the student suggestions, she elicits additional
evidence and reasoning relevant to the drawing of a warranted conclusion. Thus the overall result
of the teacher's choice of follow-up options in this and several further sequences is an extended
discussion in which various perspectives are considered in a co-constructed attempt to resolve
Alicia's problem.

Discussion

The findings of the quantitative analyses of this corpus of teacher-whole-class interaction, together
with the quoted examples, both show the ubiquity of triadic dialogue and give some idea of the
range of functions this genre was used to perform. On this basis, we have to conclude that, even
when teachers are attempting to create a more dialogic style of interaction in their classrooms, triadic
dialogue continues to be the dominant discourse genre. However, considered further, the fact that
sequences of teacher-whole-class interaction should frequently start with a question is hardly
surprising, since a question both proposes an issue for discussion and, because of its high level of
prospectiveness, requires the recipient(s) to contribute to the issue in response. In addition, the
inclusion of a follow-up move allows the teacher to work with the students response in a variety of
ways. For this reason, this format can be an appropriate operationalization of a wide variety of
tasks, even across quite different teaching philosophies.

In the context of monitoring, for example, triadic dialogue allows the teacher to test or check
students grasp or retention of taught material; and, even with new material, the same genre permits
the teacher to engage in a form of co-construction of knowledge (Heap, 1985) in which, as Newman
et al., (1989) argue, it has the particular merit of having "a built-in repair structure in the teacher's

33

last turn so that incorrect information can be replaced with the right answers" (p.127). As has been
amply demonstrated by these and other researchers, these are the functions that tend to dominate in
traditional classrooms; as has also been demonstrated, however, the consequence of asking
questions mainly for known information is that teachers do indeed limit students opportunities
to try out their own ideas and, in so doing, to progressively master the distinctive discourses of the
disciplines (Nystrand & Gamoran, 1991).

From the social constructivist, inquiry perspective shared by the members of our research group,
therefore, this restrictive consequence is a distinct disadvantage. Nevertheless, as we have found,
even in these classrooms triadic dialogue still plays an important role in initiating sequences of
discussion. However, in these sequences, it is much more common for the teacher to make moves
that enable both students and teacher to contribute substantively in an attempt to co-construct
understanding of an issue for which there is not a single correct answer and in which the goal is, if
possible, to arrive at a consensus after considering a variety of alternatives.

Certainly, the choice of initiating question has an important influence on the way in which a
sequence develops; questions that introduce issues as for negotiation are more likely than known
information questions to elicit substantive student contributions and to encourage a variety of
perspectives. However, the choice of follow-up is even more important. As the quantitative results
make clear, where student responses to questions are frequently given an evaluative follow-up, this
tends to suppress extended student participation - though, as noted, in the case of negotiatory
questions many such follow-up moves were encouraging rather than evaluating and did not have
this negative effect. Conversely, even sequences that start with known information questions can
develop into more equal dialogue if, in the follow-up move, the teacher avoids evaluation and instead
requests justifications, connections or counter-arguments and allows students to self-select in
making their contributions. When this happens, of course, the initial IRF generic structure fades
into the background and is replaced, temporarily, by a more conversation-like genre.

34

It is still too early to attempt to put forward a taxonomy of the sub-genres of triadic dialogue. For
that, it would be necessary to investigate a larger and more representative corpus of data. However,
the present analysis suggests that a useful start could be made by considering the different roles
that the teacher chooses at the beginning of each sequence and reciprocally assigns to students.
The first of these, as already discussed in the introduction, is that of primary knower. Three
possibilities are available: teacher as primary knower (and addressees as secondary knowers); a
specific student as primary knower (and teacher and other students as secondary knowers); and no
preselected knower, where all participants can offer contributions towards the co-construction of
knowledge. The second key role is that of manager of the discussion, with responsibility for
selecting speakers and deciding on the direction and pacing of the talk. For this role the default
option is that of teacher as manager, although there are occasions, such as when one or more
students are presenting to the rest of the class, when this role is handed over to a student. Even
here, though, the teacher inevitably has the final responsibility for ensuring that time is well spent
and that participants behave in an orderly manner (Peters, 1966).

