backup: adfs::0.$.chap6

27th October 1993





Early in this book, I noted a contemporary worry that computer-based

learning might evolve into a socially-isolating experience. I have

since argued that the very nature of the technology easily invites

activity that can become dislocated from teacher-led communication.

Moreover, orthodox perspectives on the psychology of learning may even

be encouraging this trend. Yet, when we examine actual classroom

practice we find that, in certain significant respects, the computer

has facilitated socially-organized learning rather than inhibited it.

This is an impression that is particularly apparent in settings of

early education. It is an impression based upon the prevalence of

computer-based *group* work: what is now commonly seen in classrooms

is computers used to support learning through peer collaboration.

The reasons for this development may well include an active attempt by

teachers to resist the isolating atmosphere of computer use. However,

it may also reflect a particular predicament they have faced in the

recent period of implementing new educational technology.

There has been official concern that computers should penetrate deeply

into educational activity, yet the scale of financial commitment

barely made this possible. Generally, the solution improvised by

schools was to spread the resource thinly through a premises and

thereby allow limited access for each pupil. In Britain this strategy

is particularly favoured by primary schools. Secondary schools have

been more inclined to cluster their machines in customised computer

rooms: but this may only have served to increase availability at the

expense of concentrating its use within specialist subject groupings -

technology classes, for instance. In either case, if pupil

opportunities for fair access are to be optimised under these

circumstances of scarce resourcing, then cultivating small group work

(rather than pursuing the model of individual "desktop teachers") may

be viewed as the natural solution.

In this chapter, I shall evaluate the state of practice and research

relating to the deployment of computers for peer-supported learning.

First, I shall review claims that have recently been made for the

quality and consequences of such collaborative work. It will be noted

that this approach to the management of computer-based activities has

attracted some justified enthusiasm. However, I shall argue that the

use of computers for peer-based learning deserves richer input from

psychological theory. Particularly where younger pupils are involved,

we still have only limited insight into what defines an "effective"

collaboration. Rectifying this requires us to develop a suitable

conceptual vocabulary for analysing the talk and action that

constitutes collaborative work. I shall identify the concepts that

are used by researchers to characterise group learning and peer

interaction: my argument will be that these concepts turn out to be

limited in their reach. In the last section of the chapter, I shall

present a perspective on collaborative learning that is consistent

with my preferred theoretical orientation: that of a cultural


In a sense, the chapter converges on a perspective for characterising

collaborative learning and suggests that computer-based work is a good

vehicle for exploring it. The overall argument can then be elaborated

and illustrated through empirical work presented in the chapters that

follow. The recurring question beneath this discussion will be: how

can new technology serve effectively to resource collaborative

arrangements for learning?






It was implied above that the pace and pattern of introducing

computers into schools has served to cultivate a more

socially-organized style of using them. I mean that logistic problems

of optimising access have encouraged their use in group work.

However, the prevailing climate of educational practice may, in any

case, have been favourable towards such developments. The tradition

of small group work is well established in British schools - at least

in the primary sector (cf. Dunne and Bennett, 1990; Galton and

Williamson, 1992). McMahon (1990) has surveyed a number of

educational documents that reveal a strong official endorsement of

collaborative structures (although he observes wryly that, while this

is encouraged for classwork, pupil *assessment* remains fiercely

individual). McMahon goes on to review the potential of new

technology for helping to realize the pressure for more group work.

In surveying primary practice, Jackson, Fletcher and Messer (1986,

1988) report that most teachers have recognized this and are arranging

for small groups of pupils to work together; individual computer work

is rare. The situation seems comparable in other countries. For

example, Sheingold, Kane and Enderwelt (1983) report evaluative

reactions from a group of North American teachers adopting computers:

these teachers were most impressed with outcomes in the social domain

- changes in classroom interaction and in pupil self esteem.

Thus, before we consider research into peer interaction at computers,

we should note that practitioners have identified (independently of

such research) a potential for the support of group work. In

considering the popularity of computer-based group work, we,

therefore, should be careful not to interpret it *merely* in terms of

a pragmatic motive: to make a scarce resource more widely available.

Teachers may be intending to pursue more focussed educational purposes

through their encouragement of such practices.

Often, computer activities have a conspicuously engaging quality

relative to other classroom tasks. Where groups of pupils are

involved, this engagement is usually characterised by lively

conversation. Research reports by practising teachers stress their

particular satisfaction with the language sustained during computer

work (eg., Dickinson, 1986; Genishi, McCollum and Strand, 1988; Hill

and Brown, 1988; Shooter, Lovering and Bellamy, 1993). This simply

complements a long tradition in early education of concern for

developing language use. Language is commonly regarded as more than

just a means of communication - it is also seen as the point at which

meanings get created, and the resource whereby intelligent action

itself may be directed. These ideas are very familiar to teachers,

particularly through the work of educationalists such as Barnes (1976)

and Tough (1977). Insofar as computers release this verbal energy

from pupils, then they seem to be creating the kind of problem solving

environment that many teachers naturally seek. Nevertheless, further

research around this topic is still necessary. We need to see whether

these scattered impressions hold up more generally; we need to be more

specific about defining effective language use in this context and,

finally, we need to determine whether there are additional good

reasons for using computers collaboratively.

The release of language in these situations may be one basis for

promoting this way of working: for, participation in such talk

supports cognitive development. Later, I shall return to this

argument and examine the more detailed reasons for endorsing it.

However, making the case for any kind of collaborative work can, in

addition, include a more ideological motive. It may be based upon our

intuitions about the ideal intellectual atmosphere for learners to

enjoy - simply as participants. So, we may indeed encourage

collaborative work because it supports individual cognitive

development, or because the world beyond school can reward such

skills. But we also might encourage it because we simply believe that

joint intellectual activity is a rewarding dimension of human

experience. However, in some tension with this, both our

psychological theorising and our cultures of educational practice are

still strongly oriented to fostering achievements that are clearly

those of individuals (hence the solitary circumstances of

examinations). So, perhaps for many observers, the most persuasive

basis for having children work collaboratively will always tend to be

this: it leads to gains in what they can do when we revert later to

examining and testing them as *individuals*.

In assessing computers as contexts for joint work we should consider

the two issues I have raised. Firstly, do computers support lively

group work of the kind we want learners to enjoy? Secondly, do these

encounters allow individual participants to gain at least as much

understanding as they would if the same work was individually

organized? I will make some remarks on each of these questions in




Quality of peer collaborations at computers

I have noted above that many teachers describe computers as useful

catalysts for generating pupil talk. However, the quality of these

experiences deserves more systematic documentation. One reason for

caution is evident in reports from those teachers who have themselves

been especially diligent in documenting the content of talk around

computers. Take, for example, a report by Farish (1989). This

teacher admitted to being gratified - at first - with what seemed to

be happening in her own primary class. The computer seemed to engage

her pupils. She then listened more closely to the computer-supported

talk: it turned out to be mainly about uninteresting details

concerning screen control and keyboard control. In fact, it was less

stimulating than talk she recorded in other (non-computer)

collaborative settings. So, there is obviously a danger of being

misled by a gloss of sheer pupil activity.

Farish's technique for observing was unobtrusive. However, the issue

of research methodology defines a further danger associated with many

such classroom studies. That is the danger of drawing optimistic

conclusions from observations where teachers themselves are very

closely involved with the children's activities. For their presence

may, therefore, serve to stimulate more on-task, productive talk than

would otherwise be expected. (Of course, professional researchers are

also vulnerable to creating this kind of distorting influence.)

However, bearing in mind these cautions, there are many favourable

reports from all sectors of education concerning the quality of group

work when it is organized around computers.

An influential early study is described in Hawkins, Sheingold,

Gearhart and Berger (1982). They provide a comprehensive account of a

class of 10-11 year olds making extensive use of Logo: a strong sense

of joint purpose was created and a culture of expertise and support

evolved. This report may have been influential in defining a model of

computer classroom effectiveness towards which many practitioners were

subsequently to aspire. However, the study is also a reminder of the

difficulty of generalizing too freely: the school was private and its

access to computers was generous. Moreover, the intervention was done

at a time when such technology in early education was still a novelty.

