Content uploaded by Rupert Wegerif
Author content
All content in this area was uploaded by Rupert Wegerif on Oct 06, 2018
Content may be subject to copyright.
Pre-print draft of
FROM SOCIAL INTERACTION TO INDIVIDUAL REASONING: AN
EMPIRICAL INVESTIGATION OF A POSSIBLE SOCIO-CULTURAL
MODEL OF COGNITIVE DEVELOPMENT
Published as: Wegerif, R., Mercer, N., and Dawes, L. (1999) From social interaction to individual reasoning:
an empirical investigation of a possible socio-cultural model of cognitive development. Learning and
Instruction. 9 (5): 493-516 ISSN 0959-4752
Rupert Wegerif, Neil Mercer and Lyn Dawes
Correspondence address:
Dr Rupert Wegerif
Centre for Language and Communications
School of Education
The Open University
Walton Hall, Milton Keynes, MK7 6AA
tel: + 44 (0) 1908 653680
fax: + 44 (0) 1908 654111
email: R.B.Wegerif@open.ac.uk
Abstract
This study explores the theory that individual reasoning ability, as measured using
standard reasoning tests, has part of its origin in dialogue with others. In the study 64 eight
and nine year old children were taught the use of 'Exploratory talk', a type of talk in which
joint reasoning is made explicit. The relationship between the talk of the children and the
solving of Raven's test problems was followed using discourse analysis of groups working
together. The findings of the study support four claims: that use of exploratory talk can
improve group reasoning, that exploratory talk can be taught, that the teaching of
exploratory talk can successfully transfer between educational contexts and that individual
results on a standard non-verbal reasoning test significantly improved as a result of the
intervention teaching exploratory talk. Our results offer support to the hypothesis that
experience of social reasoning can improve scores on measures of individual reasoning.
The stronger hypothesis that general cognitive development is a product of induction into
social reasoning is not supported.
From Social Interaction to Individual Reasoning: An Empirical Investigation of a Possible
Model of Cognitive Development
The growing influence of the socio-cultural paradigm has led to several studies of how children and
others learn to think in particular ways through being inducted into social practices and ways of
using language (see, for example, Rogoff, 1990; Resnick, Levine & Teasley, 1991; Mercer, 1995;
Wertsch, 1991). These studies of specific cognitive development raise a question about the
possibility of studying cognitive development in general. If learning to think is always induction into a
specific social context, then how are we now to understand the findings of non-culturally based
research on cognitive development? Gauvain, writing from a socio-cultural perspective, claims that
this question poses an important challenge to researchers in the field of learning and instruction:
Although the extent to which cultural considerations can be incorporated into traditional views of
development is unclear, it would be foolhardy to dismiss a long tradition of careful research
rooted in such traditions simply because it did not take into account cultural influences on
development. Thus a difficult task for the future is the reconciliation of findings from non-
culturally based research with culturally based investigations of the same and related phenomena
(Gauvain, 1995, p. 42)
In this paper we present a study that, in its theoretical basis, its methods and also its findings, goes
some way towards constructing a bridge between a socio-cultural perspective on learning to think
and the non-culturally based tradition of research on cognitive development that Gauvain refers to.
The study is based on the argument that a type of language use which (following Barnes & Todd,
1977) we call 'exploratory talk' embodies the kind of reasoning which is valued in a range of
'educated' cultural activities. In the study children were taught to use this kind of talk and the effects
of this experience on their joint activity and their ability to solve the problems of a standardised
psychological test of non-verbal reasoning (in groups and as individuals) were assessed. The test
used - the Raven's Progressive Matrices - is commonly held to test 'the ability to reason and solve
problems involving new information' (Carpenter, Just & Shell, 1990, p.404), while also correlating
with measures of academic attainment. Analysis of discourse enabled us to relate changes in the
way that children talked together while jointly solving the problems of one version of the Raven's
test to changes in their individual scores on a second version of the same test. This methodology,
combining qualitative discourse analysis with quantitative measurement and controlled experiment,
enabled us to explore connections between cultural practice, social interaction and individual
development.
Theoretical framework
The empirical study was designed to explore a model of individual cognitive development that
combines a dialogical description of reasoning with a version of Vygotsky's account of individual
development.
Resnick, Salmon, Zeitz, Wathen, & Holowchak. (1993) argue that whereas traditional psychology
has described reasoning in terms of logical rules or other formalisms various trends and arguments
in contemporary cognitive psychology point to the need to see reason as a form of social practice.
Elsewhere we have argued, influenced by Habermas's concept of communicative rationality
(Habermas, 1990, p. 89; White, 1988; Habermas, 1995) and the arguments of other philosophers
and communication theorists (for example Rorty, 1991, p. 39; Burbules & Rice, 1992), that to
describe reason as a social practice requires a description in terms of inter-personal orientations
and associated ground rules (Wegerif & Mercer, 1997a). By labelling our description of reason
'dialogical' we mean that it is not a model of reason drawn from the outside, after the event, as if
reason was a closed and finished system, but it is a description of those rules and orientations
which inform a type of dialogue from within, specifically those rules and orientations which serve to
maintain a free and open encounter between different perspectives and ideas (see Rommetveit,
1992, for further explication of the significance of the move from 'monological' to 'dialogical'
accounts of cognition).
Vygotsky's model of individual development, which has had a strong influence on the current socio-
cultural perspective, stresses that: 'all that is internal in the higher mental functions was at one time
external' (Vygotsky, 1991, p36). Vygotsky's categories of internal and external have proved
problematic for the contemporary socio-cultural perspective. Nonetheless Vygotsky's claim that an
individual's ability to perform cognitive tasks when acting alone presupposes and stems from a prior
socialisation process is still a basic tenet of the socio-cultural perspective. Some authors, following
Leont'ev (1981), refer to the movement of development which Vygotsky called 'internalisation' as a
process of the personal appropriation of cultural capital that results from a period of 'guided
participation' or 'cognitive apprenticeship' (Rogoff, 1990; Newman, Griffin & Cole, 1989; Rojas-
Drummond, Hernandez, Velez & Villagran, 1998).
