ARE WE ALL NATURAL DUALISTS?
A COGNITIVE DEVELOPMENTAL APPROACH*
RITA ASTUTI
London School of Economics & Political Science
The article takes to task the well-established anthropological claim that non-Western peoples
are free from the traps of dualistic thinking.AlthoughVezo informants in Madagascar produce
statements that could be used to support such a claim, experimental procedures that target
their inferential reasoning reveal that they systematically differentiate between mind and
body, between the biological processes that determine the organism and the social processes
that shape personhood.This suggests that there is a significant discrepancy between people's
explicit linguistic statements and their implicit theoretical knowledge. Moreover, develop-
mental data show that such implicit theoretical presuppositions are essential to the pro-
duction and transmission of cultural knowledge. Thus, Vezo children, who do not as yet
differentiate between the biological mechanism of birth and the social mechanism of nurture,
are still unable to grasp a salient aspect of Vezo culture, namely the causally integrated set
of ideas that guide the way adults classify the social world. These findings have significant
theoretical and methodological implications for the constitution of anthropological
knowledge.
In an essay written to commemorate the Malinowski centennial year in
1984, Firth (1985: 37) remarked that `cognitive anthropology is essentially
an inferential study', since anthropologists infer informants' ways of think-
ing from linguistic expressions. Anthropological inferences are supported by
the common and not entirely implausible assumption that `language provides
clues to thought'. However, having noted the complexity of the relations
between thought and language, Firth argued that anthropological `assertions
about cognition ... should be treated as tentative, not tested summations of
mental process'. Since anthropologists work with `gross, crude analytical tools
involving wholesale inference, much speculation, and some introspective
analogy', he concluded that their `generalisations should be accordingly
modest'.
Anthropologists cannot be said to have heeded this advice, as they routinely
infer patterns of thought from patterns of speech, and make generalizations
that can hardly be described as modest. Here, I will address one of anthro-
pologists' favourite assertions about cognition: the ubiquitous claim that non-
Western peoples are free from the traps of dualistic thinking, that their way
of reasoning and of representing the world is not cast within the Cartesian
dichotomy of mind and body.
 Royal Anthropological Institute 2001.
J. Roy. anthrop. Inst. (N.S.) 7, 429-447
* Malinowski Memorial Lecture, 2000

It has been remarked that social science theorizing has a widespread ten-
dency to vilify Cartesian dualism as a kind of moral abjection (Csordas 1994:
7). There is also confusion about what exactly the domain for dualism is, as
discussions freely shift from the domain of mind vs. body proper to other
domains, such as conscious vs. unconscious, thought vs. emotion, object vs.
subject. In anthropology, of course, the mind-body dichotomy has been cast
in terms of a dichotomy between culture and biology, and the many deriva-
tives of this, such as the distinction between sex and gender, person and organ-
ism, individual and society.There are, therefore, as many anthropological claims
of `monism' as there are versions of dualistic reasoning.1
The specific one I will consider here has emerged in the field of kinship
and gender studies, and is arguably one of the most robust and uncompro-
mising versions of the widespread anthropological claim that non-Western
peoples are not dualists.The claim, more precisely, is that non-Western peoples
do not draw the distinction between the biological facts of procreation and
their cultural interpretation, between biological and social parenthood, between
the natural body and its cultural meanings, between nature and culture, sex
and gender.
The battle in support of this form of non-Western `monism' has been
fought on many grounds, theoretical as well as empirical. On the theoretical
front, anthropologists (e.g. Bouquet 2000; Strathern 1996) appear recently to
have found comfort in Latour's (1993) revelation that we, Western peoples,
have actually never been modern, and that we have worked as hard at creat-
ing hybrids of nature and culture as we have at separating the two. On the
empirical front, many have followed in Schneider's (1984) footsteps and have
argued that non-Western understandings of relatedness and of gender identi-
ties are not predicated on the distinction between the biological and the social
(e.g. Carsten 1995; 1997;Yanagisako & Collier 1987). It is fair to say that this
idea has become one of the most guarded assets of the discipline, giving its
practitioners an edge in a number of philosophical and political debates (for
dissenting positions, see e.g. Astuti 1998; Lambek 1998; Middleton 2000;
Ortner 1996). By way of illustration, let me quote a passage from an influ-
ential essay by Ingold.
In the context of a general philosophical argument against various dualis-
tic fallacies that besiege the social sciences and, more specifically, against the
foundational separation between social and biological domains in kinship
theory, Ingold (1991: 362) notes that `many (if not most) non-Western peoples
... simply do not recognize anything comparable to the social/biological dis-
tinction as articulated in western discourse'.
If given a weak interpretation, this claim is both true and trivial. That the
distinction between the social and the biological, like that between nature and
culture or mind and body, is not articulated in other cultural traditions as it
is in the Western one is granted, and has been repeatedly proven. But the
claim is far stronger, and is far from trivial: namely that
for them, the human being is not a two-part entity ­ half organism, half person ­ but a
single, undivided and embodied centre of action and awareness, an organism-person, that
neither simply `grows' of its own accord, nor is `made' like an artefact, but is rather grown
through the active contributions of many people, including those who may be designated
as `parents' (1991: 362).
430 RITA ASTUTI