Finally, the project was supervised by international experts in the

field. Such circumstances are a long distance from many teacher's

experiences. On the other hand, pupil enthusiasm at least

approximating these levels has been replicated in other, less

glamorous settings.

Energetic computer-based group work has been most thoroughly

documented within the earliest years of education (Dickinson, 1986;

Genishi et al, 1988; Hill and Brown, 1988), and in the middle and

secondary years (Broderick and Trushell, 1985; Gonzalez-Edfelt, 1991;

Johnson, Johnson and Stanne, 1985; Hoyles and Sutherland, 1989; Light,

Foot, Colbourn and McClelland, 1987; Sheingold, Hawkins and Char,

1984; Trowbridge, 1987; Webb, 1984). There is very little relevant

research on this topic that considers further and higher education or

work in training communities. However, in all of the schoole cases,

the group work is described as leading to more effective (group)

products than the same work organized on an individual study basis.

In fact, in a review of cooperative learning and new technology,

Watson (1990) fails to find a single claim that the solitary use of

classroom computers by individual pupils is a preferable

implementation strategy.

This success is not an obvious outcome. Collaborative computer work

might have been vulnerable, in particular, to playful distractions or

to social dominance. Problems of this kind do seem less serious than

has been documented for other collaborative work (Bennett, 1991;

Galton, 1990). However, such problems have been reported. So, there

has been some research attention to the themes of dominance and gender

as troublesome issues in the management of collaborative work at


In relation to gender, Hoyles (1988) has suggested a strong link

between this variable and the circumstances of group versus solitary

working conditions. In particular, girls may feel more positively

towards computer tasks that are organized collaboratively. This

invites closer attention to conditions under which mixed and same sex

groups can be compared. Hughes and Greenhough (1989), Siann and

Mcleod (1986) and Underwood, McCaffrey and Underwood, (1990) all

report different outcomes of paired learning experiences for boys and

girls - depending on the configuration of the pairings they were party

to. Unfortunately, the pattern of these results is not consistent.

However, neither were the observed tasks easily comparable.

Computer-based tasks are very varied in their nature and, thus, they

may vary also in the social relations they can support among users.

Moreover, these differing findings arise from brief research-oriented

interventions carried out in uncertain relation to the mainstream of

classroom life. Pupils' attitudes to such activities (and their

behaviour within them) may well be unstable under the conditions of

such variation. At present, we can only conclude that gender

relations are an issue in forming computer work pairs, but suggest

that consequences for the quality of the work sustained are variable -

depending, at least, upon the nature of the task set.

Regarding dominance, Light and Glachan (1985) report this as a

potential problem within pairs in a situation where limited control of

the computer was possible. This factor may also interact with gender:

(primary school) boys being more likely to take control of an input

device when such access is limited (Blaye, Light, Joiner and Sheldon,

1991). Benyon (1993) makes similar observations in relation to boys

and girls using Logo and comments that teachers may be oblivious to

such details of social interaction - given their necessarily

intermittent contact with it. These observations provide a good

example of the sense in which computers might be conceptualized in

terms of the affordances they offer for collaboration. Here, the

thing itself (the existence of only one mouse control device) exerts a

significant organizing influence on the exchange between the pupils.

The outcome in this case being an occasion upon which effective

collaboration may be undermined. Elsewhere, Light, Foot and Colbourn

(1987) show how this affordance may be modified (a form of dual-key

control) and, moreover, that only under those circumstances does

collaboration on their task show an advantage for the participants

tested later as individuals.

Research of this kind - relating structural features of computer tasks

to the social organizations they support - is all too scarce.

Numerous studies praise the quality of collaboration associated with

computer-based work without identifying how the setting achieves its

effect. This is an important relation to clarify, because enthusiasm

for computer-based work is increasingly expressed in terms of its

collaborative potential. For example, Hoyles and Sutherland (1989)

discuss Logo in these terms: suggesting that its real advantage as a

classroom resource arises from its capacity to sustain social

interaction in a context of mathematical problem solving. This is a

particularly significant observation, as the popular image of Logo has

been dominated by the "tool-to-think-with" theme (Papert, 1980), - not

the theme of social catalyst. Hoyles, Healey and Sutherland (1991)

take these observations a little further by making a structured

comparison between computer and non-computer versions of the same

(maths) tasks. Their results reinforce the computer's status as an

effective setting for pupil discussion. Although, again, it is hard

to extract reasons why; they comment: '..evidently the computer

somehow draws the attention of the pupils and becomes a focus for

discussion' (p. 217). Such observations badly need elaboration.

I shall return to this puzzle of collaborative potency in the next

chapter. So far, in relation to pupils interacting together *at*

computers, we have reviewed evidence showing that this arrangement can

engage them well. This may be satisfying if we believe skills of

discussion and collaboration are valuable, and if we feel that the

experience is itself a stimulating one for those taking part. It

remains here to make a few remarks regarding the *outcome* of such

experiences as judged in terms of later achievements by the

participants tested alone.



Outcomes for individuals participating in collaborative computer work


Arranging for a task to be conducted jointly at a computer may well be

a good strategy for getting the task completed effectively. Indeed,

the overall outcome may be more creative or, by some other criteria,

more successful than alternative working arrangements that might have

been organized. This may be all we seek. Pupils have had a

worthwhile encounter with collaboration: it led to a useful result and

the experience will serve them well on other occasions of joint

working, both in and out of school. Indeed, it represents to them

something of the way a great deal of understanding naturally occurs:

it occurs within the contexts of socially-coordinated activity.

However, teachers may not always make these arrangements only in the

interests of creating this kind of authentic experience. In

particular, teachers will know that, whatever the social circumstances

for making discoveries in the rest of life, schools are places where

progress gets assessed in terms of what *individuals* achieve. In

short, achievement is often judged in tests (exams) that deny those

who take them access to peers as sources of support. We may,

therefore, wish to know whether computer-based group work results in

gains that show up under this criterion.

Simply because an encounter was effective as a collaboration does not

necessitate that participating individuals learned more than they

would have done if working alone. For example, the task may have been

split up among collaborators in ways that denied individuals

experience with all its component parts. At once, we sense that there

will be no simple generalisation arising from the question posed.

Whether collaborative experiences of learning are more effective for

individuals - when we test them as individuals - will doubtless depend

on the task involved and the management strategy of the participants.

Nevertheless, reviews of this area suggest that gains in individual

testing can generally be expected (Dalton, Hannafin and Hooper,

1989; Light and Blaye, 1989; Light and Colbourn, 1987; Rysavy and

Sales, 1991; Trowbridge, 1987). This advantage has been reported

across a wide range of research environments - although much of the

work concerns the 10-13 year age group. In tightly controlled

experimental tasks, Light, Foot, Colbourn and McClelland (1987) and

Blaye et al (1991) report individual gains associated with paired

work. Hughes and Greenhough (1989) report similar gains for Logo

within a small-scale classroom intervention. Johnson, Johnson and

Stanne (1986) describe a more extensive classroom comparison of

individually, competitively and cooperatively organized computer work

in the geography curriculum. Again, significant gains are apparent

for the cooperative condition.

One study that found no advantage for group work (and no disadvantage

either) makes an interesting peripheral observation. Light, Colbourn

and Smith (1987) found no difference in individual 11-year old's

PROLOG programming achievements when they compared pupils who

had been working either singly at a computer, as a pair, or in a group

of four. However, videotapes of the experimental sessions revealed

that the children working "alone" had engaged in a good deal of

discussion with their neighbours. This subversion of the experimental

procedure may have been disheartening to the researchers but it surely

bears useful witness to how easily discussion is supported among

children mutually engaged in computer work. Researchers with plans in

this area might beware that suppressing such inclinations in the

interest of experimental design may lead to the creation of

unrepresentative models of classroom possibilities. Pupils'

spontaneous appeal to peers as resources of support in computer-based

problem solving is a striking tendency: for it stands in contrast with

pupils' apparent unwillingness to make use of on-line help facilities

that instructional programs themselves often offer (Messer, Jackson,

and Mohamedali, 1987).



