In 1990, a conference was held at Lucy Cavendish College, Cambridge,
to explore the prospects for a new school of research, cognitive archaeology.
The fruits of that conference are now published; they may be uneven in quality
but they are provocative. Archaeology at its best is detective work that
rivals anything in science or fiction – from Crick and Watson to Holmes
and Watson. At its worst, it is imagination run wild, underconstrained speculations
which often have the added vice of permanently distorting the data through
erroneous ‘restorations’, or just spuriously authoritative labels that then
make alternative interpretations of those objects and sites all but unthinkable.

It is hard to resist the gravitational pull of a good story, apparently,
especially when one has just spent a long hot summer and a sizable grant
(or a lifetime and a fortune) painstakingly wresting an unprepossessing
pile of ancient leftovers from the earth. One has to make something from
these fragments – if not the lost city of Atlantis, then at least some exciting
conclusions about the exotic habits, beliefs or rituals of the people who
made them. So it is not surprising that the early romantic excesses of archaeology
– Agamemnon’s tomb and all that – provoked a positivistic reformation movement.
Parallel to the behaviourists’ efforts to turn their field of psychology
into hard science, the ‘processual’ school of archaeology demanded scrupulous
data gathering and forbade all but the most rigorously constructed interpretations,
echoing Lloyd Morgan’s Canon of Parsimony: ‘Thou shalt not impute more Mind
than is strictly necessary to account for the data.’ One could venture cautious
conclusions about the diet, tools and building materials, and the size of
the groups, but precious little else – next to nothing about what or how
these ancient people thought.

It is not surprising to learn that the killjoy strictures of the processualists
have been challenged by several quite different groups. One, the inevitable
ghastly gang of postmodernist relativists, has simply declared that since
there aren’t really any facts anyway, you might as well tell whatever stories
move you. Members of the other, I am relieved to say, are even more opposed
to that devil-may-care nonsense than they are to the extreme puritanism
of the processualists, and hope to bring the pendulum to a halt in an intermediate
position. They are the would-be founders of cognitive archaeology, claiming
that new methods and insights from cognitive science make it possible to
deduce more facts about ancient minds than the processualists would allow,
without falling into humanistic fantasy.

Of course, even the most severe processualist was always ready to draw
some inferences about ancient minds. The identification of an object as
a figurine (esc-hewing such grander questions as whether it is a toy or
a totem or a token in a counting system) already ascribes considerable mental
activity to the people who made the object. This can be readily gauged
by considering how much it would take to persuade us that some roughly
ape-shaped object found among apes was a figurine of their making.

The archaeologists’ debates have always been about thresholds or standards
– not about whether, but about when and why one can claim that a stone is
an artefactual blade, or a pile of bones is a ritual burial site and not
just a pile of trash. So how does cognitive science help to clarify these
debates – or move their midpoint mindwards?

Very little, on the showing of this volume. In spite of a great deal
of throat-clearing and loin-girding by the various authors, the only substantive
message from cognitive science that I could see is the valuable but vague
caveat that cognitive competence is not always general-purpose but may
instead be strongly restricted to specific topics, or specialised ‘modules’.
No archaeological evidence of specific modules (pot-making modules, agricultural
modules, copper smelting modules?) has been uncovered, but it is indeed
wise for archaeologists to beware of over-generalising about the minds of
their subjects; a brilliant fish-hook maker might be a dunce in almost every
other regard.

What does get a salutary workout is not anything specific to cognitive
science, but just good old-fashioned logic – in particular, the exercise
of reverse engineering to make deductions about the likely uses to which
things have been put in the past, and why. A paradigm case discussed by
several authors is Colin Renfrew’s deduction that the carefully crafted
cubes found at Mohenjo-Daro were calibrated weights used in a system of
measurement of enough importance to the society to warrant the specialisation
of labour that must have created them. Another is Charles Frake’s level-headed
reinterpretation of the ‘wind’ dials of ancient navigators, illuminating
both the Pacific islanders’ oft-misinterpreted system of celestial navigation
and the naming of the Mediterranean winds by different cultures.