The claim is that most, if not all, non-Western peoples do not distinguish
between the biological and bodily processes that make an organism grow and
the social and mental processes that shape personhood, between the biologi-
cal processes of reproduction and the social processes of nurture, between bio-
logical and social parenthood.
In this article I shall challenge this idea. I shall do so in the same spirit in
which Gell (1998: 126-7) has argued that ordinary human beings, whether
Western or not, are `natural dualists', naturally predisposed to split the person
into two: the mind inside, the body outside. Like many of the anthropologists
who have argued the opposite, Gell was more than aware of the philo-
sophical pitfalls of dualism. However, he reminded us that the task of anthro-
pology is not to prescribe notions that are philosophically defensible, but
`to describe forms of thought which could not stand up to much philosophi-
cal scrutiny, but which are nonetheless socially and cognitively practicable'
(1998: 17). It follows that describing `natural dualism' as a socially and cogni-
tively practicable form of thought does not amount to endorsing its philo-
sophical validity. Similarly, describing how `ordinary human beings' distinguish
between organism and person, between birth and nurture, between the
biological and the social, does not amount to endorsing the philosophical or
metaphysical validity of any of these distinctions.The issues of description and
of philosophical validity are separate and should be kept separate. Indeed, as
I see it, the real challenge for anthropologists is not to collect examples of
sound philosophical thinking among the people they study, but to establish
less crude analytical tools to generalize about the way ordinary human beings
think.
I shall thus endeavour to describe how one group of non-Western peoples,
the Vezo of Madagascar, reason about the distinction between birth and
nurture, organism and person, mind and body. Vezo ethnography provides
plenty of evidence that could be used to support the anthropological claim
that non-Western peoples do not draw such distinctions. For example, Vezo
find the suggestion that birth parents have exclusive claims over their children
morally problematic, and they express this view in a variety of ways. Thus,
they seemingly do not `see' resemblances between parents and their offspring.
Instead, they invoke a variety of mechanisms other than birth to explain the
way babies come to resemble people other than their biological parents. For
example, if a pregnant woman spends a lot of time talking to a friend, the
friend is said to `steal' some of the baby's facial traits; if she takes a strong
dislike to someone, her baby will look like the disliked person. The expecta-
tion that babies will come to resemble people other than their parents,
has the effect of socializing parenthood,2
and of extending the child's bodily
connections beyond those with its birth parents.
On the basis of this kind of evidence, it would be an easy task to show
that Vezo do not distinguish between organism and person, birth and nurture,
biological and social parenthood. I shall argue, none the less, that the kind
of evidence that I have just invoked ­ which is the kind of evidence
routinely used by anthropologists ­ does not support the inference that
Vezo have successfully escaped the traps of dualistic reasoning. By means
of an empirical investigation of informants' inferential reasoning, rather
than relying on their explicit normative statements, I shall demonstrate that
RITA ASTUTI 431

Vezo adults in fact draw the distinction between the biological processes
that determine the organism and the social processes that shape personhood.
And although what I offer here is but one carefully examined case, the
evidence that I present raises broader questions about the foundations of
anthropological knowledge. However, I also aim to move the argument
forward. I shall do so by extending the investigation to Vezo children in order
to pursue new questions about the relationship between implicit and explicit
knowledge.
The adoption task
The evidence that I shall be presenting is the result of the extensive use of
an experimental technique borrowed from developmental psychology.What I
shall describe below as the `adoption task' was part of an investigation of North
American children's intuitive understanding of birth and of the role that birth
plays in the transmission of properties from parents to offspring (Solomon,
Johnson, Zaitchik & Carey 1996).The task was originally devised to establish
at what age children develop an adult-like understanding of the transmission
of properties from parents to offspring. In agreement with previous studies
(e.g. Springer 1992; Springer & Keil 1989; 1991), Solomon, Johnson, Zaitchik,
and Carey noted that North American children as young as 3 or 4 years of
age systematically predict that offspring resemble their parents ­ for example,
that blond parents have blond children ­ and set out to probe children's under-
standing of the mechanisms responsible for the resemblance. In particular, they
wanted to establish at what age North American children adopt the view, as
adults do, that resemblance to parents involves two separate causal mechanisms
for the transmission of two ontologically distinct traits of the person. In other
words, at what age do children adopt the view that birth is the mechanism
responsible for transmitting characteristics such as bodily traits, and that nurture,
learning, and habituation are responsible for transmitting other characteristics
such as mental traits?3
The adoption task they devised pitted biological against social parentage,
and the transmission of bodily traits against the transmission of mental traits
such as beliefs; that is, things that are held to be true.They found that, up to
the age of 7, North American children are as likely to predict that birth trans-
mits both bodily and mental traits, as they are to predict that nurture trans-
mits both traits. These children, therefore, do not differentiate between birth
and nurture, bodily and mental traits.
In Madagascar, I used an adapted version of this task to establish the pres-
ence or absence of dualistic reasoning among Vezo adults and children. In its
simplest form, the Malagasy adoption task consisted in telling participants a
story about a baby born to one set of parents and raised by another:
The birth parents ­ described in Vezo idiom as `the father and mother who generated
the child' (baba sy neny niteraky azy) ­ are travelling through the forest; they are attacked by
bandits; they have time to hide the baby under a bush; the parents are killed but the baby
survives and is found and raised by another set of parents; the adoptive parents ­ described
as `the father and mother who raised the child' (baba sy neny niteza azy) ­ nurture him with
affection and love.4
432 RITA ASTUTI