Issues arising from studies of computer-based collaboration

Integrating the above remarks with those made in the previous chapter,

there are now grounds for both pessimism and optimism regarding

technology's place in any educational culture that prioritises social

processes. On the one hand, it has been argued (see Chapter 5) that

work with computers is vulnerable to exclusion from the communal

discourse that gets organized in classrooms. So, computer activities

that easily drift towards the periphery of classroom life may have

their impact reduced by this exclusion. This may leave us

pessimistic, because such talk is the vehicle whereby teachers can

create continuity of experience for pupils. On the other hand, the

material reviewed in this Chapter indicates that computers may

re-enter the social life of classrooms via their support of

collaborative work between pupils themselves. This may give cause for

optimism among those educationalists who are concerned to locate this

teaching technology in some social context.

Of course, any trend towards involving computers in peer

collaborations may not, at a stroke, neutralize all our pessimism

arising from the easy marginalization of classroom technology. Work

that children do together around this technology may well be lively

and productive at the time: yet (just like solitary computer work) it

may remain work that is never properly consolidated through

incorporation into classroom discourse. In terms of the concerns

raised in the last chapter, it may become work that still does not get

knitted into the common knowledge of classroom activity.

Those arguments need not be rehearsed again here. Collaborative

computer work is as vulnerable as solitary computer work, in respect

of becoming dislocated from a mainstream of teacher-led talk. Whether

or not this is its fate, we may still ask what formats for a

collaborative task make the experience as rich as possible for the

participants at the time. I believe pursuit of this question brings

us back to further considering the concept of intersubjectivity.

Earlier, in discussing teachers' creation of shared understandings, I

appealed to this concept. I argued that the possibility of creating a

shared cognitive context depends upon the participants' mutual

appropriation of motives, intentions and understandings. So, teachers

must become skilled at saying and doing things that precipitate and

exploit such possibilities: they thereby establish resources of common

knowledge and can build further upon them. This all reflects a human

capacity for intersubjectivity. Here (and in the next chapter), I

will develop an argument claiming that the same capacity should be

recognized as central to what takes place within peer interaction - at

least, when it is going well.

So, at a very general level, I am concerned to reinforce a useful

theoretical link: identifying socio-cognitive processes that are

common to both the interactions between teachers and pupils (see

Chapter 5) and the interactions between pupils and their peers.

Discussion of both arenas of activity can usefully call upon the same

theoretical constructs; in particular, the constructs of common

knowledge, socially-shared cognition and intersubjectivity. All of

this may serve, indirectly, to endorse peer interaction as a

potentially powerful educational resource.

I am also concerned to explore how computers may be mobilized to

become part of such a resource. Yet, given the favourable review of

computer-based collaborative work above, what more is there to be

done to encourage this? To recap: computers have been discovered by

practitioners to create lively forums of joint activity. Close

observation suggests that usually - not inevitably - these

interactions are productive and lead to (group) products that are

superior to whatever may be created by the participants alone.

Finally, the individual participants in collaborative learning tend

later to perform better on relevant tests of achievement - as compared

with peers who learned on their computers in isolation. Surely, the

task ahead is merely to investigate the parameters of this advantage -

investigating the relative potency of different software, and so


Superficially this may describe the proper research agenda. However,

if "merely" investigating the outcomes of collaborative computer work

means doing so in a piecemeal, atheoretical manner, then an

opportunity is being lost. Effective investigation of software should

converge on accounts that can systematise those specific features of

computer-based tasks that afford productive social interaction.

However, what do we understand by "productive" as applied to peer

collaborations? We might answer that question solely in relation to

the "products" of individual testing: a productive collaboration is

one that leads the individual participants to learn more - as assessed

in solitary posttests. But such pragmatic accounts avoid any

concentration on the dynamic of the collaborative act itself. We are

left with shopping lists of experimental variables that co-relate task

characteristics with collaborative productivity. With this strategy,

understanding the processes that mediate the success gets neglected.

Instead our research should aim to develop a richer *theory* of

collaboratively-organized problem solving. They should allow us to

see more clearly why some particular task "works" as joint activity.

In pursuit of this aim, I shall do a number of things below. First, I

shall consider the status of peer interaction within educational

practice very generally - with no special emphasis on technology or

psychological interpretations. Secondly, I shall review the most

influential accounts of collaborative learning that refer to mediating

psychological processes. Thirdly, I shall outline a socio-cultural

critique of this contemporary framework. This leads me to formulate

an account of learning through peer interaction that stresses the

creation of socially-shared knowledge. In the next Chapter, empirical

observations of my own will be deployed to illustrate how

computer-based collaborative learning reveals the dynamic of such







In Britain, there is little need to make new arguments prioritising

group work in classrooms. This style of working is already well

established. As far as primary practice in concerned, the principle

of such work is often associated with prescriptions contained within

the influential Plowden report on early education (Plowden Report,

1967). This official view is endorsed elsewhere: McMahon (1990) has

reviewed policy documents within British education that serve to

legitimize and encourage early group work.

The historical basis for an emphasis on group work in classrooms has

been discussed by Galton and Williamson (1992). In their view, the

origin of this commitment among British teachers had a lot to do with

a prior belief in mixed ability classrooms. Such beliefs were, in

turn, stimulated by the breakdown of a two-tier secondary education

with its associated selection testing at 11. The principle that

children would learn effectively in mixed ability settings was widely

adopted at this time. Galton and Williamson's point is that active

cultivation of group work may not, at first, have been grounded in

particular cognitive theory about its benefits.

It seems unlikely that structural changes in the education

system were the sole stimulus for fostering group work practices.

After all, the idea is surely in some tension with common sense

conceptions of learning. Such conceptions might suppose that learning

is properly managed by the official, teacher-dominated talk of lessons

(Speier, 1976). Moreover, this common sense belief in the central

role of teachers arises not just because they know more facts about

the world. Their greater procedural knowledge must also be of

advantage to learners. This is supported by research: for, even in a

content-free activity (such as planning a route), there is still an

advantage afforded children by collaborating with an adult versus

collaborating with a peer (Radzisewska and Rogoff, 1991). In the face

of this common sense, it is likely that a contemporary recognition of

learning through peer interaction must have arisen partly through

influences from educational or psychological theory.

The work of Piaget (perhaps promoted by Plowden) was a likely source

of inspiration. Here was an account of cognitive development that lay

great stress on the the child's capacity for spontaneous, *untutored*

discovery. This naturally leads to recognizing that circumstance

often refered to by the influential educationalist Frank Smith (eg.,

Smith, 1981): children are "learning" all the time, even when not

engaged with teachers or other experienced adults. Insofar as they

are immersed in the peer community of the classroom, that setting will

furnish experience of interaction that supports learning also. In

fact, given the natural social energy released within children's peer

cultures, it seems sensible to mobilise this in the service of

educational aims - to regard peers as a positive source of support

(Sieber, 1979). This should only require practitioners to evolve some

expertise regarding the management of such encounters: what sorts of

tasks are well suited to joint cognitive activity and what

configurations of individuals make the best of the opportunities.

Yet, the situation is not so straightforward. There may be a natural

social energy to children's' early exchanges, but channelling it into

the agendas of school learning may prove a more than trivial task.

So despite children's sociability, despite the good sense of

incorporating group work, and despite the pressure to do so, the

reality of how it typically proceeds is somewhat sobering. There have

been several substantial observational studies of group work in

primary schools. They have been reviewed by Galton (1990; Galton and

Williamson, 1992) and by Bennett (1991). The summary picture these

reviews present is a disappointing one. There is much investment in

creating the material conditions for group work, but there is less

evidence that these conditions support productive, on-task joint

activity. In other words, the seating has been socialized into

collaboration but the pupils have not. Bennett (1991) comments: 'What

seems to have happened in practice is that teachers have taken on

board the Plowden Report's views on having children work in groups,

but have preferred to retain individualization rather than cooperation

in that context' (p. 585).

