25th October 1993





So far, I have identified some controversies arising from the

increasing use of computers to support learning (Chapter 1); and I

have identified some traditions of psychological theory that might

help us think more clearly about these issues (Chapters 2 and 3). In

the present chapter, I shall start to discuss particular

configurations whereby computers enter into learning activities,

developing the term "collaboration" as an organizing concept. I

believe the term is central to the cultural approach in the sense that

"computation" or "construction" are central to other relevant

theoretical traditions in Psychology. However, in the end, the term

serves as a device to think with. I do not claim there is any

widely-shared commitment to "collaboration" as the key concept in

socio-cultural thinking.

Each of the remaining chapters in this book concern social

configurations for computer-based learning. In this and the following

chapters, I am concerned with the most orthodox of situations: that

involving organized asymmetry of expertise (expert and novice; teacher

and pupil). I wish to look at very general arrangements for

incorporating new technology into the teacher-pupil exchange. I have

already claimed that there are some who hope computers might actually

*become* the teacher in this interaction. That possibility is the

first to be considered: it will be discussed in the present


In the first Section below I raise the prospect of computers

simulating social processes in the tutorial sense. It must be decided

whether interactions *with* computers can capture the social quality

of traditional guided instruction. There is a kind of educational

software that is written to do this. We shall find that it is

interesting but modest in its "social" achievements. So, from the

user's point of view, computer-based instructional dialogue typically

feels limited in its reach. It is brittle and inflexible. I shall

apply the framework of cultural psychology to make sense of why this

should be so. This will require presenting the cultural analysis of

just what does constitute effective instructional dialogue - as it

occurs in those interpersonal contexts of teaching and learning that

we are familiar with. It will become clear that comprehensive

computer-based modelling of such interactions is not a realistic


In the course of considering computers and instruction in these ways,

one important theoretical concept will surface. That is the concept

of "intersubjectivity" or, briefly, shared understanding that is

mutually recognized. I suggest that intersubjectivity is central to

what occurs within instructional communication. The following

discussion of how such talk is typically organized will draw attention

to one aspect of what instruction involves. However, I will suggest

that the cultural approach to cognition has not taken real advantage

of the intersubjectivity concept. The approach could mobilise

the concept in order to extend its theoretical resources for

interpreting the management of learning. My own discussion here will

converge on a particular conceptualisation of what the act

of instruction involves: a powerful form of collaboration arising when

the human capacity for intersubjectivity is explicitly deployed to

achieve guidance within arenas of joint activity.






One response to the fear that computers will undermine the social

quality of education is to argue that such concerns will inevitably

evaporate: the present tension will be resolved and the problem will

go away. The argument proposes that what we value in the social

experience of learning will be preserved: it will merely, in some way,

be taken over by the technology. Such visions aim to respect what we

wish to retain in the current system - by supposing that it can be

simulated and not remain dependent upon the interventions of other

people (in particular, teachers). This does not seem likely to

happen in the near future but enthusiasts will simply plead for more

time. Given a few more technical developments, it is argued,

computers will engage in instructional conversations just like the

ones pupils already enjoy with human teachers. With varying degrees

of confidence, this view is voiced by those working in the tradition I

have identified as computer-as-tutor. The idea is to incorporate

whatever is important in novice-expert dialogue into the design of

computer programmes. This is often the sense of "intelligence" that

is appealed to in so-called "intelligent tutoring systems"

This approach has its enthusiasts (Quere, 1986). Appealing to certain

recent technical advances, Henderson (1986) comments:

In effect, we believe that an instructional system comprised of a

videodisc player interfaced with a microcomputer should be able to

simulate a coach or the master/apprentice relationship quite

effectively (p. 430).

Of course, it is accepted that any such enterprise must be founded on

much basic research concerning just what gets done by "coaches" or

"masters" as they interact with learners. In Lepper and Gurtner's

(1989) overview of educational computing, this is identified as one

important priority for future research: learning about conventional

instructional processes in ways that will help us to model them.

The goal of simulating tutorial exchanges has been pursued for some

time by educational software authors. Collins (1977) reports an early

and influential exploration of the problems. His work includes an

intriguing systematisation of the Socratic form of teaching dialogue

as gleaned from transcripts of authentic instruction. The aim was to

extract principles that might be modelled in an intelligent tutoring

system. Yet, 15 years on, the promise of this analysis, and others

like it, does not seem to be visible in any strong tradition of

dialogue-based teaching software. In fact, Collins himself appears to

have adopted a different approach to the design issues (eg., Collins,

1988; Brown, Collins and Duguid, 1989) - as it happens, an approach

more in tune with cultural psychological ideas.

Moreover, other authors working under the rubric of intelligent

tutoring systems have proposed a switch of focus away from the literal

simulation of tutorial dialogue. For example, in Chapter 1, I

referred to the work of Schank and others on tutoring systems designed

to support case-based forms of reasoning: styles of problem solving

that, it is supposed, are more natural for us (eg. Riesbeck and

Schank, 1991). Such systems expose learners to a range of problem

"cases" (typically as simulations) - supposing that solutions to novel

problems are naturally made with reference to private accumulations of

case-based experience. This is certainly an alternative to

traditional intelligent tutoring systems, with their stress on

modelling tutorial dialogues. However, it does not challenge the

principle of reproducing socially-organized instruction. It merely

challenges the typical conception of what tutorial interventions

attempt to do: in particular, the idea that they equip learners with

repertoires of rules of the "if x then y" variety. In the end, a

tutoring system built around theories of case-based reasoning may

still fall short as a comprehensive system of instruction. For

example, it may turn out that human tutorial intervention is an

important feature of how the learner is helped to index and retrieve

this case-based knowledge (a form of social support that we may

suppose will remain particularly hard to simulate).

For whatever reason, no species of intelligent tutoring system has

found a firm place within academic education. This mode of

computer-supported learning is found more in situations that we might

usually refer to as "training" - perhaps where circumscribed technical

skills are being developed. Thus, military and industrial

applications have been well documented. However, evaluation studies

tend to indicate their potential in these settings is also limited

(cf. Schlechter, 1986). My discussion in the sections below is, in

part, an effort to interpret this lack of success. I argue that many

difficulties arise from failing to appreciate the subtle nature of

instructional talk. Yet, I do not wish boldly to legislate against

simulation of such dialogue - as if it inevitably led to applications

of no value. It seems to me that such simulation can be attempted for

some portion of what teachers and learners might normally talk about.