What is crucial to any such interpretation of human behaviour based
on artefacts is the assumption that the person who crafted the object would
not have gone to such lengths to make these things if they didn’t strongly
believe that they worked. People have long valued nonfunctional decoration
for its own sake, but if people have devoted the bulk of their lives to
making doodads (are they weapons? calculating devices? culinary tools?)
or a single great thingumabob (a fort? a temple? a storehouse?), they presumably
thought, rightly or wrongly, that there was a pressing requirement to make
such a thing. So if one cannot show that the artefacts did perform some
valuable function, one is left having to explain how their makers could
have been so convinced of a falsehood.

At this point I detect serious confusion on the part of at least some
of the contributors to this volume. They have a tendency to reserve ‘cognition’
for such elevated or ‘cultural’ topics as religion, ritual and style of
government, as opposed to such mundane practicalities as agriculture and
self-defence – as if one could farm or hunt or build a shelter without cognition,
but needed cognition to engage in ritual when burying the dead.

Allied with this is the surely anachronistic tendency to contrast religious
practices with ‘functional’ practices. To our eyes, the systematic placement
of carefully conserved seeds into the ground in the spring is not a ritual,
while the systematic placement of ancestors bones into the ground on some
other occasion is. But this is only because we know the former ‘works’ and
the latter, presumably, does not.

The people who engaged in both practices made no such distinction. For
them a sacrificial altar and a dry storehouse were equally functional, equally
essential protections against the vicissitudes of nature. Presumably these
people really believed in the efficacy of what they were doing; they were
not, like many of today’s masters of ceremony, just ‘keeping a tradition
²¹±ô¾±±¹±ð’.

As the anthropologist Dan Sperber has noted, it is the false beliefs
of any culture that are of most interest to social scientists, for unlike
a true belief, a false belief requires a special genealogy: how did this
error continue to be propagated so securely? Here, perhaps, a behaviourist,
not a cognitive scientist, has provided the best clue. In a classic series
of experiments, Richard Herrnstein put pigeons in Skinner boxes on random
reinforcement schedules. Soon the pigeons were obsessively producing bizarre
ritualistic bobbing and weaving sequences, which in time were reinforced
by random bits of food reward. ‘It’s working!’ we can imagine the pigeons
exulting; ‘I knew if I just twisted my neck a little more to the left this
time, I’d induce the pellet-god to give me another one!’ Herrnstein aptly
called these pigeon dances ‘superstitious behaviour’, but he eschewed putting
any such imagined thoughts in the pigeons’ heads.

Random reinforcement is still the best explanation we have for how elaborate
and costly (and completely inefficacious) rituals could get started in the
first place, but of course once they do get started, they become highly
efficacious – for the group of priests or kings or others who make a handsome
living keeping the rituals going, and elaborating on them.

The processualists shunned deductions about ideology, while welcoming
deductions about practical life; if cognitive archaeology has a future,
it will be by showing how, under the right conditions, one can extrapolate
facts about ‘ideological’ features of ancient cultures by showing how they
are the likely or even obligatory extensions of the practical concerns
that shape all cognition.

Daniel Dennett is at the Center for Cognitive Studies, Tufts University,
Medford, Massachusetts.

In 1956, the mathematician John McCarthy coined the term ‘artificial
intelligence’ for a new discipline that was emerging from some of the more
imaginative and playful explorations of that new mind-tool, the computer.
A few years later he developed a radically new sort of programming language,
Lisp, which became the lingua franca of AI. Unlike the sturdier, stodgier
computer languages created by and for business and industry, Lisp was remarkably
open-ended and freewheeling. Instead of concentrating on numbers, it was
designed to take any symbols or strings of symbols (lists) as its objects,
and because its own machinery consisted of just such lists (and lists of
lists), Lisp creations easily inhabited the very world they acted upon,
and hence could reflect upon themselves and their own reflections indefinitely,
revising and reinventing themselves, breaking down the artificial barrier
between program and data.

Seymour Papert was one of the most playful of the AI pioneers, and more
than any of the others, his own reflections turned to the nature of that
playfulness and its role in learning and discovery. In 1980, he published
Mindstorms, in which he presented his Utopian vision of computers in the
classroom, of which the centrepiece was Logo, a dialect of Lisp that he
and others had developed specifically for very young children.