This conclusion should not imply that teachers have subverted the

educational plan by underpinning the surface organization with an

alternative agenda of their own - as if grouping arrangements were

then merely cosmetic gestures. It may be that teachers pursue the

goal of group work but they enjoy very little guidance on how it is

most effectively energised. This point is made in more detail by

Perret-Clermont and Schubaeur-Leoni (1981). Or the problem may be

different again: where younger children are concerned, the aim of

fostering collaborative practices may simply be too ambitious. This

might be so if Kutnick (1983) is right when describing the infant

school peer group in his observations as 'largely an ineffectual

gathering of pupils greatly dependent on the parents and the teacher'

(p. 49).

We already have reason to doubt the generality of this dispiriting

observation about young children as collaborators: earlier I

cited many reports of such children in lively joint activity with

computer tasks. The other suggestion above, that teachers are poorly

prepared to manage joint work, may be credible; although one would

expect the necessary practical insights soon to emerge through

immersion in classroom life - if there are any practical insights to be

had. I believe that the gap between expectation and reality for group

work has more to do with our failing to characterise the optimal

setting for joint problem solving. We need more sensitivity to the

kinds of tasks that are particularly potent in supporting good

collaborative interaction. At the moment, the ethnographic literature

on classroom group work warns us that such tasks may be scarce and not

in widespread use - or that the teacher's job of coordinating them is

more demanding than is realized. Although the problem may

superficially be one of achieving this fit between tasks and social

processes, our limited success in getting it right probably reflects

the lack of a persuasive theoretical account of the nature of

collaboration itself. Thus the strengthening of such theory will be

one useful achievement. It may help us notice the things that are

making a difference, or help us actively tease them out.

Naturally, I am going to suggest that a way forward does lie in

considering computers in relation to this problem. Computers may

furnish good settings for meeting the goals of effective small group

work. This possibility has already been highlighted above. However,

the simple input-outcome nature of that research may be too modest a

basis for trying to influence practice. Therefore, my aim now is to

make better sense of why computers should be effective in this way. I

believe this aim requires us to reflect, first, on the psychological

processes active in collaborations that are effective. We may then

clarify how computer-based activities manifest properties that are

suited to prompting and sustaining those processes. With these goals

in view, I turn next to consider developmental research on processes

underlying any peer-based problem solving exercise.






There are already some detailed surveys of this topic in the

literature (eg., Azmitia and Perlmutter, 1989; Cazden, 1986; Damon and

Phelps, 1989; Forman and Kraker, 1985). Therefore, I will not attempt

a comprehensive review of all empirical research that relates to the

various themes arising. However, I will abstract the important

distinctions that characterise understanding in this field, and I will

offer pointers to research that supports the main conclusions.

So far, I have been slipping between reference to "group work" and

to "collaborations" fairly freely. However, researchers in this area

have adopted informal distinctions that identify more precisely the

different kinds of group working that might be arranged. In

particular, the distinction is commonly made between interactions that

are "tutorial", "cooperative" or "collaborative". We shall be mainly

interested in the last of these, but brief mention of what the other

two involve will be useful.



Group work as peer tutoring or as cooperative learning

We might say casually that what characterises interactions in small

group work is learning rather than teaching. It is usually a

different kind of class experience than any that are managed by a

teacher. Yet, peer interactions can be organized to have a tutorial

quality. A peer can be assigned to act the role of tutor. This might

be arranged simply to free up the "official" teacher. Peer tutors may

indeed be effective because they achieve with their classmates

something of what grown-up teachers achieve. But, in addition, the

tutoring experience may be educationally valuable because of effects

that are advantageous for the tutors themselves. To take the role of

teacher may be a useful learning experience. This is a point that

arises in a number of research reviews relating to such classroom

initiatives (eg., Topping, 1992).

What is involved in the activity of teaching may include a style of

talking that is both specialised and demanding of interpersonal

sensitivity. These skills may be cultivated over a long period and

not within the easy grasp of younger children. Yet, it might be to

the advantage of young children to gain some reflective access to what

is involved. In this spirit, Palinscar and Brown (1984) have reported

a productive exercise based upon allowing learners to take control of

teaching styles of dialogue. However, this format for group work is

important but of less interest here: it is not the circumstance under

which most small group work is usually managed in classrooms.

Cooperative learning is a more typical arrangement than peer

tutoring. It refers to strategies for managing tasks that often

implicate a large group working together or even a whole class. It is

common for tasks in this tradition to be partitioned such that

different members of a group take responsibility for different

components of the task. Johnson and Johnson have reported a number of

initiatives that pay close attention to strategies for achieving

optimal organizations (eg. Johnson and Johnson, 1985). Another

concern of research in this area is motivational: how to define a

structure of rewards that sustains joint activity. Slavin and

colleagues have been particularly interested in this problem (eg.,

Slavin, 1986).

Nastasi and Clements (1991) have reviewed research in the cooperative

learning tradition and highlighted implications for classroom

practice. At first sight, such work seems relevant to our interest in

assessing computers as vehicles for peer-centred work. It does

address the issue of motivation: something that might prove important

to our understanding of effective computer use. However, the analysis

is typically at too molar a level. It is particularly concerned with

how reward structures can be defined and negotiated in relation to the

outcomes of work. Computers can be fitted into classrooms that pursue

cooperative work regimes and there may be good reasons to do so -

given the success that has been reported for this method (eg. Slavin,

1987). However, studies of these regimes do not focus down on the

interactions that are supported at the point of actually using the

technology. The grain with which cooperative classrooms are studied

in too coarse to reveal distinctive effects of computers on social

interaction. I am more concerned to capture these "local" conditions

of social exchange during peer-based work. The tradition of research

that has attended to such matters is normally termed "collaborative

learning" research.


Group work as collaborative learning

The line is thin between cooperative and collaborative learning but

one characteristic of research in the "collaborative" tradition is a

greater interest in cognitive processes - as against motivational

ones. At least, Slavin (1987) draws the distinction in these terms,

arguing that the two research traditions often talk past each other

although they are really complementary. Cooperative learning studies

help define a motivational and organizational structure for an

overarching programme of group work; while collaborative learning

studies focus on the cognitive advantages that arise within the more

intimate exchanges of working together.

Here I really want to relate to the concerns of both research

orientations. On the one hand, computer-based work does have a

focused, intense quality and it invites a research technique that

examines pupil interactions closely and in relation to aspects of the

task. This is the strategy of the cognitively-oriented collaborative

learning tradition. On the other hand, I believe there *is* a

motivational issue to be addressed in relation to work at computers,

but it is not one that is easily explored (or even revealed) within

the research strategy favoured in studies of cooperative learning.

The present discussion will converge on this dilemma.

Part of my present concern is to consolidate the case for making peer

collaborations a significant basis for organizing computer-based

interactions. I have argued that this requires us to look quite

closely at the detail of interactions that such activities afford. We

may thereby identify the distinctive strengths of computers deployed

in this manner. Research on the nature of peer interaction during

focussed problem solving has a cognitive emphasis. It identifies a

number of processes that might be powerful when released in this

social situation: I believe there are three that have been well

documented. For shorthand, I shall refer to them as processes

relating to (1) articulation, (2) conflict, and (3) co-construction.

(1) Articulation. One way in which learners might gain from working

closely on a problem with a peer is by being required to make their

thinking public and explicit. The learner's opinions, predictions,

interpretations and so on will need to be articulated for the benefit

of joint activity. These declarations may be a product of problem

solving that inherently supports the learning process. This simple

requirement defines a possible value in collaborative work that has

been identified by a number of researchers (Chi and Van Lehn, 1991;

Hoyles, 1985a; Schunk, 1986). Damon and Phelps (1989) summarize the

advantages as follows:

In order to work productively with their partners, children must

publically recapitulate their own emerging understanding of the task.

This, we believe, is a process that strongly facilitates intellectual

growth, because it forces the subjects to bring to consciousness the

ideas that they are just beginning to grasp intuitively. The

responsibility that children feel for communicating well with their

peer partners induces them to gain greater conceptual clarity for

themselves (p. 152).