Indeed, there are examples of interesting and effective applications

that appear to achieve this. However, such circumscribed successes

provide no basis for predicting the wholesale replacement of

traditional tutorial exchange - if that is part of any educational

vision. Generally, the possibility of computers reproducing the role

of teachers by supporting a genuine interpersonal experience is based

on some unlikely suppositions. I shall identify them in the sections

below. As other critics have expressed it, the idea supposes 'that

the teacher's understanding of both the subject being taught and of

the profession of teaching consists in knowing facts and rules..'

(Dreyfus and Dreyfus, 1986, p.132). If such knowledge was all that

was needed to be reproduced, then the possibility of programming a

computer simulation might be credible. But teaching surely does

involve more than dealing only in tidy rules - either rules pertaining

to domains of knowledge or (interpersonal) rules governing the

effective performance of instructional talk.

Dreyfus and Dreyfus (1983) consider the danger of conceptualizing

learning as the mastery of sets of rules defined for various

knowledge domains. I shall not rehearse their persuasive arguments

here but focus instead on the other part of what tutoring systems

typically aim to simulate: the expertise that is involved in the

social act of instruction itself. Early innovators felt able to write

papers with such titles as 'The computer that talks like a teacher'

(Feurzieg, 1964). But is this extraordinary interpersonal achievement

one that might be expressed in the rule-based formats that the

programmers of computers demand? I suspect not; and, 25 years later,

some of the same early innovators are now writing articles on

computer-based learning with such words as "apprentice" and

"practitioner" in their title (Feurzieg, 1988). However, to help

decide about these matters, we need to consider the psychological

processes that underly effective instructional exchanges. I shall do

this next. I shall then return later in the chapter to consider

again the prospects of creating pupil-computer interactions that are

supposedly "social" in character: instructional interactions *with*





In this Section I review the analysis of instructional talk that has

been emerging within cultural psychological theory. As it happens,

this is the theoretical tradition that has devoted most attention to

such matters. The discussion returns us to a consideration of

Vygotsky's zone of proximal development, a concept introduced in

Chapter 2. Cultural psychologists have various formulations of what

goes on in effective instructional exchanges: these ideas arise from

empirical observations of social events within ZPDs. I believe that

their analysis has been helpful in clarifying what is precious within

socially-organized instruction. Although, in the next Chapter I shall

suggest that the distinctions developed below reflect too great a

preoccupation with exchanges that are intense and intimate in nature -

one-to-one interactions between experts and novices, teachers and

pupils. In the end, a full understanding of the social nature of

instruction requires us to inquire beyond such a narrow focus. It

requires us to consider a wider range of interpersonal interactions

that can arise in classrooms.

Supportive encounters in the ZPD: (i) internalization


Consider a child solving a jigsaw puzzle by herself. This simple

task demands a variety of strategic problem solving activities. There

are helpful ways for the child to go about arranging pieces (say,

right side up, grouping by visual features, picking out corner pieces

or pieces with a straight edge). A pattern of attention needs to be

organized between the pieces on the table and the picture of the

completed puzzle. The child needs to remember juxtapositions that

have already been attempted...and so on.

If the child is young - say of preschool age - and if the puzzle

comprises more than a handful of pieces, we can predict that progress

will be limited. Very generally, we will always risk provoking

disengagement where a problem-solving task is set just beyond a young

child's expertise. Yet, for a good learning experience, the task must

be challenging. So, a flexible way must be found to support children

in the discovery of solutions to more complex tasks - particularly

tasks where trial and error explorations are likely to be inefficient

and tedious.

Cultural theories of cognitive development argue that humankind has

evolved a solution to this problem: one that involves cultivating

forums of joint activity. In particular, we have evolved practices

whereby individuals who are expert in some domain will collaborate in

distinctive ways with novices and, thereby, communicate their

expertise. At their most effective, these are occasions in

which experts go beyond simply *showing* the novice what is to be

done. These are occasions which are potent because expert and novice

join to construct a joint "cognitive system". It is useful to think

of such a system as having a unitary nature, although it is actually

comprised of (at least) two thinking individuals. However, although

there may be two people involved, their work need not be partitioned

and individually allocated. In an effectively organized ZPD, the

novice is assumed to be doing it along with the expert, who may be

judiciously steering or prompting. Rather than being driven by

showing and explaining, these encounters encourage the novice's full

"participation" in the problem solving act; they are conducted in the

spirit of collaboration.

The management of an encounter like the one described above, one

involving children with their mothers solving a jigsaw, has been

described by Wertsch, McNamee, McLane and Budwig (1980). Joint

activity in this case is shown to take on the quality of a

unitary cognitive system. It does this by virtue of how

responsibility is distributed for the various strategic activities

involved. The adult in this situation will take responsibility for

some of those strategic moves that seem to be currently beyond the

reach of the child - although with both participants remaining

focussed on the same goals. At other moments, the adult might do and

say things to prompt the mobilization of a strategy that is within the

child's repertoire but not spontaneously elicited by the situation


Encounters of this kind are typical of those that children may

experience in the everyday world of solving problems. Tharp and

Gallimore (1988) cite the example of an parent intervening as a child

searches for her shoes. The intervention acts at a level of

supporting the cognitive activity of remembering; it creates a

cognitive system:

...the father asks several questions ("Did you take them into the

kitchen? Did you have them while playing in your room?") The child

has some of the information stored in memory ("not in the kitchen; I

think in my room"); the father has an interrogation strategy for

organising retrieval of isolated bits of information in order to

narrow the possibilities to a reasonable search strategy. The child

does not know how to organize an effective recall strategy; the

father knows the strategy, but he does not have the information needed

to locate the shoes. Through collaboration, they produce a

satisfactory solution.

This is a mundane example from domestic life. Tharp and Gallimore

endorse a typically socio-cultural analysis by presuming such

exchanges are richly produced within formal instruction. Thereby, the

child is able to participate in substantial (but meaningfully

complete) problem solving exercises: tasks that it might be impossible

for that child to pursue alone.