The story was followed by questions which asked participants to decide
whether the grown-up child would resemble the birth or the adoptive father
on a number of different traits: bodily traits, beliefs, and skills. A control task
was devised to establish whether participants might reason differently depend-
ing on whether the link of filiation in the questions was paternal or mater-
nal; since, interestingly, there was no evidence of a systematic effect, in what
follows I ignore this variable.
The questions were formulated as follows:
The father who generated the child had pointed ears, whereas the father who raised the child
had roundish ears. In your opinion, when the child is fully grown up, will he have pointed
ears like the father who generated him, or roundish ears like the father who raised him?
and
The father who generated the child believed that chameleons have 30 teeth, whereas the
father who raised the child believed that chameleons have 20 teeth. In your opinion, when
the child is fully grown up, will he believe that chameleons have 30 teeth like the father
who generated him or will he believe that chameleons have 20 teeth like the father who
raised him?5
Note that the task set out an artificial problem, a riddle, to which partici-
pants were asked to find a solution.The aim was to keep the riddle as culture-
neutral as possible. For example, the story was devised in such a way that it
did not evoke the social and moral setting in which adoption normally takes
place among Vezo, and the traits for the resemblance questions were chosen
because they were not culturally salient: for example, nobody knew or cared
to know how many teeth chameleons have,if any.Overall,the task was designed
to force participants, as it were, to put their thinking cap on and leave their
cultural narratives to one side. Thus, the task did not require participants to
articulate verbally their customary ways of explaining things, nor to provide
an abstract account of how properties are transmitted from one generation
to the next. In the case of children, this was a clear advantage, since they might
have been unable to put their knowledge into words; in the case of adults, the
advantage was that the riddles did not directly prime their stock cultural knowl-
edge but required them to engage in inferential reasoning.
It could be argued that the adoption task imposes on the participants the
ontological categories of the researcher, the dualisms of sociality and biology,
of organism and person, of mind and body. However, even though the task is
undoubtedly constructed around these distinctions, it does not force them on
the participants. If participants do not distinguish between social and biologi-
cal parenthood, between birth and nurture, between bodily and mental traits,
they will sail through the task blissfully unaware of the distinctions being
probed.
The adoption task has four possible outcomes, defined in terms of each
participant's pattern of judgement. Participants are said to show a Differenti-
ated Pattern if they judge that the adopted child would resemble the birth
father on bodily traits and the adoptive father on beliefs.Their response implies
a distinction between two chains of causal mechanisms, birth and nurture,
which transmit two distinct traits of the person, bodily traits and beliefs.
RITA ASTUTI 433

Participants are said to show a Birth Parent Bias if they judge that the
adopted child would resemble his birth parent on all or almost all traits; they
are said to show an Adoptive Parent Bias if they judge that the adopted child
would resemble his adoptive parent on all or almost all traits. Participants who
show either a Birth Parent or an Adoptive Parent Bias draw no theoretically
driven distinction between birth and nurture as separate mechanisms for the
transmission of bodily and mental traits.
Finally, participants who judge the adopted child to resemble birth and
adoptive parents in a random combination of traits are classified as showing
a Mixed Pattern. They are most evidently clueless about the distinctions
probed by the task.Their pattern of judgement reveals that they do not reason
causally about the mechanisms of birth and of nurture; neither, of course, do
they distinguish between the two.6
`Dualistic' adults and `monistic' children
Let me now introduce the results of the adoption task used with male and
female Vezo adults, aged 21 and above.7
Although most of them were initially
bemused by my questions, mid-way through the task, when they realized that
only half of my questions were about the boy's body (vatany), while the other
half were about his mind (sainy), their faces would light up in recognition.
This moment of recognition gave rise to the following results: an over-
whelming 78 per cent of adult participants showed a Differentiated Pattern
(Fig. 1). Of the remaining 22 per cent, 13 per cent showed a Mixed Pattern,
6 a Birth Parent Bias, 3 an Adoptive Parent Bias.
434 RITA ASTUTI
FIGURE 1. Adults' patterns of judgement.

The overwhelming majority of Vezo adults, therefore, drew the distinction
between birth and nurture.They reasoned that birth is the mechanism respon-
sible for the transmission of bodily traits, whereas nurture is the mechanism
responsible for the transmission of mental traits. Their inferences ­ for
example, that the adopted boy will resemble his birth father in the shape of
his ears, but will resemble his adoptive father in the belief about chameleons'
teeth ­ were guided by a theory about the different character of mind and
body, and about the distinctive causal mechanisms affecting them. By contrast,
Vezo adults did not invoke the `monism' that informs their explicit cultural
statements, such as the explanations used to account for babies' physiognomy.
Had they done so, they would have reasoned that nurturing relations mould
every trait of the person, mental as well as bodily.
The first conclusion to be drawn from these findings, which converge with
those of Bloch's (Bloch, Solomon & Carey 2001) investigation among the
Zafimaniry, is that there is a significant discrepancy between what Vezo adults
say, the enunciations that get transmitted and are part of what anthropologists
would identify as `Vezo culture', and the theoretical presuppositions that
people mobilize to reason inferentially, irrespective of their cultural narratives.
In other words, there is a discrepancy between explicit linguistic statements
and implicit theoretical knowledge.
This finding gives support to the argument that informants' ontological
commitments cannot be inferred from their explicit discourse: it follows that
anthropologists cannot advance claims about the absence of dualistic reason-
ing in non-Western societies based on the interpretation of explicit enuncia-
tions, as is most commonly done. The result has far-reaching methodological
implications and casts doubt on some of the most attention-demanding claims
made by anthropologists about personhood, gender, kinship, the body, sub-
jectivity, and so on. My aim, however, is to put forward a rather more
challenging proposition for anthropologists: we cannot understand what our
informants explicitly tell us (let me loosely call it `culture') if we ignore the
theoretical cognitive underlay of their statements.
There is an obvious, logical reason for this. If it is true that what people
say and enshrine in their explicit cultural statements can diverge from the
implicit theories that they mobilize to reason inferentially, it follows that one
cannot correctly understand their statements without knowing how these
articulate with their implicit theories. For example, statements that appear to
blur the distinction between birth and nurture, body and mind (let me call
them `monistic statements'), would reflect people's ontological commitments
if they are uttered in the context of a monistic ontology.The same statements,
however, would have to be interpreted and understood differently if they are
uttered in the context of a dualistic ontology, in which case they contradict
people's ontological commitments. Needless to say, the meaning of an
utterance will be seriously misjudged if contradictions are interpreted as
reflections.
The logical reason why anthropologists need to take into account the
implicit cognitive underpinning of their informants' statements has been noted
before (e.g. Bloch 1998; Boyer 1994; Sperber 1985; 1997).There is, however,
a separate developmental reason which has remained unexplored, and to which
I now turn.
RITA ASTUTI 435