This summary seems to locate the strength of peer work in

self-reflective processes arising from the responsibility of

justifying and declaring your own ideas to a collaborator. The idea

is captured elsewhere in comments made by Bruner on the

general role of language in supporting learning:

It must express stance and counter-stance and in the process leave

place for reflection, for metacognition. It is this that permits one

to reach higher ground, this process of objectifying in language or

image what one has thought and then turning around on it and

reconsidering it. (1986, P.129)

However, there may be a further advantage within situations where

ideas are publically articulated: one that reflects the more

interactive (rather than declarative) quality of joint activity. That

is, it arises from being witness to your own partner engaging in this

articulation of ideas. The advantage here involves something more

dynamic in a properly social sense. It may arise insofar as the talk

of one participant serves to create for the other exemplars of

strategic moves that comprises effective problem solving.

With this suggestion, a link is made to one interpretation of

effective practice in the zone of proximal development. In Chapter 4,

the view was presented that some instructional talk worked its effect

by virtue of creating a cognitive system. A more expert partner would

be particularly suitable for prompting, elaborating and filling in the

problem solving process, thereby allowing the novice to achieve a goal

unattainable in a solitary setting. Vygotsky's own account of ZPD

configurations referred to novices with 'adult guidance or in

collaboration with a more capable peer' (Vygotsky, 1978, p. 86).

However, there is no reason why symmetrical peer pairings should not

sometimes give rise to a socially-defined cognitive system of the same

sort: one that is comparable to that traditionally discussed for

novices working with more expert partners. As before, the cognitive

benefits arising from this would be associated with processes of

internalization. The opportunity to participate in the processes of

coordinating a problem solving strategy creates the conditions for

transfer from Vygotsky's inter-mental social plane to his private

intra-mental plane.

So, it is being argued that "simply" articulating ideas is a useful

experience for problem solvers and one naturally afforded by

socially-organized tasks. It is suspected that benefits would arise

because of the pressure for reflection that is felt by the individual

talking. It is also supposed that making ideas public may allow them

to slot into an externally-located cognitive system that implicates a

partner's contributions also. Participation in such systems may then

be internalised. Does empirical work suggest that experiences of this

kind carry forward?

There are various classroom observations of group work (particularly

involving computers) that draw attention to episodes of publically

declared reasoning (eg., Hoyles and Sutherland, 1989). However, the

impact of such experiences on the participant's problem solving in

subsequent situations (perhaps alone) is harder to demonstrate. A

significant obstacle is the difficulty of systematising and

quantifying such dialogue so that the relevant correlations might be

explored. There are some studies that attempt this on a modest scale.

For instance, there are studies of children's interactions that

characterise and enumerate discrete problem solving moves (questions,

challenges, requests etc.) and then relate their occurrence with the

participants' subsequent performance on other, related problems.

Positive correlations have been reported (eg., King, 1989; Webb, Ender

and Lewis, 1986). The inference is that participation within this

kind of problem solving discourse was indeed a valuable learning

experience - perhaps for the reasons implied above. Where there is

lots of reasoned discussion there will be benefit that carries

forward. Unfortunately, it is in the nature of correlational data

that we will be left uncertain as to the precise basis of such

performance gains. These studies must encourage us in fostering

problem solving encounters in which thinking gets publically

articulated, but exactly how the benefit of this experience is

mediated remains to be demonstrated.

(2) Conflict. There is nothing strongly interactional in the above

conceptualization. The presence of a peer doubtless encourages a

pupil to articulate her ideas, but the benefits are not claimed to

arise from dialogue - from coordinated *exchanges* with that partner.

The second process to be mentioned here, conflict, does arise within

such mutual involvement.

In this case, the benefit is supposed to occur in the context of

disagreements between peers and their efforts to resolve them. It

is a convention of conversation that disagreement should prompt

discursive moves of justification and negotiation. So, the cognitive

consequences of conflict might be quite productive. This is a claim

most closely associated with Piaget. It is a claim at odds with the

common belief that Piaget denied the potency of interpersonal

experience in cognitive development. What Piaget draws attention to

within peer exchanges is the power of argument, or 'the shock of our

thought coming into contact with that of others' (Piaget, 1928, p.

204). Piaget supposed that the natural egocentrism of early

childhood thinking typically prevented younger children from

reflecting on their own thought. However, demands for justification

and accountability that will arise within peer disagreements forces

them to enter into such reflection. Moreover, as Piaget claimed,

simply encountering these differences (during argument) seems adequate

to precipitate cognitive restructuring - further processes of

negotiating an agreed consensus need not be part of what must occur

(Howe, Tolmie and Rodgers, 1990).

Piaget's proposal has led to a significant body of research exploring

the conditions under which conflict might effectively induce cognitive

change (for reviews see: Bell, Grossen and Perret-Clermont, 1985;

Doise, 1985; Doise and Mugny, 1984). Much of this work has employed

tasks typical of those used within Piagetian tests of cognitive

development. Particularly popular with researchers have been the

conservation problems: tasks designed for evaluating transition into

the concrete operational stage of cognitive development. These have

been a natural choice where the aim has been to demonstrate how peer

conflict can induce cognitive change - in terms of Piaget's own

conception of cognitive structures. However, such brief and

circumscribed problem solving encounters offer a poor representation

of interaction as it might occur within sustained classwork. Indeed,

Doise (1985) comments on how little communication often takes place in

these experimental tasks. However, one study involving a somewhat

more elaborate, puzzle-like task (Light and Glachan, 1985) does

provide us with a more convincing model for classroom activities.

Here it was demonstrated that the degree of conflict arising within a

peer pairing was positively correlated with performance later on the

task by individuals acting alone.

Conflict has, therefore, been demonstrated to be a distinctive and

productive feature of peer-based problem solving. But the extensive

research attention accorded this feature of peer interaction largely

reflects the great interest in cognitive stage transition within

Piagetian theory. It is less a reflection of the relative importance

or frequency of conflict in classroom group work. There will surely

be other characteristics of effective peer-based learning interactions

which do not implicate conflict.

(3) Co-construction. A final basis for conceptualizing the cognitive

benefit of collaborative learning lies in what is termed

co-construction. This notion often arises in discussions of peer

interaction more influenced by Vygotsky's socio-cultural thinking.

There may be some feeling that the Piagetian emphasis on conflict

suggests the inevitability of social *tension* within joint problem

solving. We may feel uneasy about this and, indeed, there are

examples of research showing that conflict is far from a necessary

ingredient of effective group work. For example, Blaye et al (1991)

observed children tackling one particular set of problems: in this

case, they found no correlation between frequency of children's

conflict and the quality of their performance in subsequent individual

performance tests.

Thus, some authors have analysed peer-based problem solving in terms

of social processes that do not simply involve adopting declarative

attitudes ("articulation" above), neither do they involve more

adversarial styles of interaction ("conflict" above). Forman and her

colleagues have reported empirical observations that illustrate a

co-constructional alternative (eg. Forman, 1989; Forman and Cazden,

1985; Forman and Kraker, 1985). They dwell in particular on how

children may take individual responsibility for complementary

cognitive functions while solving a problem. This may be

organized within the context of some overall converging discussion

about the task. Where Forman and others seem to see joint problem

solving as proceeding well is where the work gets creatively

*dispersed*: strategies of sharing responsibility serve to accelerate

the participants' joint construction of some worthwhile convergence:

a common object of some sort.

Reference to common cognitive objects echoes theoretical concepts

introduced in the previous chapter to help discuss classroom discourse

managed by teachers. In that context, I dwelt upon activities that

assisted the creation of socially-shared cognition. The two sets of

ideas are certainly in the same spirit of theorising. However, when

students of peer interaction identify an object as "co-constructed",

typically that object will be rather narrowly propositional in

character. Such objects often turn out to be procedures or

hypothesis, or other formalized elements in the structural description

of problem solving. They are discrete ingredients in the sequence of

problem solving moves. Social constructions of this sort are surely

valid and useful outcomes of joint activity. However, while I do wish

to endorse reference to jointly-constructed cognitive objects, I feel

that such constructs need to be more richly articulated than those

illustrated by the present line of theorising. In particular, they

will be more fundamental to the support of problem solving talk than

are the particular conclusions, solutions, or predictions that are

co-constructed during the course of that talk. Moreover, they may be

more hidden from us as observers. I shall return to this point in a

section below dedicated to establishing a cultural psychological

approach to peer collaboration research.