The dialogical account of reasoning outlined above implies that reasoning is embedded in a social
practice. Neo-Vygotskian accounts of individual cognitive development focus on induction into
specific social practices. Combining these two views, a dialogical view of reasoning and a neo-
Vygotskian view of development, would lead to the conclusion that learning to reason is essentially
induction into a social practice. This model may be an oversimplification but as it is an extrapolation
of trends in contemporary theory it is worth articulating and evaluating. In the strong form presented
here it is a possible socio-cultural alternative to models of general cognitive development which are
central to the non-cultural tradition of developmental psychology. Some of the conceptual issues
dealt with only briefly here are considered in more detail in Wegerif (1996a) and in Wegerif &
Mercer (1997a).
Exploratory talk and reasoning
In order to apply and test the theoretical framework we have outlined, the notion of reasoning as a
social process has to be specified rather more precisely, in terms of actual situated social practice,
than has been done by philosophers such as Habermas and Rorty. To do this we have used the
concept of 'exploratory talk’ whose origin lies in empirical studies of classroom discourse.
In an article in Learning and Instruction, Mercer (1996a) used observational research in British
primary schools to typify three kinds of talk, which he also described as representing different
'social modes of thinking'.
1. The first way of talking is Disputational talk, which is characterised by disagreement and
individualised decision making. There are few attempts to pool resources, or to offer constructive
criticism of suggestions. [...] Disputational talk also has some characteristic discourse features.
notably short exchanges consisting of assertions and counter-assertions.
2. Next there is Cumulative talk, in which speakers build positively but uncritically on what the
other has said. Partners use talk to construct a "common knowledge" by accumulation. Cumulative
discourse is characterised by repetitions, confirmations and elaborations. [...]
3. Exploratory talk occurs when partners engage critically but constructively with each other's
ideas [...] Statements and suggestions are offered for joint consideration. These may be challenged
and counter-challenged, but challenges are justified and alternative hypotheses are offered.
Compared with the other two types, in exploratory talk knowledge is made more publicly
accountable and reasoning is more visible in the talk. Progress then emerges from the eventual
joint agreement reached.
(Mercer, 1996a, pp. 368-369)
These types of talk are discussed and illustrated in detail elsewhere (Mercer, 1995, 1996; Wegerif
& Mercer, 1997a). Of the three types we claim that exploratory talk is the closest to reasoning as a
social practice.
The more detailed elaboration of exploratory talk which follows stems from three influences:
conceptual considerations, particular those raised by Habermas (see Wegerif, 1996), the literature
on research on effective collaborative learning (see review in Mercer 1995, p. 90-95) and our
experience in classrooms working closely with teachers (see Dawes, 1997). Out of this
combination of sources the following pragmatic ground rules for exploratory talk are provisionally
proposed:
1 all relevant information is shared
2 the group seeks to reach agreement
3 the group takes responsibility for decisions
4 reasons are expected
5 challenges are acceptable
6 alternatives are discussed before a decision is taken
7 all in the group are encouraged to speak by other group members
The first three rules in the list are ground rules that are shared with cumulative talk, rules that serve
to bind the group, share information together and construct knowledge together through seeking
agreement. Rules four and five focus on the explicit reasoning that characterises exploratory talk as
opposed to other types of talk. The role of challenges is important in distinguishing between
cumulative, disputational and exploratory orientations. In exploratory talk challenges stimulate joint
reasoning, in cumulative talk they are experienced as disruptive and often lead to a loss of
cooperation and a switch into disputational talk. In disputational talk participants may still offer
apparent arguments but are in fact focusing on 'winning' rather than on understanding or solving a
problem together.
Ground rule six, that alternatives are discussed, reflects the findings of research on collaborative
problem solving, particularly that of Kruger (1993) which has found that groups which do best are
those which consider alternatives before deciding. In contradistinction to some researchers (e.g.
Howe, 1992) we argue that this generation of alternative views does not necessarily imply different
initial conception of the problem by the participants in collaboration but can itself be generated by
the ground-rules of the talk. Finally rule seven was a product of empirical experience working with
groups of children. We found that offering the abstract right to participate, found for example, in
Habermas's characterisation of the ideal speech situation (1991, p. 87) was not sufficient. In
practice children needed to be actively encouraged to speak and to put forward views by their
peers. These ground rules again emphasise our focus on the generative power of the interaction as
opposed to an emphasis on the prior dispositions and views of the participants.
'Exploratory talk', in which reasoning is made visible and publicly accountable through the
discussion of alternatives, offers us an empirically grounded version of what Habermas calls
'communicative rationality'. Engaging in this type of talk actively constructs participants as
‘reasonable’, (BenHabib, 1992) that is both giving reasons for claims and being open and
responsive to the reasoning of others. While communicative rationality of this sort is historically and
culturally situated it is nonetheless valued across a wide range of contemporary contexts. The
'exploratory talk' found and promoted in the classroom situation is a version of a type of language
use given central importance in contemporary cultural activities such as science, law, government
and the negotiation of business.
Aims of the study
The study reported had two broad aims, one to do with issues of practical pedagogy the other to do
with theory. The first aim was to find out if the type of talk we are calling ‘exploratory talk’ can be
taught and if this teaching can be successfully transferred from the teacher who originated the
programme in her classroom to other teachers in different schools. The second aim was to explore
the theory that performance on individual non-verbal reasoning tests is connected to prior
participation in social reasoning as embodied in 'exploratory talk'. This theory generates the
testable hypothesis that the effective coaching of exploratory talk will increase children's individual
results on established tests of reasoning. In order to test this hypothesis it is not enough to show a
connection between coaching exploratory talk and individual test results, it is also necessary to
show, in so far as this is possible, firstly that the ground rules of exploratory talk were instrumental
in helping to solve reasoning test problems when children worked in groups and secondly that the
coaching was effective in leading to the production of more exploratory talk.