The hypothesis that I shall pursue below is that the production and trans-
mission of the conceptual systems that anthropologists deduce from their
informants' explicit cultural enunciations and describe in their monographs ­
about personhood, relatedness, identity ­ presuppose the emergence of the
theoretical differentiation between birth and nurture, mind and body found
among Vezo adults.
I shall explore this hypothesis by means of a research strategy pioneered
by Toren (1990; 1999a; 1999b), which consists in turning one's attention to
children, to ask how they come to make sense of the way adults conceptual-
ize the world in which they live. The specific question I shall ask is this: do
Vezo children distinguish, like Vezo adults, between birth and nurture, body
and mind? The answer, in short, is that they do not; and this discovery makes
it possible to compare, within the same population, a group that does not have
a dualistic ontology (the children) and one that does (the adults).
The comparison allows me to pursue a more general question about the
relationship between adults' implicit ontological distinctions and the explicit
cultural statements that get transmitted to their children. As we shall see, in
the absence of a dualistic ontology, children not only fail to understand, and
are unable coherently to reproduce, such statements, but they also fail to
understand why and how they have come to be what they are.
I shall begin by examining the results of the adoption task carried out with
Vezo children of both sexes aged 6 to 13.8
As illustrated in Figure 2, children
performed very differently from adults.
An overall 88 per cent did not show a Differentiated Pattern of judgements.
Of these, 40 per cent showed a Mixed Pattern, 30 per cent a Birth Parent
Bias, and 18 per cent an Adoptive Parent Bias. The Differentiated Pattern,
which was 78 per cent among adults, was instead a mere 12 per cent among
the children. Children were thus as likely to judge that either birth, or nurture,
or both affect the way people appear in both their bodily and their mental
traits. Unlike adults' inferences, those of children were not guided by the
theoretically driven differentiation between birth and nurture.Therefore, each
436 RITA ASTUTI
FIGURE 2. Children's and adults' patterns of judgement.

answer was settled on the basis of a variety of ad hoc reasons: simple guess,
preference for one set of parents over the other, preference for individual traits,
and so on. In short,Vezo children did not see the distinctions drawn by the
task. They thereby fulfil the anthropological dream of a population with a
non-dualistic ontology.
The hypothesis I raised earlier is that the capacity of Vezo adults to dif-
ferentiate between birth and nurture, body and mind, is a prerequisite for
the production and transmission of their cultural knowledge. To test this
hypothesis, I now turn to a particularly salient cultural trait: the set of ideas
that specify what it means to be Vezo, and how some people come to be `us'
while others come to be `them'. I do this to explore whether children, who
do not differentiate like adults, are able to understand and reproduce what
their parents teach them about their social surroundings.
Vezo social classification
The criteriaVezo adults use for classifying people are explicit and can be sum-
marized as follows: people are what they are because of the place where they
live, which in turn determines the activities they perform; group affiliation,
as well as the difference between `us' and `them', are established through the
actions that different people perform in their physical and social environment.
This way of classifying the social world is typically articulated by adults in a
variety of informal contexts which portray the contrast between Vezo people
and their immediate neighbours, the Masikoro. Adults constantly and consis-
tently invoke the principle that being Vezo is to live on the coast and do Vezo
things, such as fishing, sailing, and eating fish, whereas being Masikoro is to
live inland and do Masikoro things, such as cultivating, raising cattle, and ­ so
Vezo say ­ eating `grass' (for a detailed ethnographic account, see Astuti 1995a;
1995b).
This definition of what it means to be Vezo or Masikoro, and the fact that
people who are Vezo can become Masikoro and vice versa, implies that the
process of social classification is not mediated by birth. Birth is not the mecha-
nism that transmits group affiliation between generations: one does not need
to be born Vezo or Masikoro to be one; all one needs is to learn to do Vezo
or Masikoro things.
Vezo social classification is thus based on activity criteria: Vezo use the
performance of certain activities as the criterion for classifying people.
However, before extending this conclusion to the classification of other social
groups, I must pre-empt a potential source of confusion. It could be that Vezo
use activity criteria to distinguishVezo and Masikoro people only because they
regard both groups as sub-units of the same super-ordinate ancestral group.
In other words, they might be invoking occupation as a criterion for identi-
fication and they might be disregarding people's ancestral origins because they
assume that Vezo and Masikoro people share the same origins to begin with.9
Thus, to establish that Vezo use activity criteria to classify other social groups,
one must be able to show that Vezo informants extend that way of reasoning
to people of unquestionably separate ancestral origins. The counterpart I
chose for testing the hypothesis were the Karany. Karany are immigrants of
RITA ASTUTI 437