If the teacher assumes the managerial role, there is a third and more tactical role that has to be
enacted: that of deciding whether to act as the sequence initiator or to allow a student to take on this
role. In either case, the teacher is likely to make the first move, either by initiating the nuclear
exchange with a demand or give move, or by using a preparatory exchange to nominate the speaker
who will initiate the nuclear exchange. Finally, if the teacher initiates the nuclear exchange with a
demand, there is a further choice: whether to ask a question to which the addressee is expected to
make a substantive response, or to state a position and ask whether the students agree with it or not
(Haneda, in preparation).

By considering all the possible combinations of options associated with these three roles, it is
possible to distinguish a set of relationships among teacher, students and the knowledge available to

35

them that might potentially be discursively realized in any particular sequence. This is obviously an
important basis for distinguishing between sub-genres of triadic dialogue. However, there is still
the follow-up move to consider and, as we found, the way in which this is realized is not predictable
from the nature of the initiating move. There is certainly a positive correlation between teacher as
manager + primary knower + initiator, who asks a question requiring a substantive response' (the
default option) and the selection of evaluation as the type of follow-up move. But even when the
teacher adopts a less dominant role, he or she may still feel the need to do more than simply
acknowledge a student contribution, either by making a comment or asking a follow-up question
that invites the student to extend or qualify the initial contribution. Perhaps this is the most
fundamental role of all and what ultimately defines what it is to be a teacher in a whole class setting
(Oyler, 1996; Rogoff, 1994). It probably also goes a long way toward accounting for the
pervasiveness of triadic dialogue in a wide variety of instructional settings.

References
Bakhtin, M. M. (1986). Speech genres and other late essays. (Y. McGee, Trans.). Austin:
University of Texas Press.
Barnes, D. (1976). From communication to curriculum. Harmondsworth, UK: Penguin.
Bereiter, C. (1994). Implications of postmodernism for science, or, science as progressive
discourse. Educational Psychologist, 29(1), 3-12.
Berry, M. (1981). Systemic linguistics and discourse analysis: a multi-layered approach to
exchange structure. In M. Coulthard & M. Montgomery (Eds.), Studies in
discourse analysis . London: Routledge and Kegan Paul.
Cazden, C.B. (1988) Classroom discourse. Porstmouth, NH: Heinemann Educational
Books.
Dewey, J. (1938). Experience and education. New York: Macmillan.
Donoahue, Z. (1998) Giving children control: Fourth graders initiate and sustain
discussions after teacher read-alouds. Orbit, 29 (3): 18-21.