The main hypotheses explored in this study are:
1 that the ground rules of ‘exploratory talk help groups to solve problems working together on
group reasoning exercises;
2 that the incidence of talk showing the features of exploratory talk can be increased by the
use of specially-designed teacher-led and peer-group activities;
3 that activities designed to teach exploratory talk which have been developed and
implemented in one school by one dedicated teacher can transfer effectively to further
schools and further teachers;
4 that increase in the use of exploratory talk in group exercises leads to increased scores by
individual pupils working alone on reasoning tests.
In addition the study as a whole explored whether coaching in exploratory improved collaborative
learning within the normal curriculum. This aspect of the study is not the focus of this paper and is
commented on more fully in Wegerif, Mercer and Dawes, 1998.
Methods
Phases of the study
The empirical study had two phases. In the first phase an intervention programme to coach
exploratory talk was developed and implemented in one base school. In the second phase this
intervention programme was implemented in classes in two further schools.
Experimental design
Each run of the intervention programme in a school was treated as a separate field experiment. For
each intervention:
• each target class was matched with a control class of the same age group in another local
state school;
• target and control classes were divided into groups of 3, and in those groups the children
attempted to solve the Raven's Standard Progressive Matrices (SPM - see illustration in
figure 1). They were given this test before the intervention and then again after the
intervention had been completed in the target school;
• members of both target and control classes were given Raven's Coloured Progressive
Matrices (CPM), which is made up of similar non-verbal problems, as individuals both
before and after the intervention programme;
• one focal group of three children in each class was video-taped while doing the group
reasoning tests. (Other video-recordings of classroom events were also made, but the
analysis of that data will not be discussed in this paper.)
Participants
60 children, aged 9 and 10 years old, in three local state middle schools composed the target
classes that took part in the TRAC implementation programme. As mentioned above, each target
class in each school was matched with a control class of the same age in another local state
middle school. This produced three control classes of 64 children in all, who had no part of the
intervention programme but who were observed and tested in the same ways as the target classes.
(Because of absences on the days of testing, a few children initially included as subjects in both
control and target classes dropped out of the study.) For the purpose of the tests and many of the
exercises in the intervention programme the children were required to work together in mixed
gender groups of three (plus one or two groups of two if numbers did not divide by three). These
groups were organised by the class teacher, so as to include a range of ability in each. There were
in total 23 target groups and 25 control groups. (Because of absences, three groups in target
classes and four in control classes varied slightly in membership between the beginning and end of
the study.)
The intervention programme
The TRAC intervention programme will be described only briefly here. A teachers' guide is being
prepared (Dawes, Mercer and Wegerif, in preparation) and more detail about the content of the
programme is provided elsewhere by its main originator (Dawes, 1995, 1997). The programme
consists of a series of nine lessons. Each lesson is designed to last for about one hour and
focuses on one or more of the ground rules of exploratory talk which were outlined earlier. The first
few lessons deal with skills such as listening, sharing information and co-operating, while later
lessons encourage children to make critical arguments for and against different cases. The children
are given opportunities to practice discussing alternative ideas, giving and asking for reasons and
ensuring that all members of the group are invited to contribute. Some computer-based group
activities are included, using specially-designed software.
The use of the ground rules was taught to the children through explicit modelling by the teacher,
coaching their use in whole group and small group discussions and giving opportunities for their
use by the children working in small groups without the teacher. The explicit modelling phase
involved the teacher at the front of the class illustrating the ways in which she wanted the children
to talk together. Asking ‘why?’, using ‘because’ to give reasons for statements, asking other
children what they think, reaching agreement before making a final decision. Different teachers
modelled the language of exploratory talk differently. Some of these differences were picked up
through analysing transcripts of the talk of the children. A key lesson in the programme involved
eliciting the ground rules for the children in their own words. This was the third lesson after the
children had all had some practice in collaborative activities. In a guided discussion the teacher
drew from the class the kind of rules that they think should be used in group work. The list that
resulted was then put on the wall in large letters. In the rest of the programme and in other lessons
where collaborative learning was used the teacher or children can then refer to the rules on the
wall and say ‘remember our rules’. Although each of the three classes studied produced a different
set of ground rules they were all similar to the ground rules for exploratory talk which we listed
earlier. Here is an example of the ground rules for one of the classes in the study:
Class 5D's Ground rules for talk
1. Discuss things together. That means
* ask everyone for their opinion
* ask for reasons why
* listen to people.
2. Be prepared to change your mind.
3. Think before you speak.
4. Respect other people's ideas- don't just use your own.
5. Share all the ideas and information you have.
6. Make sure the group agrees after talking.
Discourse analysis
In each of the four target classes, one 'focal group' of children was video-recorded as they worked
on the Raven's SPM test, before and after the intervention. All the groups were arranged by the
class teacher to be both mixed ability and mixed gender. The teachers were asked to select a focal
group that was typical of the other groups in the class. The results of these groups on the group
Raven’s test were never the best results in the class but were above average for two of the three
classes. From our observation of their talk together we hypothesise that the extra attention of
having a camera focused upon them may have encouraged them to treat the task more carefully
then they might have done without the camera.
The video-tapes of these focal groups were used for qualitative discourse analysis of the kind
developed and described by the present authors and associates (Edwards & Mercer, 1987; Mercer
& Fisher, 1993; Mercer, 1996b; Wegerif & Mercer 1997b). In addition transcripts of the talk of the
children were made and a computerised text analysis concordancer was used to search for pre-
and post-intervention differences. This method (described in more detail in Wegerif and Mercer,
1997b) involves integrating qualitative analysis of the full transcript with the abstraction of 'key
words in context' in order to generalise significant features and compare different transcripts.
Although essentially qualitative this approach also facilitates linking qualitative evaluation to
quantitative descriptions of texts. For example detailed qualitative analysis of the deliberations of all
the groups revealed that the word ‘think’, as in ‘I think’, was being used to put forward reasons. This
word was then included in a computerised search for key features in the talk of the focal groups
that indicated exploratory talk and so could be used to provide a quantitative measure for
comparison with other transcripts.