This conception of joint activity underpinning instruction is

appealing precisely because it seems to describe expert-novice

interactions in a wide variety of learning situations. Thus, it does

refer to classroom exchanges (eg., Palinscar and Brown, 1984; Tharp

and Gallimore, 1988) as well as more everyday and informal joint

encounters (Greenfield, 1984; Rogoff, 1990). Further, it offers a

basis for characterising variation in instructional success: for

example, it may assist the optimising of learning opportunities for

groups of people with special educational needs (Wood's research

(1989) is an example of this in relation to deaf people). Wood also

demonstrates that when tutors instruct in the manner outlined here,

children may regard them more warmly (after the teaching sessions)

than than they regard other tutors who adopt a more directive or

didactic strategy.

So, to summarise, it is argued that in this zone of proximal

development social interaction may serve to create a unified cognitive

system. Then it is supposed that the public nature of constituent

moves within this system (the various talk and action) can promote a

process of internalization by the novice participant. What is

performed in the arena of joint action gets internalized into the

private world of the novice's own mental life. Under such

circumstances, individual cognitive resources are first experienced on

this public plane of collaboration; they are then adopted as private.

It is tempting to suppose that this traffic from inter- to intra-

mental functioning simply implicates a process of "modelling". This

may be part of the story. In a busy zone of proximal development, the

learner is exposed to exemplars of strategic problem solving.

Questions are asked, directives are issued, remembering is invoked,

summarising classifications are deployed, and so forth. Simply being

witness to such activities may sometimes be enough to make new

resources available to the onlooker. However, it is widely argued

that simply modelling such processes for learners is rarely enough.

The possibility of internalization is claimed to depend upon active

participation within such encounters (Wertsch and Bivens, 1992).

Rogoff (1990) captures the spirit of this idea by referring to

successful tuition within a ZPD in terms of "guided participation".

The concept of internalization has been much appealed to by

researchers sympathetic to cultural theorising or - more generally -

seeking some socially-grounded account of cognitive change. However,

I believe that unless the definition of "internalization" is made very

broad (and, thus, not very useful), this emphasis leads us to neglect

other socially-distributed processes associated with instructional

talk. I shall illustrate these in the following two sections.



Supportive encounters in the ZPD: (ii) semiotic mediation


There is something accessible and appealing about the internalization

concept as reviewed above. Our private cognition is traced to public

events: activities in relation to which we were witnesses or

participants. Other theorists have enjoyed similar approval with

related formulations: notably, Mead (1934) with a conception of

thought as something derived from public discourse - conversation with

'the generalized other'. However, it has proved hard to study this

internalization through a fine-grained analysis: hard to trace

convincingly the origins of cognitive change in a manner that Vygotsky

characterised as demanding 'microgenetic analysis'.

Perhaps for such reasons, there has emerged a further (complementary)

perspective on how we may characterise the formative nature of these

(ZPD) social interactions. It is one that less readily suggests the

notion of internalization. Moreover, it suggests "provocation";

rather than the "assistance" of Tharp and Gallimore's (1988) 'assisted

performance'. Indeed, it might be viewed as a more cognitive account

in the traditional sense. For, it gives emphasis to the elicitation

of private cognitive processes: but processes that are prompted by

social participation.

Stone and Wertsch (1984) fix the relevant idea as follows. They draw

attention to the manner in which instructional dialogues are often

characterised by prolepsis. This term refers to communication in

which interpretation of the message requires some grasp of the

speaker's presuppositions - understandings which are left unstated.

Such messages may be termed underspecified or richly presupposing.

Consider a casual example involving parent and child:

C: Where did you put my shoes?

P: Over by the animals

C: (Pause) my Heavy Metal poster, you mean.

The parent's underspecified answer here forces the child to pause and

seek reflective clarification. So, the child's participation in this

brief exchange effectively prompts a resolving inference regarding the

parent's opinion of musicians on a rock band poster. If the child had

not "calculated" the meaning of her parent's answer through this

route, then the adult might have gone on to supply more of the context

to his reply - although necessarily *after* its original utterance.

Prolepsis illustrates a species of dialogue typically of interest to

conversational analysts. Our talk is normally saturated with it.

However, in the special cases I am considering here, it is being used

in a contrived way: deployed just at the thresholds of mutual

understanding. In my parent-child example (and within encounters that

will arise in the course of formal instruction), shared understanding

that normally supports the comfortable continuity of talk seems to

have been violated. The violation causes tension; the tension demands

repair, and work gets done by the listener to achieve this.

Note that such instructional devices are quite compatible with the

strategic management of problem solving discussed above under

"internalization". The point is that internalization focuses

on the *content* of strategic interventions (eg., questions that

mobilise organized recall). There is room for variation in the way

that these strategic interventions are verbally realized. Exploiting

prolepsis may serve to make the point of the intervention more vivid

for the learner. Thus, Stone (1985) uses an example to illustrate

prolepsis that echoes situations discussed in the last section. A

teacher gives an instruction in relation to solving a jigsaw puzzle:

"put in the next piece". Stone comments: 'This directive presupposes

an understanding of the task's overarching goal, that is to use the

model as a guide for defining the location of the pieces' (p.135).

Perhaps the learner has been verbally prompted to generate this idea

(in order to create some options for reacting to what the tutor has

just said). In doing so, the learner has participated in a (modest)

strategic move appropriate to making progress with this kind of puzzle

at this kind of juncture. The link to Vygotsky's ZPD, supposes that

dialogues conceived for instructional purposes are particularly rich

in such disruptions. Moreover, dealing with the disruptions is a

potent experience for the learner.

Rommetveit (1979b) offers a fuller argument for identifying processes

of this kind as central to human communication. I have cited an

example drawn from speech, but he notes that these processes are

frequently invoked in fiction and drama. Members of an audience may

realize that they have understood more than actually has been said.

The author has taken contextual information for granted, but prompted

its recovery within the audience's effort of interpretation. The idea

that listeners (and readers) are active in this sense - spontaneously

making inferences about discourse and text - is a familiar one to

cognitive psychologists (eg., Bransford, Vye, Adams, and Perfetto,

1989; Sperber and Wilson, 1986). However, the idea that instructional

processes might involve the organized mobilization of such devices is

more novel.

In a recent paper, Wertsch and Bevins (1992) pursue Rommetveit's

interest in identifying prolepsis as a basic communicative resource.

They suggest that the effects noted here for the verbal devices of

discourse have parallels within a wider range of communicative media.

Specifically, they propose a relation between (proleptic) talk of the

kind illustrated above and our experiences with certain expository

text: particularly with that property of text that allows it

sometimes to serve as a "thinking device". Thus, these authors are

considering the manner in which cognitive work is generated within

interaction - but from a broader definition of what can constitute an

interactional context. For them, it is not restricted to the

prototypical ZPD of two or more people in problem solving discourse.