Summary comments on processes in collaborative work

We set out concerned to justify the further cultivation of peer

interaction *at* classroom computers. Research in the "collaborative

learning" tradition proved most relevant to our interests because it

studies intimate and focussed interactions of the sort necessitated by

computer-based work. In reviewing this research, we have encountered

a number of ways in which joint work can be cognitively stimulating.

One implication running through this research literature is important

to stress. Peer-based work is identified as being more than a mere

approximation to some other, more ideal, arrangement for learning -

namely, the same interactions involving a more expert (adult) partner.

I have already commented that there are persuasive grounds for

arranging peer interaction as an experience that is worthwhile in its

own right. Children can usefully be offered opportunities to

coordinate their problem solving skills with peers. It is a rewarding

and a valued achievement to be able to do so effectively. However, it

is also argued that such collaborative problem solving is uniquely

suited to the cultivation of certain forms of thinking. That is,

there may be varieties of intellectual stimulation arising in work

with peers that are less readily furnished within interactions

involving more expert adults.

Piaget (1928) makes this point quite strongly in comparing

the two kids of interaction. He argues that the implicit authority

characterising children's asymmetrical relations with adults can be

counter-productive in some problem solving situations. It is the

symmetry (in authority) of peer-based discussion that most effectively

forces useful reflection. It encourages active evaluation of the

status of one's own ideas - as legitimate alternatives to that of a

partner. The tension arising from two *like* minds coming into

conflict prompts resolving argumentation and reflection - rather then

deference to authority.

Moreover, a distinctive role in cognitive development for peer

interaction need not depend only on argument and conflict (the

features of interaction stressed in Piaget's version of this insight).

Research suggests further grounds for recognizing a selective

advantage to problem solving with peers. Advantages, in other words

that are not necessarily mediated only by the resolution of

disagreements. It is true that the asymmetry of relations with an

adult may generally be advantageous in situations of guided

participation (such as those discussed in relation to the zone of

proximal development). However, this advantage is not inevitable.

Kruger and Tomasello (1986) made a formal comparison between children

engaged in (moral) reasoning both with peers and with a familiar

adult. They showed that the two settings released different

contributions from the children. In particular, the asymmetry of the

adult encounter seemed to inhibit certain forms of reasoning: certain

spontaneous logical operations directed towards a partner's

contributions were more likely with peers than adults. Thus, as a

forum for "articulating" (as discussed above) particular kinds of

reasoned discourse, peer interactions may offer precious





It might be expected that students of peer interaction would be

among the most assertive in promoting socio-cultural interpretations

of learning. The cultural perspective - as outlined here in chapter 2

- dwells upon the essentially social nature of instruction and

learning. Should not research activity into peer interaction promise

some sort of commitment to a truly social characterisation of


It appears that serious commitments of this kind are scarce. Studies

of classroom collaboration are a disappointment, if we are looking for

genuinely radical theoretical perspectives on cognition. Perhaps it

testifies to the deeply set nature of our individualistic theorising,

but there has been little sign of researchers in this area developing

a more socially-grounded interpretation of cognitive activity. I will

argue in this section that a truly useful interpretation of research

on collaboration requires us to develop a conceptual vocabulary for

talking about cognition as a distributed, or shared, achievement.

Let us approach this by considering the way in which claims about the

impact of collaboration are derived from research. The predominant

orientation towards the issue of peer interaction is to frame

questions of the form: "What is the effect of collaborative learning

upon fostering ability X?" Often such questions will be motivated by

concern to make recommendations for optimal classroom practice.

Sometimes the question will be directed towards evaluating theoretical

claims regarding the origins of certain cognitive changes (eg. that

they may originate within social conflicts). In any case, the natural

design for a research exercise around this topic can be summarized in

the following notation that describes a typical three-stage


(1) T1 >> IND >> T2

(2) T1 >> PIa >> T2

(3) T1 >> PIb >> T2

In the simplist of research designs, comparisons might be made of

events arising in condition (1) vs. condition (2). In each case,

subjects are pre-tested (T1) on some problem. The groups differ

according to how they are occupied next, during an intervention stage.

In condition (1) they work independently at an activity relevant to the

problem solving skill. In condition (2), they engage in some version

of peer interaction (PIa) around the same activity. Of special

interest will be performance of subjects (working alone again) on a

post-test (T2). In the comparison of (1) and (2), there might be

post-test advantages associated with participation in condition (2).

In this case, it would be claimed that the post-test achievements were

facilitated by peer interaction (PIa). Such interaction led to more

progress than that associated with solitary learning (IND).

Alternatively, in the comparison of (2) and (3), there might emerge

distinguishable advantages associated with one of the two arrangements

for organizing the interaction (PIa or PIb). In which case, the

research would have made progress towards identifying the particular

conditions of interaction that were most favourable for learning in

this forum.

Such an interest in pinning down the particular basis of an advantage

for collaborations might be pursued in a third kind of design.

Without any necessity for pre-testing, very detailed recording of

events during peer interactions themselves might be made in order to

correlate such measures with individual performance on post-tests.

For example, if certain talk within an interacting group was

positively correlated with later test performance of the individuals

participating in that group - then we might suppose that this kind of

talk was a potent factor mediating the advantage of collaboration. We

might then take steps to cultivate it.

This all appears suitably elegant and plausible. Moreover, I do not

wish to imply that real insights can not be derived from research

designs of this sort. They may well help specify for teachers,

some possible consequences of some kinds of classroom arrangements

for pupil working. Yet, in terms of developing an underpinning

theory, the individualistic orientation of this research strategy

needs to be recognized. This orientation may cause us to miss

opportunities for re-framing some of these issues and, perhaps,

investigating them from new perspectives. My strategy here will be to

think further on the way in which we currently conceptualise events

associated with "PI" conditions and "T2" conditions - as notated in

the scheme above.



Interpreting the post-intervention performance

Post-testing (T2) is perhaps the most plausible feature of all in this

system and it might seem the most difficult to challenge. Probably we

are persuaded of the crucial need for this testing because we are

committed to the idea that what matters is how individual pupils

function when left on their own - hopefully doing better as a result

of our interventions. It would be claimed that, in the end, we must

equip pupils for deploying their individual cognitive resources (to

"think independently") in all the situations that this will be

required of them. And, of course, life will furnish many such

situations; not the least of which will be the formal examinations

through which pupils are typically assessed. In fact, our research

post-tests may feel very similar to some of those examinations and,

so, our discovery of interventions that optimise performance on such

occasions can surely only be of benefit to pupils.

There is one kind of challenge to this reasoning that I do not

wish to dwell upon. It notes that intelligence is actually more often

exercised in collaborative settings than within solitary reflection.

Therefore, it is claimed, what should be of interest to researchers is

post-tests of how children perform *collaboratively*. If a post-test

demanding unsupported solitary thinking fails to show an advantage of

learning through peer interaction, we need not be concerned: why

should the one kind of achievement be evident in what is a different

arena? I find this challenge relevant but not very comprehensive.

It is surely true that educational practice must recognize that much

problem solving does now require the ability to gather and coordinate

with others. This may imply that cultivating (and assessing) such

capabilities might deserve more attention from research and practice,

but it does not imply we should make this an exclusive preoccupation.