Raven's Progressive Matrices
Raven's Progressive Matrices (RPM) consist of graphical puzzles of a type illustrated in figure 1.
They are widely used in education and psychology as a test of 'non-verbal' reasoning. Raven's
tests are particularly appropriate for exploring the link between language practices and the non-
culturally based tradition of research in cognitive development, because they correlate well with
other similar tests of reasoning and with measures of academic achievement (Raven, Court &
Raven, 1995, Richardson, 1992, p. 129). Carpenter, Just & Shell write:
The centrality of the Raven test indicates not only that it is a good measure of intelligence, but also
that a theory of the processing in the Raven test should account for a good deal of the reasoning in
the other tests (Carpenter, Just & Shell, 1990, p. 428)
Richardson (1992) has argued that RPM do not measure abstract mental processes but rather the
ability to read a particular kind of representation. He has demonstrated that if the same logical
problems as those found in Raven's tests are presented in a way that makes more 'human sense',
using pictures of cars and teddy bears for example instead of abstract shapes, children respond
very differently with a different distribution of test-scores. We do not disagree with these criticisms
of the way that Raven's test results, and similar tests, have been interpreted. Our interest in
Raven's is as an index of a valued kind of cognitive ability in a particular cultural context. Our study
uses RPM to investigate the relationship between this valued individual cognitive ability and the
way children talk together.
We used two similar Raven's tests, the Standard Progressive Matrices (SPM), consisting of 60
problems, and the Coloured Progressive Matrices (CPM) consisting of 36 similar problems. The
CPM is appropriate for younger children (up to the age of 12) while the SPM is appropriate for any
age. Test scores from CPM and SPM can be translated into a common scale through the use of a
table provided (Raven, Court & Raven, 1995, p. 64). We used the SPM for children working
together in groups of three, giving each group a single book and a single answer sheet and
encouraging them to talk together in solving the problems. We also gave the same children the
CPM exactly three days later, asking them to work individually and following the guidelines for the
administration of the test given in the manual (Raven, Court & Raven, 1995).
Results
Different approaches were used to investigate the different hypotheses. Our results will therefore
be presented in relation to each hypothesis and the methods involved in testing it. Although the
main methods used have been described above this results section will also refer to the design of
each of these three investigations. Because of the methodological pluralism of the study, rigid
divisions between method, results and discussion would not help achieve clarity in our presentation
of the findings.
1. Evaluating exploratory talk
Our first hypothesis was that following the ground rules of 'exploratory talk' would help groups of
children to jointly solve reasoning problems. The best way to assess this hypothesis is through a
detailed analysis of the connection between the type of talk children used and their solving of
reasoning test problems.
We selected the target group that had achieved the greatest pre/post-intervention change in group
score on the Raven's SPM test to be the focus of this study. The group score had increased from
39 to 47 (the SPM has 60 items and so is measured on a 60 point scale). A comparison between
this group score and the scores of individual members on the other version of the same test, the
Raven's CPM (which has 36 items and so is scored out of 36) was made by converting both scores
to a common 36 point interval scale, following the procedure described in the Raven's test manual
(Raven, Court & Raven, 1995 p. 64). In the pre-intervention tests the group score was lower than
the highest individual score (31 to 32). In the post-intervention tests, however, the group score was
slightly higher than the highest individual score in the group (34 compared to 33). This suggested
that the striking improvement in group score after the intervention could not be accounted for by a
change in the quality of reasoning of one individual in the group, but was a product of a change in
the way the group reasoned together. To investigate this hypothesis further, we focused our
discourse analysis on the talk of the group when they dealt with eight problems of the Raven's
SPM test. These were problems that had been answered incorrectly in their pre-intervention
attempt, but had been correctly solved by the group after the intervention. Here, to illustrate our
analysis, are examples of the group's pre-intervention talk (which failed to produce the correct
answer) and also of their post-intervention talk (which succeeded in finding the correct answer)
when dealing with the same problem ( B12, shown in Figure 1 below).
Figure 1 Problem B12. Raven's SPM
Transcript 1 : pre-intervention talk in Group 1 School A1 on Raven's SPM problem
B12
(Note: for the sake of intelligibility, punctuation has been added to all transcripts. Contextual
information is presented in parentheses)
George: B12
Susan: We haven't done that.
Trisha: (giggles) this is where your Mum can see what you're really like at school George.
Trisha: Square and diamond, it's 2
George: No it's not
Trisha: It is 2
George: No it's not
Trisha: It is
George: No it's not
Susan: It's that one 6
Trisha: It is
George: No it's not it's got to be a square and a circle
Trisha: Its that, it has to be that, it has to be that, it has to be 6 because look they've only got that
(pointing to the pictures)
Susan: Look first they are starting with one of them things over (pointing) and then it has to be
black
George: Right, 6
Susan: No it isn't George
Trisha: That's number 2 because it goes bigger and bigger and bigger (Trisha is looking at the next
question)
Susan: It isn't George look at that one, no it isn't George
(George is writing '6' , which is the wrong answer. Susan pushes him)
George: Susan! All right someone else be the writer then
Susan: Me (forcibly takes paper)
Trisha: No, Susan you have to sit in this chair to be the writer (Trisha takes pen)
Trisha: Give it to George because he's quicker at it
Susan: Give me the pen (takes the pen)
Susan: Are we finished on that one, are we on that one now
Trisha: You're not allowed to do it
George: Let Trisha have a go when we get up to E
(Turn to next exercise)
Susan: Well what do you think it is you dur brain? (addressed to George)
Transcript 2: post-intervention talk in Group 1 School A1 on Raven's SPM problem
B12
Susan: Set C
Trisha: No we ain't done B
Trisha: That has got to be a diamond, a square with a diamond with a circle in that one, number 6,
do you agree?
George: No, what do you mean?