It might embrace interacting with written materials.

Influenced by the semiotician Lotman, they refer to the 'dialogic

function' of such thinking devices. This function of text is more

provocative towards the reader; it demands the construction of shared

meanings. The contrast is with univocal texts: they are more

like receptacles and communicate more in the mode of passive

transmission. Both texts and talk may be too easily thought of as

merely narrowly univocal in this sense. Often they may entail more

dynamic communicative properties: devices that allow both the spoken

and written word to mediate *interaction* - to precipitate new

meanings through active engagement. So, a close analysis of

instructional texts and instructional discourse may be helpful if it

serves to reveal a potential dialogic function.

The dialogic presupposition and exploitation of shared understandings

is a necessary basis for human communication. Moreover, this

manipulation of shared understanding applies beyond the arena of

instructional language. A culture's various communication media offer

a whole range of dialogic devices with which individuals may manage

their interactions with others. So, prolepsis may well be apparent in

the talk of a teacher, but it is also evident in the text of an

author, or even the advertising images of a graphic designer.

The case of advertising illustrates well the reach of semiotic

mediation as an explanatory concept. A concrete example may be

useful. At the time of writing, passengers using Edinburgh station

may be intrigued by one large poster among the various promotional

hoardings in the station concourse. Three sets of (real) objects are

attached to the poster backcloth: a tangle of red trouser braces, a

jumbled group of traditional black telephones and an empty gilt

picture frame. Underneath is the legally required health warning that

implies a tobacco product. I have an idea about what specific product

is being advertised. But that idea emerged only after a period of

reflection - during which time the product was necessarily drawn into

the foreground of my consciousness. Perhaps this is the designer's

purpose. There is just enough presupposition to evoke active

cognitive work on the images and their associations. The viewer is

led to a precipice of understanding and, thus, some reflective

engagement with the product has been achieved. At least, it was

achieved in my own case; it might not always be so. The danger with

this device, as applied to advertising, is that the "precipice" may be

very different for different consumers. Care is needed in

identifying the background knowledge necessary for an image to

precipitate successful engagement. The knowledge presupposed by a

given image may not be widely shared. To be engaging in the present

sense, the image may need to be conceived differently for different


This problem faced by the advertising designer is usually less keenly

felt by the teacher - who, I am suggesting, may sometimes be doing

similar things. In most educational contexts, teachers have more

privileged access to what their pupils already know. They also have

situational access to the focus of a pupil's attention, and to the

extent of that pupil's motivation as it stands at the moment of

instruction. So, communicative devices of the kind discussed above

may be more finely judged. The central point is that in text, in

images and in talk, the effective deployment of semiotic mediation

entails judicious reference to shared understandings. For it is the

successful matching of a message to this existing mutual knowledge

that is important. That matching is what allows the message to elicit

cognitive work in the reader, the viewer or the listener. These

encounters thereby provoke reflective engagement of a kind sought by

the agent of communication. Indeed, the broad scope of this idea

should warn us that, as described thus far, it can do no more than

orient us to a significant phenomena within communication. Much

research must be done to clarify exactly how particular dialogic

devices may achieve distinctive effects of this kind. In terms of our

present interest in schooled learning, we would want particularly to

pursue this in respect of the context and character of talk that is

instructional. In summary, then, this conception of "semiotic

mediation" must become a further ingredient of any cultural

characterisation of effective instructional dialogue - along with the

idea of internalization.



Supportive encounters in the ZPD: (iii) appropriation


Appropriation is yet a third concept that cultural theorists have

deployed to help think about instructional discourse. It is taken

from the work of Leont'ev (1981). Appropriation arises from the

sense-making efforts of both teachers and pupils as they engage within

the contexts of learning. As such, it has recently been very fully

exemplified by Newman, Griffin and Cole (1989) who report an empirical

study of teaching in an elementary school. I will take their work as

a basis for introducing the concept.

First, Newman et al locate their study of instructional processes

within the theoretical framework of the zone of proximal development.

So, they define this conception in terms familiar enough:

The concept of ZPD was developed within a theory that assumes that

higher, distinctively human, psychological functions have

socio-cultural origins. The activities that constitute a zone *are*

the social origins referred to; when cognitive change occurs not only

*what* is carried out among participants, but *how* they carry it out

appears again as an independent psychological function that can be

attributed to the novice (1989, p. 61).

Appropriation is included in the 'activities that constitute a zone'.

Two common features of instructional strategy are identified by these

authors as underpinning it. The first refers to that quality of

indeterminacy characterising a great deal of instructional talk -

indeed, characterising a great deal of interpersonal social life more

generally. The participants may be approaching their interaction from

different positions of understanding, but they are temporarily caught

up going along with each other - trying to create some stable and

common ground. This open nature of such social situations is seen by

Newman et al as a positive force: it invites a variety of negotiable

options to be pursued within the interaction. It is the very thing

that allows the participant's differing starting points to be


The second feature of such situations that is important arises from

the first: from the fact that the participants may not initially

understand each other at all well. In this situation, a meeting may

nevertheless be achieved if the partners are prepared to appropriate

from each others activities: to behave *as if* there were more common

ground than, in reality, there is. Encounters of this kind are

central to early psychological development. As Vygotsky (1978)

comments: 'From the very first days of the child's development his

activities acquire a meaning of their own in a system of social

behaviour..' (p.30). The point is that this meaning is often

something to be negotiated in a collaborative way. So, Newsom (1978)

notes the manner in which the parents of infants will be actively

interpretative in their reaction to a child's behaviour. Parents

ascribe meaning and intent to a degree that exceeds the child's actual

capabilities. There is a compelling tendency towards such creative

attribution of meaning and intent. In a sense, the infant's behaviour

is thereby "appropriated" to the purposes and frameworks of the adult.

Newman et al suggest this is a common state of affairs between expert

and novice in the ZPD; a common characteristic of organized


The analysis is partly inspired by an earlier account of interactions

in a preschool setting by Gearhart and Newman (1980). These observers

were impressed by how teachers would often interpret what a child was

doing in a manner that presupposed the teacher's own perspective. The

child is scribbling: the teacher asks "What is it?" Such a question

presupposes a planful activity on the part of the child - an

interpretation of the activity from the teacher's point of view. In

terminology developed in the previous chapter, the pupil comes to this

situation with established functional cognitive systems: including,

say, one concerned with making scribbled marks. The pupil's progress

depends on the prior existence of such systems and upon the interest

of other people in elaborating them through appropriation: for

example, by them reacting to scribbles as an effort at

representational drawing. Finally, the pupil's experience of social

communication prompts retrospective sense-making along just such

intended lines.