36

Eggins, S., & Slade, D. (1997). Analysing casual conversation. London: Cassell.
Engestrm, Y. (1990). Learning, working and imagining: Twelve studies in activity theory.
Helsinki: Orienta-Konsultit.
Engestrm, Y. (1991). Non scolae sed vitae discimus: Toward overcoming the
encapsulation of school learning. Learning and Instruction, 1, 243-259.
Galbraith, B., Van Tassell, M.A. and Wells, G. (1997) Aprendizaje y enseanza en la zona de
desarrollo proximo (Learning and teaching in the zone of proximal development). In
A. Alvarez (Ed.) Hacia un curriculum cultural: La vigencia de Vygotski en la
educacion. Madrid: Fundacion Infancia y Aprendizaje,pp. 55-76.
Halliday, M. A. K. (1984). Language as code and language as behaviour: A systemic
functional interpretation of the nature and ontogenesis of language. In R. Fawcett,
M. A. K. Halliday, S. M. Lamb, & A. Makkai (Eds.), The semiotics of culture and
language, Vol.1 . London: Frances Pinter.
Haneda, M. (2000). Negotiating meaning in writing conferences around students' goals for
revision. Ph.D. thesis, OISE, University of Toronto.
Heap, J. L. (1985). Discourse in the production of classroom knowledge: Reading lessons.
Curriculum Inquiry, 15(3), 247-279.
Hume K. (1998) Co-researching with students: An activity for a knowledge building
community. Networks, 1 (Sept. 1998) http://www.oise.utoronto.ca/~ctd/networks
Kamberelis, G. (1995). Genre as institutionally-informed social practice. Journal of
Contemporary Legal Issues, 6, 115-171.
Labov, W., & Fanshel, D. (1977). Therapeutic discourse: Psychotherapy as conversation.
New York: Academic Press.
Lemke, J. L. (1990). Talking science: Language, learning, and values. orwood, NJ: Ablex.
Leont'ev, A. N. (1981). The problem of activity in psychology. In J.V.Wertsch (Ed.), The
concept of activity in Soviet Psychology (pp. 37-71). Armonk, NY: Sharpe.
Martin, J. R. (1992). English texts. Amsterdam: John Benjamins.

37

Mehan, H. (1979). Learning lessons: Social organization in the classroom. Cambridge,


MA: Harvard University Press.
Mercer, N. (1995). The guided construction of knowledge. Clevedon UK: Multilingual
Matters.
Miller, C. R. (1984). Genre as social action. Quarterly Journal of Speech, 70, 151-167.
Moll, L.C. (Ed.) (1990). Vygotsky and education: Instructional implications and
applications of sociohistorical psychology. New York: Cambridge University Press.
Newman, D., Griffin, P., & Cole, M. (1989). The construction zone: Working for cognitive
change in school. Cambridge: Cambridge University Press.
Nystrand, M. (1997). Opening dialogue: Understanding the dynamics of language and
learning in the English classroom. New York: Teachers College Press.
Nystrand, M. & Gamoran, A. (1991) Student engagement: When recitation becomes
conversation. In H.C. Waxman & H.J. Walberg (Eds.) Effective teaching: Current
research. Berkeley, CA: McCutchan Publishing Corp.
Oyler,C. (1996) Making room for students: Sharing teacher authority in room 104. New
York: Teachers College Press.
Peters, R.S. (1966) Ethics and education. London: Allen & Unwin.
Rogoff, B. (1994) Developing understanding of the idea of Communities of Learners.
Mind, Culture and Activity, 1 (4): 209-229.
Sacks, H., Schegloff, E. A., & Jefferson, G. (1974). A simplest systematics for the
organizing of turn-taking for conversation. Language, 50(4), 696-735.
Shechter, M. (1998) Mrs. Frisby and the grade four children. Orbit, 29 (3): 22-25.
Sinclair, J. M., & Coulthard, M. (1975). Towards an analysis of discourse: The English
used by teachers and pupils. London: Oxford University Press.
Vygotsky, L. S. (1978). Mind in society: The development of higher psychological
processes. Cambridge, MA: Harvard University Press.

38

Vygotsky, L. S. (1934/1987). Thinking and speech. In R. W. Rieber & A. S. Carton (Eds.),