Trisha: OK no it's got to be square
Susan: I think it's number 6, that's the one
George: No it ain't
Susan: I think it's number 6
Trisha: No 'cause it's got to swing round every time, so there is a circle in it
Susan: Yes but it hasn't got a circle in there has it and that one has (indicating)
(3 sec pause. Concentrated faces)
George: I think it's number ...
Trisha: I think it's number 4 to be honest
George: I don't I think it's number 6
Trisha: I don't, I think it's number 3 look because that one (pointing) has that in the middle and it's
got a half one in the middle
George: Complicated ain't it?
Susan: No because that one is that, I think it's that one
George: No because look at that and look at that (pointing) they are the same, you can't have two the
same and it's got that one on, look Sue, it's got that one on and it's got that one on so its out
of them three.
Susan: That one, one, 'cause that's a ..
George: Yes but it's got to be that
Susan: It’s that because look that's got a square so it's just got to be empty
George: With no circle in so it's just got to be an empty square
Susan: No they are just normal boxes
Trisha: Look that's got a triangle, that's got a square, look that's got a square with a diamond with a
circle in, that's got a square with a diamond in and that's got a square with a circle in so
that's got to be a square
George: I don't understand this at all
Trisha: Because look on that they've taken the circle out yes? So on that you are going to take the
circle out because they have taken the circle out of that one
George: On this they have taken the circle out and on this they have taken the diamond out and on
this they have put them both in, so it should be a blank square because look it goes circle
square
Susan: It’s got to be a blank square. Yeah it is. Mrs Dawes is coming. (inaudible whisper)
George: Do you agree on number 5, do you agree on 5?
(George writes '5', which is the correct answer)
George: Who is doing C? Susan, right let's have these here, C1
Commentary
In the pre-intervention talk of Transcript 1, George challenges Trisha's first suggestion ('It is 2')
without giving a reason. Trisha offers no further justification for her suggestion. This leads into a
series of exchanges typical of the type of talk we call 'disputational', in which participants simply
assert their opposed views without reasoning. Susan then suggests 'It is that one 6' and this is
taken up by Trisha, and both she and Susan offer reasons. '6' is apparently agreed upon, and
George writes it down. However, Susan then appears to change her mind without saying what her
new opinion is (or she may be objecting to him writing the answer down before checking properly
with her and Trisha: no reason is made explicit). There is then a dispute about who should be
writing the answers on the answer sheet.
Transcript 2 illustrates some ways that the talk of the same children changed after doing the TRAC
programme. Compared with their pre-intervention talk, there are more long turns at talk, as more
elaborate explanations are given. Again, Trisha is the first to propose an answer, but this time she
does this not as a statement ('it is 2') but as an elaborated hypothesis with a question encouraging
debate ('That has got to be a diamond, a square with a diamond with a circle in that one, number 6,
do you agree?'). George asks for more explanation. Attempting to be explicit, Trisha appears to
see that she is wrong and changes her claim. George and Susan again engage in a 'disputational'
exchange but this is short-lived. After a pause (for thought?) the children return to using language
to think explicitly together about the problem. They come to agree that it is a kind of subtraction
problem, and so find the correct answer.
Many more of the essential features of exploratory talk - as represented by our 'ground rules' - are
evident in the post-intervention talk than in the pre-intervention talk. Explicit reasons for claims are
given, challenges are offered with reasons, several alternatives are considered before a decision is
reached, and the children can be seen seeking to reach agreement together. Explicit reasoning
may be represented in talk by the incidence of some specific linguistic forms, and we can see here
some 'key features ': the hypothetical nature of claims is indicated by a preceding 'I think', reasons
are linked to claims by the use of 'because' or ''cause' and agreement is sought through the
question 'do you agree?'. Explicit reasoning requires the linking of clauses and leads here to the
incidence of more longer utterances in the post intervention talk. This group solved a total of eight
new problems in the post-test which they had failed to solve in the pre-test. When we compare talk
that led to the group solving these problems correctly and with talk which led to wrong answers, we
find that there is a clear association with the relative incidence of these key linguistic features. This
can be seen from Table 1 (below) which compares the number of long utterances (where 'long' is
defined through taking an arbitrary cut-off point of being 100 characters in length or more when
transcribed), and the incidence of 'because', 'agree' and 'I think'.
Table 1 - Incidence of key features: comparing talk leading to correct answers with
talk leading to incorrect answers for Group A1 (Trisha, Susan and George).
Key linguistic feature Incidence in talk leading to
incorrect answers
Incidence in talk leading to
correct answers
long turns at talk 0 11
'Because' and 'cause' 6 26
'I think' 1 24
'agree' 3 18
This specification of some of the differences between successful talk and unsuccessful talk can be
used to provide us with an instrument to measure, if only in a very approximate way, features
indicative of the effectiveness of the talk of the children. This is applied in the next section where
we look at the evaluation of the effectiveness of the intervention programme in changing the talk of
the children.
2. Evaluating the intervention programme
Our second hypothesis was that the incidence of exploratory talk could be increased by the use of
specially-designed teacher-led and peer-group activities, and our third was that the effective
teaching of exploratory talk could transfer from the school where the programme originated with a
dedicated teacher to other schools. These claims for the programme were tested by performing a
computer-based discourse analysis of the children’s talk while they worked together on the Raven's
SPM test, before and after the TRAC intervention. As is explained below, it was also tested through
comparisons of group test scores before and after the intervention.
A method of discourse analysis using computer-based concordancer was applied to exploring the
differences between transcripts of the talk of focal groups of children in the four target classes
working on the same standard task (Ravens SPM) both before and after the intervention.
Above we saw from an extract of transcript that the exploratory talk which helped the children to
solve the problem led to certain indicative linguistic features, those given in Table 1. Elsewhere
detailed analysis revealed that modals, 'would', 'could', 'should' and 'might' were sometimes being
used, as well as because, to introduce reasoning and so these were included in the list of indicative
key words. Generalising these terms to all the transcripts collected of the talk of the focal groups
produced the results given in Table 2.