It is this overarching *contextualising* feature of appropriation that

distinguishes it from the other two categories of instructional talk

discussed above: those were more motivated by assisting strategic

control of activity at the moment-by-moment level. Newman et al

regard the appropriation process as 'a "stand-in" for the child's

self-discovery' (p.142). This indicates the shift of balance in

cultural theorising away from more pupil-focussed models of learning

characteristic of constructivist perspectives. Now it is assumed that

socially-organized practices are central to the learning process.

However, this is not to undermine the creative dimension of

intellectual development. Creativity exists within the appropriating

social interactions themselves, and the new functional systems that

result do equip the learner for self-discovering opportunities - as

might well be the case for the example of drawing.

Modest exchanges of the kind illustrated in this section are widely

reproduced within classroom encounters. They depend upon both

partners appropriating the activities of the other - acting as if they

were all "somewhere else". The somewhere else, of course, is generally

some approximation to that place where instruction is carefully

leading. Newman et al note an intriguing paradox in this process:

'for a lesson to be needed, in say, division, it must be presumed that

the children cannot do division; but, for the lesson to work, the

presumption is that whatever the children are doing can become a way

of doing division!' (op. cit., 64).






I have sketched here three varieties of instructional talk that have

been identified by Psychologists of a socio-cultural persuasion.

These ways of talking support three possible kinds of instructional

influence. Firstly, *internalization* is associated with the

creation of joint problem solving formats, or cognitive systems.

These allow the novice to witness and participate in more advanced

strategic cognition on a public, inter-individual plane. Secondly,

prolepsis illustrates a form of *semiotic mediation* available as a

conversational device to prompt the novice into private cognitive

reflection. Thirdly, *appropriation* is a related device whereby

(perhaps over a more extensive section of talk) a collaborator will

act "as if" a partner's intentions and motivations matched their own.

This 'strategic fiction' (Newman et al, 1989) allows them each to act

as if their partner's behaviour was locatable within their own goal

structure; thereby achieving more effective direction and coordination

of purposes.

Of course, these distinctions do not exhaust the possibilities

for instructional talk. Direct explanation, or exchanges that

encourage passive modelling may well be adequate to promote learning -

although, perhaps learning of more modest scope. The point is that

talk of the kind identified above comprises what is typically found

when local circumstances have been chosen that are most favourable for

full concentration on the business of "teaching" (cf. Edwards and

Mercer, 1987). The effects of such talk on pupil achievements is hard

to evaluate cleanly (Schaffer, 1992). However, there is empirical

evidence that, when variation is studied, the features discussed above

are indeed a powerful basis for supporting learning (Freund, 1990;

Rogoff, 1990; Smagorinsky and Fly, 1993; Wood, 1988).

So. three cultural approaches to systematising instructional talk have

been summarised here. Commentators may differ in which they stress.

I believe that the three are mutually compatible and that the

distinctions are necessary. But it would be valuable to attempt some

integration: to show that they share a common conceptual core. I

shall explore this possibility in the remainder of the present

section; suggesting that integration is possible through an appeal to

the peculiarly human capacity for *intersubjectivity*. This is a

concept of great topical interest to psychologists; it may identify an

inter-individual achievement that is uniquely human (Humphrey, 1976).

Particular interest is apparent among developmental psychologists who

are concerned to trace its ontogeny (Astington, Harris and Olson,

1988), as well as the consequences of its disturbance (Frith, 1989).

A parallel interest is shared by linguists, who are concerned to

understand how it is maintained within routine conversation.


The nature of intersubjectivity


I shall use the work of a linguist to introduce the idea here. In a

series of influential articles, Rommetveit has discussed

intersubjectivity in the course of characterising the management of

everyday communication. He expresses it in relation to the

interpersonal business of creating shared reference:

A state of intersubjectivity with respect to some state of affairs "S"

is attained at a given stage of dyadic interaction if and only if some

aspect "A(i)" of "S" at that stage is brought into focus by one

participant and jointly attended to by both of them (Rommetveit,

1979a, p.187)

This identifies the concept with a state of mutual understanding

and encourages us to regard cognition as something that may be studied

as "socially shared" (see Resnick, Levine and Teasly (1991) for a

cross section of reactions to such an idea). However, Rommetveit's

definition is not entirely satisfactory here. It orients us in one

particular direction - towards a kind of common object that can be the

shared reference point of communication. The object metaphor is

slightly mysterious, although useful for some purposes. I shall

return to it later in this chapter. What I prefer to stress here are

the social psychological processes through which we create and sustain

such common objects of attention. Rommetveit implicates them in

defining a state of 'perfectly-shared social reality'. Such a state

exists at a point of communication:

...if and only if both participants at that stage take it for granted

that "S" is "A(i)" and each of them assumes the other to hold that

belief (op. cit., p.187).

This mutual recognition of a partner's understandings is more what I

wish to highlight here: the registration by one person that

particular mental states (particular beliefs or intentions etc.) exist

within another person. This registration has a recursive character.

I may acknowledge that you have certain understandings; you may

acknowledge that I have made that acknowledgement, and so forth. Such

mutual projection of mental states may be exploited to finely tune the

communication between us. As Davidson (1992) puts it:

Sociality and rationality combine to produce curiosity about what is

in others' minds and motivation to formulate a *fit* between one's own

thoughts and the thoughts of respected others - in other words, to

create intersubjectivity (p.31).

This human concern for mutual recognition of mental states offers the

basis for integrating our three categories of instructional talk. I

believe that each of them depends upon an ability (and inclination) of

both teachers and pupils actively to mobilise intersubjectivity. Of

course, the 'motivation to formulate a fit' in this spirit may

sometimes have to be skilfully provoked and encouraged. In any case,

on this analysis what "instruction" turns out to involve is the

skilful deployment and organization of human intersubjective


This claims more than the simple truth that instruction should take

into account the current (cumulative) state of a learner's knowledge.