The collected works of L.S. Vygotsky, Volume 1: Problems of general psychology .
New York: Plenum.
Wegerif, R., & Mercer, N. (1997). A dialogical framework for researching peer talk. In R.
Wegerif & P. Scrimshaw (Eds.), Computers and talk in the primary classroom .
Clevedon, U.K.: Multilingual Matters.
Wells, G. (1993a). Reevaluating the IRF sequence: A proposal for the articulation of
theories of activity and discourse for the analysis of teaching and learning in the
classroom. Linguistics and Education, 5: 1-38.
Wells, G. (1993b). Working with a teacher in the zone of proximal development: Action
research on the learning and teaching of science. Journal of the Society for
Accelerative Learning and Teaching, 18: 127-222.
Wells, G. (1996). Using the tool-kit of discourse in the activity of learning and teaching.
Mind, Culture, and Activity, 3(2), 74-101.
Wells, G. (1997). From guessing to predicting: Progressive discourse in the learning and
teaching of science. In C. Coll & D. Edwards (Eds.), Teaching, learning and
classroom discourse: Approaches to the study of educational discourse. (pp. 67-87).
Madrid: Fundacin Infancia y Aprendizaje.
Wells, G. (1999) Dialogic inquiry: Towards a sociocultural practice and theory of
education. New York: Cambridge University Press.
Wells, G. (Ed.) (2001) Talk, text, and inquiry. New York: Teachers College Press
Wertsch, J. V. (1991). Voices of the mind : a sociocultural approach to mediated action.
Cambridge, MA: Harvard University Press.
Wood, D. (1992). Teaching talk. In K. Norman (Ed.), Thinking voices: The work of the
National Oracy Project . London: Hodder and Stoughton for the National
Curriculum Council.

39

Appendix
DEVELOPING INQUIRING COMMUNITIES IN EDUCATION PROJECT (DICEP)
ABBREVIATED VERSION OF THE CODING SCHEME

The instrument assumes that coding will be carried out by (a) segmenting the transcript into its successive
constituent sequences, and then (b) for each sequence - and the exchanges that make it up - entering the
appropriate code in each of the numbered columns in the coding protocol corresponding to the categories
in the coding instrument. The hierarchical relationship between the units for analysis can be diagrammed
as follows: (parentheses and dotted lines indicate optional status)

Episode

Sequence 1

...

(Sequence n)

(Preparatory Exch) Nuclear Exch (Dependent Exch) (Embedded Exch)

Initiating Move

Responding Move

(Follow-up Move)

In this abbreviated version of the coding scheme, only those categories referred to in the text are included.
To facilitate cross-referencing from the text, the major categories are also numbered sequentially.

1. Episode Activity Orientation


O

Organizing

Problem-Solving

Planning

Generating

Reporting

Constructing

Launching

Formulating

Monitoring

Reviewing

40

2. Exchange Type
N

Nuclear

Dependent

Preparatory

Embedded

Give

3. Prospectiveness
D

Demand

Ack

Acknowledge

4. Function
(a) Teacher Only Moves
Nom

Nomination

Expos Exposition

(b) Assumed Known Information


F

Fact

RGAns

Rule-governed answer

Connection

PubEv

Report of public event

Conventional expl.

(c) Personal
E

Experience

Imagination

(d) For Negotiation


Opin

Opinion

Pred

Prediction

Conj

Conjecture

Expl

Explanation

Conn Connection

Sugg Suggestion

(e) Action
Act

Action

Bid

Bid

Intend Intention

41

(f) Clarification
Clar

Clarification

5. Follow Up: Give


(a) Evaluation
Acc

Accept

Rej

Reject

Ref

Reformulate

Praise Praise

(b) Comment
Exem Exemplification

Ampl Amplification

Conn Connection

Sum

Summarize

(c) Metatalk
Metacog

Metacognitive

Metatop

Metaorg

Metaorganizational

6. Follow Up: Demand


(a) Comment
Exem Exemplification

Ampl Amplification

Conn Connection

Opin Opinion

Just

Sum

Justification

Summarize

(b) Clarification
Rep

Repetition

Conf Confirmation
1997 Gordon Wells and DICEP.