Table 2 Incidence of key linguistic features in the talk of all target focal groups
Key feature School A1 Group 1 School A1 Group 2 School B1
Group 1 School C1
Group 1
Pre Post Pre Post Pre Post Pre Post
because 25 100 53 40 12 45 25 30
agree 7 87 20000 2
I think 7 87 516 012 44 47
would 1 15 11 25 141 7
should 1 1 79022 1
could 2 14 317 010 1
might 1 2 23021 1
Totals 44 306 83 110 13 66 73 89
Table 3 Incidence of key linguistic features in the talk of all control focal groups
Key feature School A2 Group 1 School A2 Group 2 School B2
Group 1 School C2
Group 1
Pre Post Pre Post Pre Post Pre Post
because 28 15 34 25 28 17 30 24
agree 15 6 12 20 100 0
I think 35 18 27 44 351 3
would 7 3 31325 2
should 15 3 16 2000 2
could 6 9 709613 4
might 3 2 15 1302 1
Totals 109 56 114 93 47 30 51 36
Table 4 Incidence of long utterances for all focal groups
Condition School A 1&2
Group 1 School A 1&2
Group 2 School B 1&2
Group 1 School C 1&2
Group 1
Pre Post Pre Post Pre Post Pre Post
Target 1 35 24 36 0 6 13 25
Control 13 6 25 16 20 18 10 13
It can be seen from Table 2 that, overall, key features in the talk of the target focal groups
increased after the intervention. The same increase was not found in the talk of the control focal
groups shown in Table 3. A two way analysis of variance, with repeated measures on pre- and
post- intervention counts, was performed. This showed that the difference between the target and
control condition changes in the total count of key features was statistically significant (F = 3.9 one-
tailed p = 0.048). The same analysis on the number of long utterances (Table 4) also showed a
significant result (F = 8.9 one-tailed p = 0.012).
One can see that the increase varies across the schools, with an especially large shift in School A
Group 1, and a very small shift in School C Group 1. Key features also have different patterns of
incidence in the talk of each group (so for example in School A1 there is a high post-intervention
rate of use of 'agree' by School A Group 1 but not for School A Group 2) . On the basis of our
observation of the various teachers carrying out the TRAC programme, we speculate that these
differences are at least in part due to teachers placing more or less emphasis on the use of different
kinds of language. In School A, class 1, the teacher stressed the importance of reaching agreement
and modelled the use of the question: 'do you agree?'. Our analysis suggests that the teaching in
School A, class 2 might have lacked this element.
The use of key terms by the control groups reduced over the period of the intervention. This
reduction was not statistically significant. We have no explanation for it. On possible hypothesis is
that these groups found the task and situation less engaging the second time around than the first.
If this is true it would have implications for the scores on the group Raven’s tests. This unexplained
reduction in the use of key terms by the control groups means that our claim to statistical
significance for an increased use of key terms by target groups needs to be treated with caution.
It is also noticeable that the control groups began the intervention with a greater use of indicators of
exploratory talk than the target groups. These indicators are only a rough guide to the exploratory
talk taking place. However this difference in the amount of indicators of exploratory talk was
reflected in higher scores on the group Raven’s test in the pre-test for the control focal groups than
for the target focal groups.
Group test scores
Table 5 Group scores on the Raven's SPM test.
School Number of
groups Mean Pre-
intervention
score (out of 60)
Mean Post-
intervention
score (out of 60)
Difference
A1 (target) 12 38.92 44.00 5.08
A2 (control) 11 43.00 44.45 1.45
B1 (target) 5 42.40 46.80 4.40
B2 (control) 5 40.20 39.40 -0.80
C1 (target) 6 45.66 47.33 1.66
C2 (control) 9 43.77 46.22 2.40
Total target 23 41.43 45.48 4.05
Total control 25 42.72 44.08 1.36
Overall the scores of the target groups increased more that the scores o the control groups which is
in line with the hypothesis that coaching exploratory talk will improve scores on reasoning tests.
However the control group for school C1 actually increased more than the target school. An
analysis of covariance (with post-test scores as the dependent variable and pre-test scores as the
covariate) indicates that the difference between the gains of the target and control groups was not
significant (F= 2.43; one-tail p = 0.06).
3. Evaluating changes in individual performance
Table 6 gives the increase in individual test scores for both target and control classes. An analysis
of covariation (again with post-test scores as the dependent variable and pre-test scores as
covariate) revealed that the gains made by the individual target class children were significantly
greater than those made by children in control classes (F= 3.141; one-tail p = 0.04.)
Table 6 Individual scores on the Raven's CPM test.
School Number of
children Mean Pre-
intervention
score
(out of 36)
Mean Post-
intervention
score
(out of 36)
Difference
A1 (target) 35 28.91 31.31 2.40
A2 (control) 31 31.39 32.89 1.00
B1 (target) 15 32.73 34.53 1.80
B2 (control) 14 31.28 32.21 0.93
C1 (target) 10 33.10 34.70 1.60
C2 (control) 19 32.86 33.95 1.09
Total target 60 30.56 32.68 2.11
Total control 64 31.87 32.89 1.01
Discussion
It might seem strange that the individual increases in reasoning test scores should prove significant
when the group increases were not significant. The socio-cultural model of learning reasoning
which we are exploring would predict that the teaching exploratory talk should have its first, more
direct, impact on group reasoning, and then secondarily on individual reasoning. However the
design of our study meant that there were approximately three times as many individuals as groups
and this probably explains why the group results did not reach statistical significance while the
individual results did. Although the increase in group scores was not significant there is some
evidence to support the hypothesis. In the two schools where our observations suggested that the
TRAC programme had been most carefully and comprehensively carried out (Schools A1 and B1),
group scores on the Raven's test increased by over 10%.