The key element of his design was Turtle graphics, an inspired interface
that made children’s interactions with Logo not just visible, but instantly
comprehensible – feelable, you might say. The tales he told of those early
encounters were compelling. These became an important ingredient in the
barrage of persuasive literature that led teachers and schools all over
the US, and indeed the world, to invest huge sums in ‘computerising the
classroom’. Thousands of teachers tried their hand at Logo in the classroom,
with mixed results.

I was one of them. About ten years ago, I was part of a team that developed
and taught an introductory course in computer science aimed at university
students who hated and feared computers but whose parents, in many cases,
had said ‘You must learn about computers before you graduate.’ These students
were seasoned veterans of what Papert calls School – experts at piling on
the facts, drilling for the big test – and they were pathologically uncomfortable
in any setting where they had to think. (My impression then was that many
of them, given a choice between solving a simple puzzle and memorising two
pages of the phone book, would gratefully choose the latter project.) I
had read Papert’s book and discussed its ideas with him at length. Logo
was designed for five-year-olds, so it might, I thought, be just right for
my university students. In fact, it was spectacularly successful. The students
forgot their phobias and inhibitions and took flight, creating a trove of
idiosyncratic projects, effortlessly learning the fundamentals of programming,
and building a robust base on which we could then help them to construct
a more ‘adult’ set of edifices. So I am one of the many who can personally
attest that Logo, in the right circumstances, does work wonders.

The right circumstances are hard to come by, alas. In the years since
Mindstorms, Papert has participated in a host of projects, large and small,
designed to implement his ideas, and has received a wealth of feedback,
much of it deeply discouraging. But one can learn even more from ‘mistakes’
than from a string of successes – that is a central tenet of Papert’s vision
of learning, and he practices what he preaches. So this sequel engagingly
recounts what he has learned, and especially the mistakes he made along
the way. His own thinking has undergone a transformation; he is still an
infectiously optimistic visionary, but a wiser one.

Logo has now joined forces with Lego, the plastic building blocks, and
a new wave of delectable settings for learning has been created and explored.
Papert favours parables; in these pages you will find no statistical surveys
of the effects wrought by Logo-Lego projects and no data on control groups
with which to compare the inspiring tales he tells. Indeed, at several points
he says that he is quite sure that no test could be counted on to reveal
by before-and-after comparison the benefits he has anecdotally conveyed.
As you read further you come to realise that this is not a dodge on his
part, but a fundamental implication of his message: the sort of testing
required by such attempts to measure success destroys the very conditions
for learning that the computer, at its best, can create.

There is a recurrent pattern that has bedevilled many – in the end,
almost all – attempts to use computers in education; no sooner are the classrooms
equipped than the dead hand of educational bureaucracy begins to impose
conditions that systematically squander the power the equipment promised
to deliver. Of all the teachers who enthusiastically took up the themes
of Mindstorms, ‘many felt seduced and abandoned by the talk of a computer
revolution as the use of the computer became routinised’. Instead of railing
at the system or blaming the teachers or administrators, Papert looks at
the broader problem, and sees that it springs from deeply held – and of
course ill-examined – assumptions about the point of School.

Consider the awkward confrontation you would inevitably create were
you to implement one of Papert’s fundamental principles of learning: take
your time. Learning happens best when you can browse around in a problem
space, savouring the shapes, fiddling with the bits and pieces, twiddling
the knobs – but always, always, taking your time. So there you are, noodling
away contentedly and constructively when the bell rings, and you have not
finished your assignment. Well? What sort of payoff does society want? And
how soon? It is always a nagging problem.