Ident Identification

Metatopic

42

Notes
1

The research reported in this paper was supported by grants from the Spencer Foundation to the

Ontario Institute for Studies in Education, (Principal Investigators: Patrick Allen, Myriam Shechter
and Gordon Wells). However, the views expressed are those of the authors and not necessarily
those of the Foundation. We should like to thank all the teachers and students who participated in
the project and made the data available for analysis. We should also like to thank those who coded
the discourse data: Hameed Esmaeile, Nancy Reiner, Frances Giampapa, Lynn Nigalis, Myriam
Shechter, and Dale Vaillancourt. We should also like to express our gratitude to the anonymous
reviewers for their constructive suggestions and for their helpful prompts to return once more to the
primary data.
2

A similar role is played by the chair of a large committee; he or she directs the movement through

the agenda, nominating speakers and often adding metacomments on the progress of the
discussion.
3

In this and the following transcripts, the following conventions are used: < > enclose segments

where the transcription is in doubt; * indicates an unintelligible word; CAPS indicate a segment
spoken with emphasis; underline indicates segments spoken simultaneously; . a period marks
approximately one second of pause.
4

Several students in turn may answer a single question, with or without the question being

repeated. In this case, the coding for the exchange type includes & to indicate that the current
exchange can be treated as, in some senses, paratactic with the preceding one of the same type.
5At

first sight this example does look like a case of guess whats in teachers mind. But it takes

on a different slant when it is set in context. Immediately preceding the teachers first question, a
student had asked why, if reinforcing concrete made it stronger, the bridge beside the Oakland

43

stadium had collapsed. With that information, it is possible to see the teachers line of questioning
- and her negative evaluation of some of the student answers - as an attempt to help the student
questioner to make the connection for himself. A further point that emerges from a closer look at
this example is that the individual exchanges (separated by blank lines above) are not unconnected.
Rather, they constitute a series of prompts that eventually leads another student to offer an answer
that can be expanded by the teacher as an explanation for the bridges collapse.
6

In some cases, sequences include more than one nuclear exchange as, for example, when the

teacher solicits opinions from several students before moving on to compare their merits.

Hallidays (1984) terms were Demand, Give, Accept. However, since accept is a sub-

category of evaluation, we have chosen the term Acknowledge for the weakest level of
prospectiveness.

What is at issue here is not whether the teacher actually 'knew' the answer, but whether s/he was

genuinely willing to give serious consideration to students' beliefs, opinions and explanations and to
expect their peers to do so as well.

It is important to note that, although discourse data were frequently used as evidence in the

teachers individual inquiries, there was no attempt made to persuade them to adopt any particular
style of discourse. Furthermore, the decision to investigate issues concerning the follow-up move
was made in the last year of the project, by which time the vast majority of the recorded
observations had already been made. In other words, the results reported in this paper were not
biased by teachers awareness of the focus of the analyses that were ultimately carried out.

44

10

It is customary in investigations involving coded data to carry out an analysis of inter-coder

reliability. However, in the present case, this was neither feasible nor theoretically justifiable. As
was explained above, even participants do not always agree on unit boundaries or on the functions
of moves. The same is equally true of observers or coders, since they have no privileged access to
what participants intended or understood. Rather than train further coders to carry out a reliability
check, which would have introduced yet a further level of differential interpretation, the principal
investigator checked each coded episode and, where necessary, recoded segments judged to have
been misconstrued. The proportion of coding judgments that was revised was considerably below
5%. Although this procedure was less than ideal, it guaranteed greater consistency in the coding
overall. (This issue of reliability in the coding of discourse data is discussed in a forthcoming
paper.)

11As

can be seen from table 2, the frequency of questions requesting personal information was, in

general, very low. For this reason, only known and negotiatory information questions were
included in subsequent analyses. In addition, a small proportion of teacher-initiated sequences did,
in fact, start with a Give move rather than a Demand. However, since initiating Give moves by a
teacher do not normally receive any overt student acknowledgement - and therefore no teacher
Follow-up - sequences initiated by a teacher Give move were not further coded, nor were they
included in the analysis.

12

Of course, the teacher's enthusiastic acceptance of the student's contribution is itself an evaluation

of the student's participation. But this is more related to the teacher's role of manager than to that of
(authoritative) primary knower.

13

See Wells (1993b) for an extended discussion of this episode.