The purpose of this study was not to show that it is possible to increase Raven's test scores
through a training programme in non-verbal reasoning skills. This possibility is well known and has
been shown by other studies (for example Riding & Powell, 1985). The difference with our study is
that the training programme focused entirely on language use, teaching the ground-rules of
exploratory talk without the use of materials or exercises which were in any way similar to a
Raven's test. Children in the target classes had no more experience of doing the Raven' test, or
any similar kinds of non-verbal problem solving, than those in the control classes. The individual
Raven's tests were carried out three days after the group Raven's tests in all cases: yet their
individual test results improved more, to a statistically significant extent.
It has been suggested to us (by other researchers) that the statistically significant results of the
target classes may reflect the appropriation by target children of strategies for solving non-verbal
problems used in the group Raven's tests, rather than simply the appropriation of 'dialogic' ways of
reasoning from the use of exploratory talk. However, even if this is the case, it does not necessarily
undermine the support for our hypothesis. We have demonstrated that the strategies used by
target children to share and account for problem-solving strategies in the group Raven's test did
reflect the influence of the teaching programme. If individuals in the target classes were able to
appropriate task specific strategies more effectively than those in the control classes, this is
therefore likely to be because of the more effective way that they jointly constructed knowledge
when they talked together around the task. Whether the causation was direct or indirect, the results
of the experimental study support the view that the statistically significant improvements in the
individual test results were a result of our language-based intervention.
The TRAC intervention programme worked well in the school where it originated. Here it was
implemented by an enthusiastic teacher-researcher who was committed to the programme and to
the study. The transfer to two further schools proved problematic and raises issues for future
research. If we judge by the group test scores (Table 3), the intervention programme appears to
have had an impact on group process in school B while there is little evidence of an impact in
school C. This conclusion was supported by our field notes and is also supported the indicators of
exploratory talk (Table 2). This difference may be accounted for by the different motivation of the
teachers or by the different pedagogic strategies they adopted. A planned follow-up study will focus
more on teacher motivation and on the pedagogy. However another possibility is also suggested by
the fact that the schools involved varied in the socio-economic background of their intake. School C
and its and control school serve a catchment area of predominantly affluent, middle class
communities, while the other schools serve populations of lower average income where rates of
unemployment are high. One possible explanation for the failure of the intervention programme in
School C1 is that this kind of coaching has the most positive effect on children's joint attempts to
solve the Raven's problems when the children involved are from lower socio-economic groups.
That is, it is possible that the middle class children were already more familiar with the 'ground
rules' of exploratory talk than those from the lower income families. This is supported by the
observation that a number of individuals in Schools A1 and B1 who got a very low result on the pre-
intervention test improved dramatically after coaching in exploratory talk suggesting that their initial
failure was due to the lack of quite simple strategies which the intervention course was able to
provide them with. Although this was an unexpected result it conforms to a large literature on the
impact of socio-economic class on language acquisition and academic achievement, particularly
the work of Bernstein (1992; 1996).
We have concentrated in this paper on the particular effects of the intervention on children's talk
and reasoning while doing the Raven's test. However we were also concerned to explore the
impact of teaching exploratory talk on the way children work together and reason together on tasks
directly related to the school curriculum. To this end we investigated the impact of the TRAC
intervention programme on children's talk and learning in classroom activities in the area of science
and of citizenship. Here we found a similar improvement, from an educational point of view, in the
quality of their interaction and their collaborative learning. These findings have been published
separately (Wegerif, Mercer & Dawes, 1998).
Conclusion
The four main findings of the study are that children's use of exploratory talk can be increased
through teaching, that the programme to teach exploratory talk can transfer from its place of origin
to other schools (although this transfer can be difficult to achieve); that exploratory talk helps groups
to reason together, and that individual scores on a standard non-verbal reasoning test can be
significantly increased by teaching children to use exploratory talk.
We began this paper with the outline of a possible socio-cultural model of learning to reason as
induction into the dialogical practice of reasoning together. The study described was designed
partly to explore the extent to which this model of learning to reason could provide an account of
general cognitive development. The results of our study indicate that, at least in part, the kind of
reasoning ability involved in individual non-verbal reasoning tests is mediated by social interaction.
This result is interesting because non-verbal reasoning tests have been taken by some to be the
paradigmatic context for the assessment of individual reasoning ability independent of culture and
language use. However the exact nature of any link between teaching the use of exploratory talk
and individual test results needs further elucidation. One plausible hypothesis is that the group
experience in the target classes prior to the individual post-intervention testing, in which knowledge
was much more explicitly shared and accounted for than in the control classes, helped target class
children appropriate specific problem-solving skills (such as considering alternative options before
choosing an answer) and develop a greater awareness in general of the nature of the task.
However, this interpretation does not undermine our socio-cultural claims of a relationship between
the 'intermental' and the 'intramental' in learning and development, as it still supports the view that
reasoning skills may be developed through social 'collective thinking' activity. The findings of the
study are compatible with the socio-cultural position, that success at any cognitive task is a situated
achievement in which many contextual factors play a part.
In the introduction Gauvain was quoted calling for the need to find routes to reconcile what she
called 'traditional views of development' with a more culturally based approach. The findings of our
study give modest support to the idea that we should place a traditional account of cognitive
development within a historical and cultural context in which this account is seen, at least in part, to
reflect a process of induction into certain valued ways of using language. The study was not
detailed enough, and its findings were not unambiguous enough, for us to claim that they either
confirm or deny the strong hypothesis that learning and using dialogical reasoning is the origin of
what is described by some as the process of general cognitive development. However pursuing
aspects of this hypothesis led us to findings which are highly relevant to these issues, and also led
to the development of new methods which may be used in the future to explore further the
relationship between language use and cognitive development.
Acknowledgements
We are grateful for the support of the Economic and Social Research Council (ref. R000221868),
for the active cooperation of teachers in Milton Keynes, for Karen Littleton's contributions to the
project and for Richard Joiner's help with the analysis of the data.
References
Barnes, D., & Todd, F. (1978). Communication and learning in small groups. London:
Routledge & Kegan Paul.
BenHabib, S. (1992). Situating the self: gender, community and postmodernism in
contemporary ethics. Cambridge: Polity Press.
Bernstein, B. (1996). Pedagogy, symbolic control and identity. London: Taylor and Francis.