Lying behind this and other confrontations is an assumption that Papert
calls the Gothic cathedral model of learning. Suppose educating a child
were like ‘building a Gothic cathedral out of 40 000 blocks of stone. Clearly,
strict organisation is needed to perform such a task. One cannot have individual
workers deciding that they want to put a block here or there just because
they are inspired to do so.’ Instead of conceding that education must, after
all, be a building process conducted under the wisest set of controls we
can muster, Papert takes on the task of persuading us that education is
not that sort of task at all. We must dare to let go – not always, but as
much as we can bear. We must adopt a more ‘systemic’ view of learning as
a variegated family of ’emergent’ phenomena. This is a truly radical idea,
which Papert cannily makes more palatable by an analogy. His revolutionary
proposals would create a market economy of educational experimentation,
in contrast to today’s traditional educational hierarchy, which he likens
to the Soviet Union’s disastrous planned economy: ‘. . . while our economic
system, with all its faults, is above the threshold of functionality and
theirs was below it, our education system falls on the same side of the
line as the Soviet economy’. The timing of Papert’s proposal could not be
more ticklish: just as the US is finally coming to grips with the spiralling
cost and diminishing effectiveness of a health-care system that has been
shaped by the pressures of a market economy, he encourages us to make a
leap of faith, and trust in the distributed wisdom of many local experiments
under minimal control: ‘The Rigorous Researcher will object to the populist
tone of this argument. It is appropriate to buy a food processor or garlic
press on the basis of individual whim, but education is more serious. Every
child deserves the best. Science should be used to find out what is the
best, and then everyone should adopt the proven methods.’

Papert has a good retort. What makes the Rigorous Researcher think there
is a best way? Herbert Simon, another of the founders of AI, has built his
distinguished career around showing the importance of Voltaire’s maxim:
‘The best is the enemy of the good’, and Papert would replace the hyper-rationalist
quest for the best – which has given us a series of stultifying, straitjacketed
systems of education – with a free-roaming, ‘cybernetic’, feedback-guided
tracking of the good. Many will be tempted to dismiss Papert’s proposed
revolution as a trendy amalgam of politically correct themes, unabashedly
fes-tooned with the buzzwords of feminism, empowerment, multiculturalism
and anti-elitism – Dionysus si! Apollo no! and all that. But in fact he
has given these themes a much more unified, circumscribed and cogently reasoned
justification than one generally encounters.

And then there are the details. There is just no escape from his shrewd
analyses of the crushing obstacles to learning that stand untouched by traditional
pedagogy. To my mind, the most important of these analyses are his observations
on the crippling taboos against self-exposure. We learn in School to conceal
our own ignorance and confusion, and this not only inhibits us from exploring
the very moves that would be crowned with success, but saps our self-confidence.
With no fund of shared experience of screwing up, we are apt to harbour
wildly unrealistic fantasies about the intellectual prowess, the clarity
and rigour, of our teachers and peers. This Victorian prudishness about
our own cognitive disabilities is built right into the fundamental structure
of School, and it spawns a host of secondary effects, all debilitating.
There are some wonderfully liberating passages in which Papert, as usual
practising what he preaches, describes his own confusions, false starts
and insecurities, and then recounts the childlike solutions he discovers.
We should all be such children.

I am a believer, but the adult in me says that responsible scepticism
has still not been met halfway. Take a few thousand children and feed them
a lovingly prepared diet of Logo-Lego, and it is not surprising that you
can harvest a few – or even a few hundred – inspiring success stories. What
happened to the other children? I don’t think anybody knows, at least not
in detail. Presumably they were at least not harmed in any way, but were
there enough benefits to justify expanding these programmes? When the thrill
of pioneering has to be traded in for year-in, year-out practices that can
be reliably adopted by thousands of less inspired teachers, will the same
wonderful effects be in evidence? Papert is cautiously optimistic, and he
suggests that in any case the current systems are so toxic for so many children
that we risk little by taking the leap, but he offers no blueprints and
no guarantees – it would fly in the face of his radical message to offer
either. For the time being he is content to make it clear that there must
be feedback if these experiments are to sort themselves out in a desirable
direction, but he has little to say about how to design the appropriate
feedback mechanisms. We can hope that his next book will have more to say
on this important topic.

Daniel Dennett is the cofounder of the Curricular Software Studio at
Tufts University in Massachusetts, where he is Distinguished Professor of
Arts and Science. He is the author of Consciousness Explained (1992).