Indo-Pakistani descent. They are town dwellers; they are wealthy and well
educated; they are shopkeepers and moneylenders. They are predominantly
Muslim, and although fluent in Malagasy, they speak their own language
among themselves. Somatically, they differ from the Vezo in that their skin is
light and their hair straight. In Western folk-sociology,Vezo and Karany would
be regarded as separate racial groups.
The preliminary issue that needed to be established was therefore this: do
Vezo adults reason about Karany in the way they reason about Masikoro? Do
they classify Karany in terms of what they do rather than in terms of their
ancestral origins? In sum, do the activity criteria for social classification extend
to ancestrally unrelated social groups?
To explore this question, I used a new version of the adoption task in which
the adoption occurs across the Vezo-Karany divide: a boy is born to one set
of parents, either Vezo or Karany, and is adopted by parents of the opposite
group. Participants were asked the standard set of resemblance questions about
bodily traits, beliefs, and skills, but were also asked about a number of con-
trastive traits that are considered typical of Vezo or Karany people, such as
their occupation and religious beliefs. For example: will the adopted boy know
how to fish (Vezo typical) or to keep shop (Karany typical)? Will he believe
that corpses must be kept overnight in the village (Vezo typical) or will he
believe that one has to bury them straight away (Karany typical)? Finally,
I asked the crucial question about the child's identification: would he be
Vezo or Karany? As a control, I also ran an identical task in which adoption
occurred across the Vezo-Masikoro divide.10
The adults' results from the two versions of the task are presented in Figure
3. The performance on the Vezo-Masikoro version was unsurprising: as
expected, 87 per cent of adult participants judged that the adopted child, irre-
spective of birth origins, would grow up to be Vezo if raised by Vezo parents
in a Vezo village, and Masikoro if raised by Masikoro parents in a Masikoro
438 RITA ASTUTI
FIGURE 3. Judgements of social classification by adults.

village. The child was judged to have acquired all the traits typical of his
adoptive group.
Similarly, if more strikingly, 75 per cent of adult participants in the Vezo-
Karany version of the task also judged that the adopted child, irrespective of
birth origins, would grow up to be Karany if raised by Karany, and Vezo if
raised by Vezo (Fig. 3). As before, on virtually all traits, the child was judged
to resemble his adoptive father.
This is a remarkable result. The fact that adults reasoned about Karany in
the same way in which they reasoned about Masikoro vindicates the general
claim that Vezo make use of activity criteria to classify people into social
groups. More importantly, it shows that Vezo adults reason about social
categories coherently, drawing on a consistent, integrated theory of social
classification. The cross-over along the Vezo-Karany divide is not an occur-
rence that my informants have ever experienced in real life, nor do they ever
contemplate it in their conversations. Thus, unlike participants in the
Vezo-Masikoro task whose replies came spontaneously and without effort,
participants in the Vezo-Karany task had to respond to an entirely novel ques-
tion. They did so by asking themselves: is the adopted boy going to acquire
the traits typical of the people he grows up with? The answer to this ques-
tion was straightforward because they knew that traits such as occupational
skills or religious beliefs are acquired by living in a certain environment. It
followed that the boy would be Vezo or Karany like his adoptive father,
because what people turn out to be is determined by the typical traits that
they acquire through practice. In other words, adult participants abstracted a
set of theoretical principles from a set of known circumstances, and extended
such principles to make predictions about as yet uncharted social territories.
It is precisely this ability to mobilize implicit theoretical knowledge to
make novel inferences that children still lack. Children who grow up in Vezo
villages are bombarded with messages that people are either Vezo or Masikoro
because of what they do. Children hear adults discuss how Vezo ways differ
from Masikoro ways; they are frequently told that they are Masikoro when
they fail to act in Vezo fashion, and that they are becoming Vezo when they
succeed. The idea that being Vezo or Masikoro depends on what people do
is thus `in the air'; it is therefore to be expected that Vezo children will absorb
it and be able to reproduce it.
To establish this fact and, more significantly, to establish whether children
can apply the abstract theoretical principles underlying this idea to unexpected
circumstances, I extended the two adult tasks to the children.
The most striking finding is that children, unlike adults, performed very
differently in the two versions of the task. Thus, 65 per cent of the children
judged, like adults, that the adopted child will be Vezo or Masikoro like his
adoptive father, irrespective of birth origins (Fig. 4). By contrast, in the Vezo-
Karany task only 27 per cent of the children gave adoptive judgements; 73
per cent of them reasoned instead that the child will be like his birth father,
irrespective of where he grew up. In this case, the overwhelming majority of
children reasoned very differently from adults.
Let me first turn to the task in which children were asked to reason about
the Vezo-Masikoro cross-over, where 65 per cent of them gave adoptive
judgements. One might conclude from this that the majority of Vezo
RITA ASTUTI 439

children have an adult-like understanding of how to classify people. However,
two aspects of their performance indicate that this is not so. Both relate to
forms of logical inconsistency.
In the first place, children who judged that the boy would be Vezo or
Masikoro like his adoptive father did not extend this judgement consistently
to typical traits, such as occupational skills and religious beliefs. Having stated
that the boy would be Masikoro like his adoptive father, for example, the same
children stated that he would know how to sail and fish like his birth father.
They thus failed to comprehend the chain of causal mechanisms (i.e. place of
residence and specialized activities) that underlies the adults' theory of social
classification.11
Secondly, children applied the standard adult explanations, that people are
Vezo or Masikoro because of where they live and because of what they do,
to justify both adoptive and birth judgements.Thus, for example, children who
judged that the boy would be Vezo like his birth father, explained that this is
because the boy knows how to fish and sail, despite the fact that he was raised
in the interior surrounded by cattle and cornfields.12
In sum, these inconsistencies suggest that children, irrespective of their
judgements, were simply mimicking statements that they did not fully com-
prehend. Thus, even children who appeared to give the `right' sort of
answer ­ that is, the same answer as that given by adults ­ display no under-
standing of the theoretical underpinnings implied by their judgements.
Because they grow up hearing that people are Vezo if they fish or Masikoro
if they cultivate, they are able to reproduce these statements when primed by
familiar questions. But children's grasp of the adult theory stops here, well
short of the consistency that characterizes the way adults classify the social
world.
Indeed, children's inconsistency is the key to understanding their response
to the Vezo-Karany task. In this task an overwhelming 73 per cent of chil-
dren judged that the adopted boy will be Vezo or Karany like his birth father.
440 RITA ASTUTI
FIGURE 4. Judgements of social classification by children.