Making such a static characterisation of what-is-known is very much

the kind of achievement that computer-based tutoring systems might

attempt - rather as if knowledge was merely a commodity to

be inventoried. The claim for intersubjectivity within instruction

refers to efforts that go beyond this. In particular, it identifies a

capacity for interpreting mental states that will arise during the

dynamic of the tutorial exchange itself. Teachers will need to be

good at this; for the investigative activities of their pupils are not

always transparent for interpretation - not always predictably

strategic. Instead, what pupils are seen to do may often appear

uncertain, volatile and improvisatory. Thus, interpreting learners'

momentary psychological states can be inherently difficult; and it

does need to be sustained "on-line", in the course of evolving,

situated interactions.

Moreover, interpreting a learner's situated intentions, beliefs and

motives is only the first part of an effective instructional

intervention. For, on the present analysis, the act of instruction

that follows such an interpretation is not always some

straightforward, interjection: some inevitable consequence of the

learner's situation as it has been evaluated. Effective instructional

support is not always a question of teachers successfully formulating

their interpretations of what learners are doing and, then, delivering

efficient, unambiguous direction. Such a forumula-derived approach to

interventions (optimising them) might be the very strategy adopted by

a programmed tutor. Yet, as illustrated above, the most effective

interventions may not be those that are optimised. The best thing

for a teacher to say at some chosen moment may not be the thing that

is maximally informative. Instead, the remarks that may be most

helpful are those that are studiously chosen to be incomplete or

otherwise imperfect; chosen because they are provocative of further

engagement by the learner. Such cultivated imprecision seems to be

exactly what is found by researchers who have closely observed

tutorial discourse under ideal conditions. The apparently laudable

and precise patterns of feedback, correction, diagnosis and

demonstration that ITS designers strive to achieve do not seem to

characterise what expert tutors actually do (Lepper, Wolverton, Mumme

and Gurtner, 1993).

Human tutors seem to do something slightly different and, I suggest,

their achievements that result depend upon an inherent capacity for

intersubjective understanding. So, successful practice must depend

upon a *motivation* to mobilise this human capacity and upon some

capacity to deploy it skilfully. Such skill will derive from

histories of interacting with learners in various domains of

knowledge. It may also derive from a teacher's own experiences, at

other times, of being a pupil. The whole process can be said to

depend upon the 'projective work of the imagination' (Harris, 1991).

At the root of all that might be achieved is a distinctive mutuality:

the engaged pupil will complement what the motivated tutor does in the

course of some supportive intervention. That is, the pupil is active

in recognizing and reacting to the tutor's interpretative attitude and


Each of the three modes of instructional interaction described above

may now be examined within this framework. Take the case of

interventions in support of cognitive internalization. A sensitivity

(in both partners) to ongoing events will be necessary to create joint

activity that captures the interplay of a genuine "cognitive system".

From the tutor's viewpoint, the demands of the task and the timing of

support must be set to generate real collaboration; thereby, affording

the novice opportunities for internalization. So, judgement will be

exercised by the more expert collaborator; judgement will define the

exact points at which strategic intervention would help. Sensitivity

to the longer-range history of what the pupil knows may certainly be

relevant. But so also will be sensitivity to the task-in-progress: to

what the novice has experienced and attempted, to what they know of

the task "at that moment". At the same time, the pupil must interpret

the tutor's interventions as, somehow, being about what they

collaboratively are trying to achieve (perhaps resolving feelings that

an intervention may not seem in harmony with those shared intentions

as they are inferred at the given moment).

A similar analysis applies to the process of semiotic mediation as it

occurs within instruction. The underspecified, presupposing talk that

characterises prolepsis will work when judged well enough to bring

novice partners to a sort of "precipice" of understanding - where it

is possible for the necessary cognitive reflection to be successfully

precipitated within them. This is surely embraced by Rommetveit's

(1984) remarks on speaking and listening (in the context of defining


...encoding and decoding are complementary processes. Encoding

contains always a component of anticipatory decoding and decoding

takes the form of reconstructing fragments of an intended message


The teacher's trick here is to encode in a way that enriches the

learner's work of decoding - the trick of effective *anticipatory*

decoding. Note, again, this is not necessarily a question of being

maximally clear and informative: a degree of opaqueness may be what is

important in stimulating decoding work that is creative. In sum,

there is a distinct sensitivity that must be mobilized during this

kind of instructional exchange - a social commitment that lifts the

exchange above more didactic forms of instruction to become something

more like a "collaboration". It is a sensitivity that depends upon

mutual recognition of intention, motive, belief and understandings.

Finally, the concept of appropriation clearly indicates the necessity

of intersubjective relations within instruction. Appropriation

involves a conspiracy: the strategic fiction of a coordinated task is

created before the learner actually has an authentic sense of that

task. As Newman et al (1989) put it:

..children can learn new goals and ways of doing things when their

responses are appropriated into a system of which they were not

previously aware. Because the teacher interacts with the child...the

child can learn retrospectively what his response means in the system

as understood by the teacher (p.142)

This detective work evidently presupposes an active curiosity about

the intentions and understandings of the other. Discussing the

origins of intersubjectivity in infancy, Newsom (1978) appears to be

discussing the very same phenomenon of appropriation (although,

without identifying it as such):

..only because mothers impute meaning to 'behaviours' elicited

from the infants, is it that these eventually do come to

constitute meaningful actions so far as the child is concerned


So, I am suggesting that the various ideas from cultural psychology

regarding instructional dialogue can be usefully synthesised by

reference to the notion of intersubjectivity. This may now help us in

relation to the topic addressed at the start of this chapter: the

viability of reproducing within computer-based learning the

traditionally social processes of instruction. By this I mean the

ambition to program a computer to simulate a particular form of

dialogue: the potential for talk that bears some resemblance to what

learners normally enjoy with their human teachers.




Intersubjectivity and computer-based tutoring

I have been discussing the ways in which intersubjectivity saturates

orthodox instructional talk. I believe that this feature of such talk

undermines the possibility of comprehensively simulating it on

machines. This is not to deny that sometimes we find parallels

between achievements inspired by engagement with a computer, and

achievements prompted by human tutors. So, sometimes it might be

helpful to claim that a computer is acting as a "scaffold" for a

pupil's learning (cf. Hoyles and Noss, 1987). The discussion of

semiotic mediation above predicts this possibility: it leads us to

expect that the programmed structure of a computer activity could be

provocative or supportive of a pupil's constructive efforts during

independent learning. Such a computer program could therefore be said

to scaffold the learning. However, while such piecemeal (and,

perhaps, teacher-orchestrated) impacts are real enough, they are not

unique to this particular technology: other structured problem solving

environments may be provocative in a similar fashion. Thus, while

such computer-based contributions to the support of learning are

important, they do not correspond to the simulation of an authentic

instructional intervention.