Yet I do believe the emphasis on typical post-intervention testing is

still problematic and that the problems are made visible to us when a

proper cultural orientation is adopted. The cultural analysis of

cognition assumes its inherently social nature. But this does not

amount to making only a narrow link between cognition and social

interaction - a "social" nature in the face-to-face interpersonal

sense. So, situations demanding cognitive competence will be socially

situated in that, indeed, they may require interpersonal negotiations;

but they will doubtless also call upon the manipulation of particular

artefacts, or the exercise of particular rituals, or the deployment of

particular symbol systems - all in the service of particular goals

defined within particular social or institutional relationships...and

so forth. In short, such occasions of cognitive testing will always

require testees to coordinate and deploy a given set of mediational


Consider this applied to a concrete case (Howe, Tolmie and Rodgers,

1990). Here certain forms of collaborative discussion around a

physics problem of sinking and floating was shown to be a helpful

experience for pupils. That is, it facilitated performance in a

post-test requiring individuals to make decisions about what will sink

and what will float. This is useful: for one thing, it points at

possible classroom interventions that might be expected to improve

pupil performance on certain sorts of tasks. But exactly what has

been shown to have been facilitated in this sense; what has been

learned? A lot hinges on our intuitions about the relation between

the particular context of the post-test, and contexts typical of other

situations in which problems relating to the physics of floatation

might arise. The research-based test is just one occasion for such

thinking to take place. In this case, it does happen to have an

interpersonal dimension that will be relevant: the questioning is

managed by a particular individual in a particular relation to the

pupils - the experimenter. But the test also involves the

manipulation of particular artefacts in a particular problem space

(various objects around a water tank); testing that is continuous with

a classroom agenda; and all of this, in turn, is embedded in

"schooled" life and values.

Thus, the circumstances of research post-testing (or, indeed, school

examinations) seem rather circumscribed from the perspective of a

cultural psychology. A number of studies have now demonstrated how

the particular circumstances of controlled testing can exert a great

influence on how children think in such settings (eg. Donaldson, 1978;

Siegel, 1991a, 1991b). Tests conceived for research purposes

typically just characterise one kind of context in which cognitive

activity must be organized. They are too easily conceptualised as a

convenient instrument for assessing some "general" cognitive advance

that is supposed to have taken place. So, the temptation is to see

them as unproblematic evaluations of (private) cognitive resources

that are relatively context-independent: the test being just an

arbitrary sampling exercise on these resources.

How else can the outcome of experimental interventions be evaluated?

By taking seriously the issue of context: evaluating the range of

conditions under which the experiences of an intervention surface to

guide action elsewhere. There seems little interest among

psychological researchers for developing their work in this direction.

I believe this reflects a theoretical prejudice towards isolating

relatively context-independent cognitive achievements. This

encourages self-contained interventions and self-contained evaluation

instruments. When reproduced within classroom practice, it might

leave us with pupils who have "knowledge in pieces" (DiSessa, 1988) -

rather than truly versatile ways of thinking. So, the authentic

circumstances for promoting cognitive development may reside in the

knitting together of contextualised experiences - particularly as this

may be achieved through the discourse making up a classroom's communal

life. This possibility is exemplified in a paper by Edwards (in

press), in which he outlines an agenda for analysing conceptual

development in terms of children's participation in such discursive

practices. This offers an alternative to our preference for always

addressing these issues by sampling an inner world of discrete

conceptual acquisitions.

The more conventional practice of psychological research is not a

*fatal* methodological attitude. In the case of the sinking and

floating example, the study remains interesting and the results are

compelling. For, often what gets done by way of testing carries some

real face value; it seems to promise achievements in other situations

that are not formally evaluated. Yet, a narrowness of reach remains

the significant point. It is rarely thought necessary to pursue the

question of what has been learned - as if it could be something that

was actually bound up with contexts of acquisition or practice, and

that such embedding might usefully be documented. It seems very

seductive to suppose that a successful post-test confirms that the

testee now has some new logical "bit" in place; or some new cognitive

skill has now become available. Such characterisations may reflect

the influence upon us of traditional theorising in cognitive

psychology. That tradition tends to view the achievements of

organized learning as private, contained and relatively context free:

something to be sampled in controlled tests.

Interpreting the joint activity within peer interaction

The same biases of interpretation arise in analyses of events

within the peer interactions themselves - as typically organized for

experiments. In the notation used above, the intervention stages of

PIa and PIb are often characterised in terms that foreground

activities of the *individual* participants in the collaboration. The

collaboration may be expressed as a sum total of certain sorts of

discourse events that were observed. This gives us only limited

insight into what is typically happening in these collaborations - and

how it may be cognitively significant. I shall follow up this claim

in more detail here, giving no further attention to the problems

arising from the post-testing of interventions. Generally, I am more

concerned with weaknesses in interpreting the collaboration stage of

the standard empirical procedure - rather than with weaknesses in

interpreting the post-test stage. This is because any prescription

for involving computers within collaborative work will need, most of

all, to be based upon an adequate account of the interactional

processes active during such work.

My plan here is, first, to indicate what it means to claim that

typical analyses of collaborations are individualistic; then, to

speculate as to what a mutualist or more socially-grounded alternative

analysis would be like.

I believe the orientation to individualistic accounts arises from the

tacit influence of two influential concepts from the mainstream of

psychological thinking: "stimulus" and "skill". To be more specific,

I shall suggest that the potency of a collaboration is typically

characterised as some pattern of discrete here-and-now social events

(stimuli). These events act upon (and, perhaps, change) the

collaborators. This is too limiting a conception. To enrich it,

certain recent accounts have made reference to intersubjective

achievements that can be mobilised during the joint activity.

However, this idea can also be limited in its reach; if such

achievements are only framed in terms of social attributes (skills)

brought to the situation by the individual collaborators.

Let us consider the influence of "stimulus" conceptions first.

Research that attempts to understand peer-based problem solving has

typically concentrated on the content and organization of the talk

that occurs within it. The common strategy is to classify and

enumerate individual utterances made by the participants (eg. King,

1989; Webb et al, 1986). This allows any such occasion to be

summarised as a profile of categorized talk. It may then be found

that certain prevalent categories in such a profile are positively

correlated with later, post-test performance. I suggest that a very

common attitude towards this correlation is one assuming that the talk

was effective because it acted as a *stimulus*: it served to prompt

some kind of cognitive effect for the collaborating individual. A

significant property of an event conceptualised in this manner is that

it tends to be taken as having no history or context: it is an event

only of the here-and-now.

For example, the Piagetian emphasis on processes of social conflict

tends to get elaborated in these terms. Piaget's account is taken to

mean that, during collaboration, things that are said by one

participant (challenges, denials, assertions etc.) can serve to

precipitate useful cognitive restructuring in that individual's partner.

Such talk is, thereby, a "stimulus" for cognitive change.

As discussed earlier in this chapter, another analysis of

collaborative talk stresses the value of simply articulating one's

thoughts. In this model, an individual will benefit from entering

into a problem solving situation that prompts participants to make

their ideas explicit. Again, the social structure of the encounter

facilitates cognitive change in the individual. In this case, we

might say that the individual's own talk is the stimulus for change:

utterances are elicited by the situation and thereby (may) prompt

cognitive restructuring.

My problem with accepting this kind of analysis is its implication

that all that collaborators are ever doing is generating a corpus of

discrete conversational events. Researchers have then to determine

what circumstances of joint activity are potent in influencing the

composition of this corpus. They must determine what circumstances

facilitate the production of utterances with those properties that

post-tests reveal are predictive of a desired performance. Yet, what

collaborators are actually doing may often go beyond this. What is

said by them at any given moment may gain its potency not just from

its immediate pragmatic content (assertion, challenge, hypothesis, or

whatever). A given utterance may be important for its contribution to

the construction of a more sustained socially-shared framework of

interpretation - some form of understanding that is common and that is

recognized by the participants as being held in common.

This is a conception that I wish to take up and illustrate in the next

chapter. It invites us to view the process of collaborating as

involving a discursive achievement: the extended construction of some

degree of mutual knowledge. Then, analysis of collaborations in these

terms requires us to go beyond the categorisation of individual

utterances. Instead, the talk would be approached as discourse:

organized and evolving in time. How we understand particular

contributions should pay some respect to the context of common

understanding that that they are helping create in this way.