Thus, unlike adults, children did not extend to Karany the same activity cri-
teria for social classification normally applied to Vezo and Masikoro people.
This confirms that such principles are still beyond their grasp. Children did
not mobilize and extend their knowledge of how people come to be Vezo
or Masikoro, because their knowledge is still atheoretical and dependent on
specific statements they have heard in specific contexts. Children were unable
to resort to theoretical principles and to reason analogically about Karany,
because, as yet, they do not have a causally integrated theory of social classi-
fication to underpin what they say.13
To summarize the argument so far: first, I used the adoption task with Vezo
adults and children to establish the presence or absence of dualistic reasoning.
The results showed that adults differentiate between birth and nurture, bio-
logical and social parenthood, body and mind; children do not. Secondly, I
examined a salient cultural trait, namely the way adults and children classify
people, and thereby establish the criteria that make `us' different from `them'.
The comparison between adults' and children's performance in two tasks, one
that targeted familiar social distinctions and one that targeted entirely novel
ones, revealed that Vezo adults rely on a coherent and integrated theory. By
contrast, children do not.
And so, what makes adults' classification consistent and children's inconsis-
tent? Or, to ask the same question from a developmental, rather than static,
perspective: how is the adults' coherent, integrated theory of social classifica-
tion constituted and acquired by children? The standard anthropological
answer to questions such as `How do children become competent adults?' and
`How do they acquire their culture?' is that this occurs as children gradually
learn to see things the adult way, as a result of growing up in their cultural
environment. For example, as they grow up, Vezo children are exposed to a
culturally specific way of classifying people, and they gradually assimilate it
and learn to reproduce it.There is clearly some truth in this.Were Vezo chil-
dren to grow up in a different cultural environment, they would undoubtedly
come to classify people very differently; for example, by assuming that social
identity is fixed at birth as people inherit from their forebears the essence that
makes them what they are.
The comparative perspective I have adopted here, however, enables me to
ask rather different questions and to suggest different answers. The questions
are: what is it that adults have and children do not that allows `the adult way
of seeing things' to emerge? Or, to put it the other way around: what do
children not have that would allow them to compose the fragments of knowl-
edge that are `in the air' into an integrated set of consistent propositions?
The results of the Vezo-Masikoro task have shown that there is quite a lot
that Vezo children learn directly from adults' statements, and learn relatively
early.When they are teased because they are Masikoro or praised because they
are becoming Vezo, they learn that being Vezo and Masikoro has something
to do with what they do. Anecdotal evidence suggests that children who are
actively engaged in Vezo activities, such as fishing or trading fish, are in fact
more likely to answer the question `Why are youVezo?' in typical adult fashion
by showing the scars that the fishing line has left on their hands, or by boast-
ing about the profits they have made at the market, rather than by saying, as
less experienced children do, that they areVezo because their parents are.There
RITA ASTUTI 441

is, however, a limit to how much children can learn in this direct way. The
limit, set by what they do not know, stops them from composing into a theory
the fragments of knowledge that they absorb from Vezo adults.
The coherence and predictive potential of Vezo adults' theory rests on a
chain of causal mechanisms linking place of residence and activity to social
classification. At a more abstract and foundational level, what guides and gives
coherence to adults' social classification is the theoretically driven differentia-
tion between two separate causal mechanisms, birth and nurture, for the trans-
mission of ontologically distinct traits of the person.Adults have a general rule
of thumb for classifying people: where people live and what they do will tell
you who they are. However, they extend this simple rule by adding the capac-
ity to differentiate between traits that are fixed at birth and traits that are
transmitted through nurture, learning, and habituation. It is this differentiation
that allows them to classify people by systematically disregarding the former
traits (for example, skin colour), and focusing solely on the latter (for example,
occupational skills).
By contrast, children have, at best, learnt some isolated facts: that certain
people are Vezo because they live on the coast and fish, and others are
Masikoro because they live in the interior and cultivate. Because children's
factual knowledge is not yet integrated with the theoretical presuppositions
that underpin the way adults classify people, this knowledge is applied
randomly and inconsistently; consequently, it has neither explanatory nor
predictive power. Children do not appreciate the difference between birth
and nurture as causal mechanisms for the transmission of ontologically dis-
tinct traits of the person.Therefore, from the fact that people areVezo because
they fish, children are unable to infer the general conclusion that people must
be classified into social kinds by means of traits transmitted through nurture.
This inference is beyond children's conceptual understanding, because it
presupposes prior dualistic ontological commitments that they, as yet, do not
make. Only once they make such commitments does their knowledge of
certain isolated facts about the social world become a coherent and predic-
tive theory of the social world.
Conclusion
Vezo children have taught us a few general lessons. I have shown that their
ability to produce and reproduce coherent cultural knowledge (such as the
Vezo system of social classification) depends on implicit theoretical presuppo-
sitions (such as the differentiation between birth and nurture). This conclu-
sion has important implications for our discipline.
Anthropologists routinely deduce their informants' conceptual systems
(concepts of the person, of the self, of relatedness) and their ontological
commitments (the absence of the social-biological distinction) from knowl-
edge that their informants communicate to them and that ostensibly contains
everything that anthropologists need to know. In so doing, however, anthro-
pologists are in danger of grasping only fragments, like children, and of missing
out on the presuppositions that underpin adult knowledge and that reconsti-
tute those fragments as coherent, predictive theories.
442 RITA ASTUTI