The hope for such simulation probably springs from a certain way of

conceptualising human communication. For example, a pioneer of

intelligent tutoring remarks: ' interaction is basically

a communication between two information structures' (Carbonell, 1970,

p.194). The problem is not that this is inherently mistaken - it

depends on our particular reading of key terms - but it is a way of

framing communication that easily encourages certain other

perspectives that do seem wrong. So, it might suggest that the task

of a cognitive science is to implement data structures that correspond

to the configuration of any given human "information system". Then,

on this view of things, communication between that person and any

other such "system" (including a computer) might be readily handled by

a set of rules governing interaction (rules for information transfer).

Clancey has summarised the important theoretical contrast that this


The view that knowledge is stored suggests that interactions between

people are structured and controlled by pre-exisiting structures

stored in the head. The opposing view is that neural and social

structures coordinating our behavior come into being during our

interactions. (Clancey, 1992, p.148)


It is intriguing that the information-storage metaphors of the first

view are so appealing to us as ways of characterising communication

and instruction. Even the most distinguished of cognitive

psychologists may feel compelled to capture an exasperation about

teaching and learning by deploying this mechanical imagery: 'I find

it terribly frustrating, trying to transfer my knowledge and skill to

another human head. I'd like to open the lid and stuff the program

in' (Simon, 1983, p.27). Perhaps our exposure to means of

transferring material in the physical world encourages such imagery.

Moreover, the movement of data within information technology is

readily expressed in the language of "stuffing it in". Yet this will

not do: the currency of education is different. Knowledge is not so

neatly circumscribed as to allow complete and unambiguous stuffing

under some human lid. The problem has been discussed at some length

in an influential thesis by Suchman (1987). She comments about the

meanings we exchange (within any act of verbal communication):

..the communicative significance of a linguistic expression is always

dependent upon the circumstances of its use...the significance of an

expression always exceeds the meaning of what actually gets said, the

interpretation of an expression turns not only on its conventional or

definitional meaning nor on that plus some body of presuppositions,

but on the unspoken situation of its use (p.60).

This "situated" nature of human communication is fully explored by

Suchman. She argues that action's *inherent* uncertainty requires

that we turn from simply explaining it away 'to identifying the

resources by which the inevitable uncertainty is managed' (p.69). A

significant claim is that these resources 'are not only cognitive, but

interactional' (p.69). However, just what are these 'interactional'

resources that must be mobilized during routine communication?

They are resources that arise from the dyadic, *in situ* character of

communication. Suchman includes the prosodic and temporal structures

of talk that give it an ensemble quality; also the informal

understandings that define specialised rights and agendas within the

conversational ritual. However, she also refers to the 'local

coherence' or relevance that conversational partners invariably create

within the sequential organization of their talk. Less is said about

the mechanisms for controlling this coherence, but I suggest it is

only made possible by the "intersubjective attitude" that has been

discussed in this section. Again, in terms of the conceptual

vocabulary developed here, Suchman's account of communication is one

in which 'interpreting the significance of action is an essentially

*collaborative* achievement' (p.69, my emphasis). In terms of the

purposes of education (creating shared knowledge), and the tutorial

methods employed (instructional discourse), Davidson (1992) expresses

the position we are reaching here:

...different individuals invent similar answers to a given problem.

The intersubjective attitude supports this inventive process because

it enlivens curiosity about possible discrepancies between one's

beliefs and those of others (p.34)


My analysis implies that interactions *with* computers can not

reproduce an at-that-moment richness of dialogue that characterises

teacher-led instruction. Debating the possibility of such simulation

brings into focus a deeply-rooted theoretical difference. On the one

hand, we find developments guided by traditional cognitive

psychological perspectives: that is, conceptions of knowledge as

stored representations, with thinking as involving their manipulation.

On the other hand, we find the sobering influence of socio-cultural or

situated theories championing an opposing view.

..all processes of behaving, including speech, problem-solving, and

physical skills, are generated on the spot, not by mechanical

application of scripts or rules previously stored in the brain.

(Clancey, 1991, 110).

In cases that are being increasingly studied, knowledge and the

structures of situations are so tightly bound together that it seems

best to characterize knowledge as a relation between the knowing agent

and the situation, rather than as something that the agent has inside

of himself or herself. (Greeno, 1989, p.313)

My own discussion here has concentrated on the implications of

this opposing position for the goal of reproducing instructional

discourse. Earlier in this Chapter, I catalogued distinctive features

of such talk. These features surely draw our attention towards an

imbalance in the interactional resources available to a human pupil and

a tutoring computer. Investigating this asymmetry underpins the

agenda of situated theories in this area: '(to) locate the limits of

that sense-making ability for machines in the limits on their access

to relevant social and material resources, and [to] identify the

resulting asymmetry as the central problem for human-machine

communication' (Suchman, 1993, p.73).

So, I have stressed that instructional talk seems to be a

collaborative, situated achievement: one founded upon human

intersubjectivity. My view is that this excludes its comprehensive

simulation within tutoring systems. However, the challenge to

machine-based tuition runs deeper than this issue of reproducing

instructional dialogue. For, the underlying theoretical tension I

have explored above is also relevant to other assumptions guiding the

design of computer-based tutoring.

The ITS designer traditionally has focused on three problems. How to

model tutorial dialogue is certainly one. The others concern how to

model some domain of knowledge (i.e., what experts know), and how to

model what is currently known by a given learner. The idea that

either novice or expert knowledge might be captured in this

computational form is evidently encouraged by traditional cognitive

psychology. So, ITS researchers will express the target of their

modelling thus: 'Much of what constitutes domain-specific

problem-solving expertise has never been articulated. It resides in

the heads of tutors, getting there through experience, abstracted but

not necessarily accessible in an articulatable form' (Sleeman and

Brown, 1982, p.9). This characterisation surely exemplifies the

influence of certain cognitive psychological theories: theorising of a

kind (sceptically) characterised by Winograd and Flores as supposing:

'Knowledge is a storehouse of representations, which can be called

upon for use in reasoning and which can be translated into language.

Thinking is a process of manipulating representations' (Winograd and

Flores, 1986, p.73).