Burbules, N., & Rice, S. (1991). Dialogue across differences: continuing the conversation.
Harvard Educational Review, 61(4), 393-416.
Carpenter, P., Just, M., & Shell, P. (1990). What one intelligence test measures: a
theoretical account of the processing of the Raven Progressive matrices test. Psychological
Review, 9(7), 404-431.
Dawes, L. (1997). Teaching talk. In R. Wegerif & P. Scrimshaw (Eds.), Computers and talk
in the primary classroom, (pp. 189-196). Clevedon: Multilingual Matters.
Edwards, D., & Mercer, N. (1987). Common knowledge: The development of understanding
in the classroom. London: Methuen/Routledge.
Habermas, J. (1990). Moral consciousness and communicative action. Cambridge: Polity
Press.
Habermas, J. (1991). The theory of communicative action. (Vol. 1). Cambridge: Polity
Press.
Habermas, J. (1992). The unity of reason in the diversity of its voices, Postmetaphysical
thinking, (pp. 67-94). Cambridge: Polity Press
Heath, S. B. (1983). Ways with words: Language, life and work in communities and
classrooms. Cambridge: Cambridge University Press.
Howe, C., Tolmie, A., Anderson, A., & Mackenzie, M. (1992). Conceptual knowledge in
physics: The role of group interaction in computer-supported teaching. Learning and Instruction,
2(2), 161-183.
Kruger, A. (1993). Peer collaboration: conflict, cooperation or both? Social Development,
2(3), 165-185.
Leont'ev, A. N. (1981). Problems of the development of mind. Moscow: Progress
Publishers.
Mercer, N. (1995). The guided construction of knowledge: talk amongst teachers and
learners. Clevedon: Multilingual Matters.
Mercer, N. (1996). Sociocultural perspectives and the study of classroom discourse. In C.
Coll & D. Edwards (Eds.), Discourse and learning in the classroom, (pp. 13-23). Madrid: Infancia
and Aprendizaje.
Mercer, N. & Fisher, E. (1993). How do teachers help children to learn? An analysis of
teachers' interventions in computer-based activities. Learning and Instruction, 2(4), 339-355.
Newman, D., Griffin, P., Cole, M. (1989). The construction zone: Working for cognitive
change in school. Cambridge: Cambridge University Press.
Raven, J., Court, J., & Raven, J. C. (1995). Manual for raven's progressive matrices and
vocabulary scales. Section 2: Coloured Progressive Matrices. (1993 ed.). Oxford: Oxford
Psychologists Press.
Raven, J., Raven, J. C., & Court, J. (1995). Manual for raven's progressive matrices and
vocabulary scales. Section 1: General Overview. (1995 ed.). Oxford: Oxford Psychologists Press.
Resnick, L., Levine, J., & Teasley, S. (Eds.). (1991). Perspectives on socially shared
cognition. Washington D.C: American Psychological Association.
Resnick, L., Salmon, M., Zeitz, C., Wathen, S., & Holowchak, M. (1993). Reasoning in
Conversation. Cognition and Instruction, 11(3&4), 347-364.
Richardson, K. (1991). Reasoning with Raven - in and out of context. British Journal of
Educational Psychology, 61(2), 129-138.
Riding, R. J. & Powell., S. D. (1985). The facilitation of thinking skills in pre-school children
using computer presented activities. Educational Psychology, 5( 2), 114-125.
Rogoff, B. (1990). Apprenticeship in thinking: Cognitive development in social context.
Oxford: Oxford University Press.
Rogoff, B., Gauvain, G., & Ellis, C. (1991). Development viewed in its cultural context. In P.
Light, A. Sheldon, & M. Woodhead (Eds.), Learning to think, (pp. 292-339). London:
Routledge/Open University.
Rojas-Drummond, S., Hernandez, G., Velez, M., & Villagran, G. (1998). Cooperative
learning and the appropriation of procedural knowledge by primary school children. Learning and
Instruction, 8(1), 37 - 63.
Rommetveit, R. (1992). Outlines of a dialogically based social-cognitive approach to human
cognition and communication. In A. Wold (Ed.), The dialogical alternative: Towards a theory of
language and mind, (pp. 19 - 45). Oslo: Scandanavian Press.
Rorty, R. (1991). Objectivity, relativism, and truth: Philosophical papers. (Vol. 1). Cambridge:
Cambridge University Press.
Vygotsky, L. (1991). The Genesis of higher mental functions. In P. Light, S. Sheldon, & M.
Woodhead (Eds.), Learning to think, (pp. 34-63). London: Routledge.
Wegerif, R. (1996a, September). Reason and difference: a dialogical alternative to Piaget
and Vygotsky. Paper presented at The 7th Socio-Cultural Conference: Piaget and Vygotsky
Centenary, Geneva.
Wegerif, R. (1996b). Using computers to help coach exploratory talk across the curriculum.
In M. Kibby & J. Hartley (Eds.), Computer Assisted Learning: selected contributions from the CAL
95 Symposium, . Oxford: Pergamon.
Wegerif, R., & Mercer, N. (1997a). A dialogical framework for researching peer talk. In R.
Wegerif & P. Scrimshaw (Eds.), Computers and talk in the primary classroom, (pp. 49-65).
Clevedon: Multi-lingual Matters.
Wegerif, R., & Mercer, N. (1997b). Using computer-based text analysis to integrate
quantitative and qualitative methods in the investigation of collaborative learning. Language and
Education, 11(3), 271-287.
Wegerif, R., Mercer, N., & Dawes, L. (1998). Software design to support discussion in the
primary classroom. Journal of Computer Assisted Learning, 14(3), 199-211.
Wertsch, J. (Ed.). (1985). Culture, communication and cognition: Vygotskian perspectives.
Cambridge: Cambridge University Press.
Wertsch, J. V. (1991). Voices of the mind. New York: Harvester.
White, S. (1988). The later work of Jurgen Habermas. Cambridge, MA: Harvard University
Press.