Independent support for the argument that comprehension of much of
what is communicated, to children and anthropologists alike, relies on implicit
theoretical assumptions, comes from pragmatics, the study of how speech
utterances are interpreted by the hearer. Just as the comprehension of the
simplest linguistic behaviour relies critically on non-linguistic presuppositions,
so cultural communication, which involves fantastically complex utterances,
must do the same.
Vezo children, therefore, have taught us that Malinowski was right relent-
lessly to search for a new scientific method that would free the trained
researcher from the deceptions of what he called the question-and-answer
method. Malinowski was mainly concerned about the fact that this method
elicited informants' normative statements and therefore distorted the under-
standing of practical social life. He seemed less concerned about its effect
on the understanding of cognitive processes, perhaps because he assumed that
they were universal, and because he knew no existing method that would
satisfactorily reveal them. Participant observation, great linguistic proficiency,
a systematic method for recording ethnographic data, are some of the method-
ological tools that we have inherited from Malinowski.
As I have shown, these tools are essential but not sufficient to describe our
informants' forms of thought. If we wish to provide such descriptions accu-
rately, we should be prepared to advance our methodology by co-operating
with other disciplines such as cognitive psychology, thereby keeping to the
innovative spirit with which Malinowski created our discipline.
NOTES
This article is based on research in Madagascar generously funded by the ESRC (Grant
R000237191) and the Nuffield Foundation. It is the result of a collaborative project between
anthropologists and cognitive psychologists: Maurice Bloch (LSE), Susan Carey (NYU), Gregg
Solomon (MIT), and myself. Susan Carey and Gregg Solomon originally devised the experi-
mental procedure, which Bloch piloted among the Zafimaniry of Madagascar (Bloch, Solomon
& Carey 2001), and which I have used extensively among the Vezo, with both children and
adults. Carey and Solomon provided the necessary training in experimental methods and helped
adapt the procedures to Malagasy circumstances; to this, I added my long-standing relationship
of mutual trust with Vezo villagers, and they added their goodwill and open-mindedness.
Solomon undertook the statistical analysis of the data, and he are co-authors of a forthcoming
publication that presents in detail the data on which I base my analysis. As to the practical
administration of various versions of the experimental procedure to over 180 participants
over the course of five months, I must acknowledge the help of a very patient and, when it
came to children, indispensable research assistant: my son, Sean. Finally, Lorenzo Epstein has
given me support and inspiration every step of the way. I thank him for this and much more.
I dedicate this article to my parents, Carlo and Fernanda Astuti.
1
For example, anthropologists working on emotions have claimed that non-Western
ethnopsychologies do not separate emotion from thought (Lutz 1988); others have described
dividual constructs of the person (Marriot 1976; Strathern 1988) that dissolve the dualism
of individual and society; many others have documented a variety of sociocentric or multiple
conceptions of the self which have provided the ethnographic backing for theorizing the new
epistemology and metaphysics of the `mindful body' (Scheper-Hughes & Lock 1987).
2
On the significance of such process of socialization, see Astuti (2000).
3
This question was raised in response to other researchers' conclusion that Western children
as young as 3 or 4 have a biological understanding of the transmission of properties from parents
to offspring; in other words, that they understand that offspring inherit from their biological
RITA ASTUTI 443

parents the innate potential to develop certain traits rather than others (e.g. Springer 1992;
Springer & Keil 1989; 1991). A dissenting voice (Carey 1995) pointed out that this conclusion
was not granted, since to claim that pre-schoolers have a biological understanding of inheri-
tance is to claim that they know, at a minimum, that birth is the causal mechanism implicated
in the resemblance. That birth causes resemblance is certainly part of Western adults' under-
standing of inheritance, but the evidence does not grant the inference that this is also the case
for young children. Indeed, the latter may have no causal explanation at all for outcomes that
they are none the less able to predict, such as the fact that blond parents have blond children;
or they may invoke alternative causal mechanisms (such as love or the simple desire to be like
their parents) that have nothing to do with birth. To conclude that children understand the
transmission of properties from parents to children as a biological phenomenon, one would
need to demonstrate that the children themselves explain the resemblance between parents and
their offspring by invoking causal mechanisms that are uniquely biological. The adoption task
was devised with this aim in mind.
4
The story was composed so as to avoid the conflation of the birth-nurture contrast
targeted by the task. For example, the story emphasized that the two sets of parents were unre-
lated and unknown to each other, and that the child grew up having no contact with the birth
parents or with any of their relatives. The death of the birth parents reduced the likelihood
that the participants would assume that the child had any further interaction with the birth
parents. A further reason for introducing the death of the birth parents was to distinguish the
story presented in the task from other, culturally salient stories about babies being abandoned
in dangerous locations in the forest because they had been born on days that are inauspicious
according to local divination systems. If such abandoned children are found alive, their inaus-
piciousness is thought to have no effect on their adoptive parents. When Vezo informants tell
stories of such instances, they point out that once the child is grown-up and ready to be of
help, the birth parents will claim the child back (and the child typically refuses to follow them).
The story used in the task was thus devised to reduce the likelihood that participants would
understand the birth parents as having any lingering claim over the child.
5
The study was balanced across participants according to a Latin-Square design in order to
control for the potential confounding factors of whether the bodily traits were presented before
or after the beliefs (skills were always introduced last), and which value of a pair of features
was attributed to the birth parent. Preliminary analyses conducted on these factors revealed no
significant effects on the results discussed here.
6
The results presented are based on the administration of a version of the task in which the
adoption occurs between two sets of Vezo parents. In this version, participants were presented
with a total of twelve traits: four bodily traits, four beliefs, and four skills. Participants were said
to have shown a Differentiated Pattern if they judged the boy to resemble the birth father on
at least three of the four bodily traits and on none of the four beliefs, or on all four bodily
traits and on not more than one of the four beliefs. A participant has a probability of .04 of
showing such a pattern by chance. As, according to Western folk biology, there is no reason
that skills should be considered a priori to be biologically inherited or not, they were not
considered in determining the Differentiated Pattern. Participants were considered to have
shown a Birth Parent Bias if they judged the boy to resemble the birth father on at least ten
of the twelve features, and to have shown an Adoptive Parent Bias if they judged the boy to
resemble the adoptive father on at least ten of the twelve features (p=.02).
7
The version of the task on which the following results are based was administered to a total
of 31 adults (mean age=42 years, range=22-70).
8
The adoption task used with children was the same as that used with adults, except for
the fact that, to make the task easier, I used photographs of two different villages: one was
the village where the child in the story was born, and the other where he was raised.
Throughout the telling of the story and the questioning, I pointed to the relevant picture in
turn. In turn, children could answer by simply pointing to the picture associated with the trait
or parent of their choice, without having to remember any of the specific traits; this made the
task much less taxing on their memory and verbal skills than might at first appear. Before pro-
ceeding with the resemblance questions, children were asked a few comprehension questions
to make sure that they had understood the story. If they failed, the story was repeated a second
time; if they failed a second time, they were dropped from the study. As I piloted the task, it
became obvious that children younger than 6 or 7 would not be able to pass the compre-
444 RITA ASTUTI