The alternative to this has thinking as situated activity.

Individuals are constantly responding to a dynamic environment:

engaged in a dialectic with the material and social world.

"Knowledge" thereby becomes an activity, not a storehouse to be

replicated. It is always a creative construction within the

here-and-now: guided by past interactions, but shaped by demands of

the moment. The situated approach does not thereby deny the

possibility of cognitive modelling - the ambitions of creating

artificial intelligence. However, the methods appropriate for

creating certain circumscribed *artificial* intelligence may not be

the same methods appropriate to modelling *human* intelligence itself

(Norman, 1991). So, the achievements of AI to date - while often of

real practical value - are typically brittle and inflexible. A

situated theory predicts limitations within any design enterprise

where symbol structures are created to describe functional relations

in only a narrow domain. Instead, it encourages modelling of the

intelligence manifest in behaviour that is *adaptive* towards the

environment: the capacity for responding to circumstances as they

arise - simply "dealing with" the world (Sterling, Beer and Chiel,

1991). If we must strive to construct artificial intelligences,

then this might imply starting with such "open systems" as those

required merely to get around physical space: computational insects

perhaps (Beer, 1990).

Such an alternative perspective, when adopted for intelligent tutoring

machines, does not encourage designs-for-learning grounded only in

computed databases of knowledge in catalogue form (although such

inventories may be a real resource to refer to within a broader

learning context). Neither, as was argued above, does it encourage

attempting to reproduce the social interactions that constitute

instruction. In short, it does not encourage pursuit of comprehensive

computer-based tuition.

This more situated view of cognition can claim some converts. Certain

early researchers in the ITS tradition recently have altered their

approach towards marrying up computers and education. For example,

Brown, Collins and Duguid, (1989) have sought new inspiration from

studying learning within informal settings: out-of-school learning.

Some software developers are particularly interested in the conditions

of learning from apprenticeship relationships - such as have recently

been documented by culturally-oriented researchers (eg. Rogoff, 1991;

Rogoff and Lave, 1984). Attention to these out-of-school achievements

encouraged Brown et al (1989) to pursue a new approach to the

deployment of new technology *in* school. Their basic idea is that

the versatility of computer-based environments (particularly

simulations and microworlds) can offer a rich repertoire of authentic

situations in which pupils' thinking can flourish and develop. This

approach is guided by the cultural tradition of theorising both

because it emphasises access to mediational means, and because it

views that access as a situated achievement. These researchers remain

concerned to cultivate the abstract modes of thinking that schooling

has always pursued, but they challenge any notion that this must be

cultivated within relatively context-free tasks. Thus, they promote a

strategy whereby the power of new technology can be directed towards

furnishing a rich variety of contexts - situations - in which the

learner can interact.

Yet, the role of social interactional processes within this so-called

"cognitive apprenticeship" strategy remains neglected. Where it has

been addressed, it seems to invite a retreat to former traditions of

ITS design that focus on computer-based dialogue. So some researchers

in this tradition are now considering how a computer can be programmed

to prompt and intervene - but as an apprenticeship master rather than

a tutor (cf. Katz and Lesgold, 1993). There may be a place for such

initiatives. But the arguments above apply: the social nature

of tutorial dialogue will not be reproduced wholesale. Certainly,

there should be no implication that such social interactional

encounters do not arise within the out-of-school settings of cognitive

apprenticeships. It should not be supposed that processes of

intersubjectivity are irrelevant to these informal learning

arrangements - just because there is no "teaching" going on. Far from

it: an untutored achievement such as mastering one's native language

may be so very impressive simply because it *is* organized within a

rich framework of intersubjectivity (Bruner, 1983; Bernstein, 1981)





These various doubts about prospects for replacing the interpersonal

basis of learning should not be read as part of another sweeping

rejection of computers in education. For one thing, existing ITS

programs can remain a useful and proven resource (Anderson, Boyle,

Corbett and Lewis, 1990). There is no reason to doubt that they have

a valuable niche within in a broader context of instructional support.

So, situated theorists themselves may continue to be architects of

such systems (eg., Clancey, 1988). The point is to question whether

they represent truly comprehensive alternatives to traditional,

socially-grounded structures for learning.

More generally, the value of designing sophisticated computer-based

learning environments is certainly not in question here. However, we

do need to be clear about where the most significant increases in

sophistication can be achieved. The cultural analysis of cognition

has implications for where the creative effort of design might be best

concentrated. Reproducing tutorial dialogue may be an area where some

progress can be made but, I suggest, progress will be limited and

striving for it may not be the most cost-effective way forward. What,

then, is a better way? In trying to respect the social character of

educational experience, we should not suppose that creating

opportunities for interaction *with* a computer is the only option.

Given the central place of these social processes in instruction and

given their subtle nature, I would encourage a move away from design

strategies based exclusively upon interacting *with* computers,

towards solutions that consider computers as a *context* for social

interaction. Our aims then would be different. They would no longer

be directed towards displacing instructional interactions. They would

be more concerned to establish how computer activities can serve as an

*occasion* for classroom discourse: a setting in which certain kinds

of potent socially-organized experience can be arranged. We would be

turning our attention towards the social interactional possibilities

that the physical presence of this technology affords.

In the remainder of this book, I wish to review this possibilitiby in

respect of various configurations of interaction that the technology

may support. The first concerns interactions involving both teachers

(experts) and pupils (novices). Possibilities considered in later

chapters concern interactions between pupils themselves (novices with

novices). In all cases, I will refer back to the central place of

intersubjectivity and socially-shared understanding as introduced


The theme of the next Chapter concerns socially-based instruction as

it might exist in harmony with computer-based activity - rather than

being supplanted by it. This harmony entails social interactions

occurring *in relation to* computers (rather than, in some contrived

sense, *with* them). On such occasions, the exchange between teacher

and learner is retained as central to the educational activity.

However, it is an exchange that is not governed by computers, but

catalysed or mediated by them. The underlying computer-based

experience may still involve forms of pupil-machine dialogue. But now

the technology becomes a focus for a parallel interaction: joint

activity that teacher and pupil organize between themselves. An

encounter with the computer is, thus, assimilated into the broader

social fabric of educational activity. This may seem a

straightforward arrangement, but I shall argue that attending to what

needs to be done (within social interaction) under these common

circumstances is easily neglected - given the dis-located style of

working that computers naturally encourage.