Of course, such shared understanding is a species of mental context

comparable with that discussed in the last chapter: there it was

identified with what teachers typically have to do through their

classroom talk. I am anxious to encourage an approach to what happens

in peer interactions that learns from this analysis of more

instructional forms of conversational interaction. In both cases, the

achievement of new understandings rests upon the possibility for

adopting an intersubjective attitude (Davidson, 1992).

I believe this has been recognized by some commentators on this topic.

However, I am concerned that the insight should not be assimilated to

a more traditional form of psychological analysis. This is the reason

for my suggesting above that we might be wary of the concept of

"skill" as directing the way in which we thought about intersubjective

achievements. I shall conclude this chapter by identifying what this

line of tempting thinking involves and how my own reference to

intersubjectivity differs from it.

An appeal to intersubjective processes might enter at the point where

we consider why effective talk occurs within some collaborating groups

and not within others; or why it appears at some stage of childhood

development but not earlier. A natural move to make when

contemplating such issues is to refer to psychological attributes of

the participating individuals themselves. Some notion of differential

interactive *skill*, for example, might be cited to account for

individual variation in collaborative success or variation associated

with the age of participants. Thus, Brownell and Carriger (1991) have

analysed events within collaborative encounters from such a

developmental point of view. They stress changes within individuals

that mediate the impacts of these occasions:

As they get older, children become better able to comprehend the

behavior and intentions of another, as well as better able to affect

the other's behavior and to communicate about their own behavior and

desires. We are additionally led to conclude, then, that age-related

social and cognitive skills contributing to peer collaborations

influence what and how children learn from these collaborations (p.


A similar orientation is apparent in a review by Tomasello, Kruger and

Ratner (in press): achievements within collaborative learning are

associated with developmental advances in children's "social-cognitive

abilities". In these interpretations, the success of peer interaction

as a setting for learning is linked to skills and competencies

brought to the collaborative arena by particular individuals, or by

children at particular stages of development.

My own view is that it is very helpful to make links of the kind that

these commentators are exploring. Thus, students of collaborative

learning gain from having their attention drawn to research concerning

children's growing interest in the mental life of other people (i.e.,

children's theories of mind as reviewed in, for example, Astington et

al, 1988, or Dunn, 1988). It is helpful to recognize that what

children are inclined to do in the situations studied by

theory-of-mind research is very relevant to what they do when solving

problems with peers. Moreover, it is quite proper to explore age

differences and socio-emotional variation as it might bear on the

success of collaborative problem solving. So, we do wish to develop

an interesting association between children's varying understanding of

mental states in others and their potential for collaborative

learning. I am less happy that the way forward is to conceptualise

as if it were socio-cognitive skill in this domain. Something is

changing among and within individuals, but reifying it in this way may

distract us from really understanding the nature of that change.

Perhaps this is the very move that is naturally encouraged by

theoretical traditions that still keep cognition and social context


For example, it may be suggested (cf. Tomasello et al, in press) that

achievements in certain collaborative contexts become possible for 6-7

year olds because this is the age at which children acquire a certain

understanding of mental life in others ("second order mental states",

to be precise (Perner, 1991)). A useful correlation is thereby noted.

However, such state-like characteristics of individuals tend to carry

too much causal responsibility in our explanations; perhaps leaving us

too complacent that our account has solved a problem. In the present

example, we might ask: What should be explored further, now this link

between collaboration and theories of mind has been noticed?

The answer in this case, I suggest, is that we should investigate the

social processes whereby children's growing awareness of mental states

is mobilised into the particular service of organized problem solving.

The idea here would be that such intersubjectivity emerges at an

early age, reflecting something basic to human nature (Trevarthan,

1988). Through participation in cultural activities it is deployed

for new purposes of interaction. So, the earliest understandings of

other minds may be fostered in very contained, domestic social

settings (Dunn, 1988). However, children will then usually come into

contact with culturally-organized pressures of a powerful kind. For

example, pressures that force these new understandings to embrace

coordinated activity around schooled kinds of problems: to embrace

collaborative learning in fact. So, if intersubjectivity does become

a resource to support collaboration, it is because the conventions,

rituals, institutions and goals of organized social life arrange that

it should do so. This is the phenomenon we need to understand.

Teachers and others need to understand how best to mobilize an

intersubjective attitude towards the particular purposes of joint

problem solving.

A point that I wish to make here is that our natural focus on

individuals (and their social or cognitive attributes) tends to

distract us from pursuing research agendas of the kind sketched above.

Individualistic accounts furnish one way of talking about our

observations. That way tends to deal in a currency of

self-contained, central psychological characteristics of individual

actors. It is not suggested here that this language is simply wrong -

and that some definitive experiment might show this. The question is

whether these concepts furnish the most productive way of

systematising what we have observed.





In this chapter, I have reviewed a background of research that

concerns collaborative learning. This literature reveals a number

of important findings. Firstly, that group work is commonplace within

the organization of early education and generally prized throughout

formal schooling. Secondly, ethnographies of actual classroom life

indicate that it is hard for young pupils to sustain joint

investigative talk within these structures. Thirdly, controlled

studies of collaborative problem solving indicate that pupils often

learn more from these socially-organized tasks than they do from tasks

tackled in solitary working arrangements. Finally, studies

correlating collaborative strategies with posttest achievements

suggest the importance of a number of interactional variables -

particularly, experiences of publically articulating ones reasoning,

experiences of conflict and experiences with the co-construction of

problem solving resources.

I have also reviewed here material suggesting that computers may be

one potent setting for the support of collaborative work. This is of

special interest, given the findings that such work is not inherently

easy to cultivate in classrooms. However, there are also hints that

while this technology may sustain interactions that are lively, this

does not necessarily mean that they are rich in a cognitive sense.

Sheer energy is not all that we may wish to release when organizing

work in this way. The invitation is to research more closely the

dynamic between social exchange and the structure of computer-based

tasks. Yet, I have argued that the terms in which collaborative

interactions typically have been analysed seem too narrow to provide

real insights. It is hard to see that the coding and counting of

collaborative utterances is going to shed genuinely new light on our

concern for the exact place of computers in all this.

I believe that what is needed is a richer conceptual framework for

characterising the psychological processes that are involved.

Most research at present is concerned to characterise collaborative

interactions in strongly individualistic vocabulary. Collaborations

become arenas in which certain classes of event are likely to occur:

in particular, certain classes of utterances. Their effects on

individual participants are then to to be assessed in subsequent

tests. A more socially-grounded account is possible and, in

particular, there are examples of analyses that have explored the

theme of intersubjectivity in relation to collaborative discourse. I

am still not satisfied that these are doing the job that is needed.

Typically, such accounts set out in the right way. It is supposed

that collaborating partners each have capacities to project mental

states in the other and that this facilitates their "co-construction"

of problem solutions. However, what is analysed as co-constructed

here turns out to be "social" in only a rather narrow sense. It is

the particular solution (hypothesis, prediction or whatever) that the

shared problem demanded. The social character of that achievement

is judged to reside in the partitioning of task responsibility and the

subsequent coordination of individual insights. That is what the

co-constructing attitude is said to support. I would like, in what

follows, to pursue a further sense to this intersubjectivity.

What else could we mean by socially-shared cognition? What other kind

of object or outcome could be co-constructed within peer interaction?

As hinted above, the position I wish to develop is another view of

what must invariably be constructed within collaborative encounters:

it is a view influenced by the idea of "common knowledge" as described

in the previous chapter. That was a concept developed to help clarify

the nature of collaborations between teachers and pupils. However, it

may also refer to socially-shared cognition that exists between

collaborating peers. I suggest that the success of such encounters

will often reside in how effectively the participants co-construct a

shared mental context for their problem solving efforts. So, the

cognitive object that gets created will be this common ground. It

becomes a point of shared reference and a resource for prompting

individual contributions to the problem solving discourse.

The particular question of concern to us here is how educational

technologies mediate in the creation of such socially-shared

cognition. More particularly, what are the characteristics of

computer-based activities that might afford such successful

achievements within peer interaction? I shall illustrate some

possibilities in the next chapter.