hension questions. From the start, Vezo adults expressed the view that I would be wasting
my time with such children, since they are not yet `wise' (mbo tsy mahihitsy) and do not know
anything at all (tsy misy raha hainy). In fact, these younger children were simply too shy and
too intimidated by the testing procedure.Vezo children are hardly used to being asked genuine
questions, questions to which they are expected to have independent answers, and their response
to my questioning was total passivity. The fact that the sample only includes children of
both sexes aged 6 to 13 (n=40; mean age=9 years, range=6-13) raises an important issue,
namely that the Vezo sample is very different from that of Solomon, Johnson, Zaitchik, and
Carey's studies, which recruited North American children as young as 4. This difference is a
crucial factor in the comparison between Vezo and North American children's performance.
However, the older age of the children in my sample does not affect the comparison between
Vezo children and Vezo adults, which is what I am pursuing here. I shall therefore set this issue
to one side.
9
This is the interpretation typically found in the colonial and ethnographic literature on the
Vezo, as discussed in Astuti (1995a; 1995b).
10
All participants were set two tasks: the previously described task in which adoption
occurred across two sets of Vezo parents, and one of the tasks in which adoption occurs across
the Vezo-Karany or the Vezo-Masikoro divide (i.e. V-V and V-M or V-V and V-K). The two
combined studies were balanced across participants according to a Latin-Square design in order
to control not only for the potential confounding factors described already (note 5), but
also of whether the Vezo-Vezo task was administered before or after the Vezo-Karany or the
Vezo-Masikoro task and of whether the birth parent was Vezo or Karany (in the case of the
Vezo-Karany version), and of whether the birth parent was Vezo or Masikoro (in the case
of the Vezo-Masikoro version). Preliminary analyses conducted on these factors revealed no
significant effects on the results discussed below.
11
Thus, 80 per cent of children who judged that the boy will be Vezo or Masikoro like his
adoptive father also judged that the boy will resemble his birth father in at least some of the
traits that are typical of his group, with a mean attribution of 3.6 traits out of a total of 6. By
contrast, no adult participant ever judged that the adopted boy will resemble his birth father
in any of the typical traits.
12
One can speculate that the reason for such bizarre performance is that these children first
judged the boy to be Vezo or Masikoro like his birth father, and that, when asked to justify
their choice, they simply fell back on the stock explanation, heard many times before, that
people are Vezo or Masikoro because of what they do, the sort of explanation that is `in the
air'.
13
Children's performance on theVezo-Karany task suggests that they might have been guided
by a default assumption about social identity, that people are what they are because their birth
parents are that way. This finding gives partial support to a modified version of Hirschfeld's
(1996) claim that there are innately supported specific competencies that underlie how humans
reason about social kinds, what he terms the `human kind competence'. I intend to discuss
Hirschfeld's claim in greater detail in a forthcoming publication.
REFERENCES
Astuti, R. 1995a. People of the sea: identity and descent among the Vezo of Madagascar. Cambridge:
University Press.
------ 1995b. `The Vezo are not a kind of people': identity, difference and `ethnicity' among
a fishing people of western Madagascar. American Ethnologist 22, 464-82.
------ 1998.`It's a boy!'`It's a girl!' Reflections on sex and gender in Madagascar and beyond.
In Bodies and persons: comparative perspectives from Africa and Melanesia (eds) M. Lambek & A.
Strathern, 29-52. Cambridge: University Press.
------ 2000. Kindreds and descent groups: new perspectives from Madagascar. In Cultures of
relatedness: new approaches to the study of kinship (ed.) J. Carsten, 90-103. Cambridge:
University Press.
Bloch, M. 1998. How we think they think: anthropological approaches to cognition, memory and
literacy. Boulder, Colo.: Westview Press.
RITA ASTUTI 445

------, G.E.A. Solomon & S. Carey 2001.An understanding of what is passed on from parents
to children: a cross-cultural investigation. Journal of Cognition and Culture 1, 43-68.
Bouquet, M. 2000. Figures of relations: reconnecting kinship studies and museum collections.
In Cultures of relatedness: new approaches to the study of kinship (ed.) J. Carsten, 167-90.
Cambridge: University Press.
Boyer, P. 1994. The naturalness of religious ideas: a cognitive theory of religion. Berkeley: University
of California Press.
Carey, S. 1995. On the origin of causal understanding. In Causal cognition: a multidiscipli-
nary debate (eds) D. Sperber, D. Premack & A.J. Premack, 268-302. Oxford: Clarendon
Press.
Carsten, J. 1995.The substance of kinship and the heat of the hearth: feeding, personhood, and
relatedness among Malays in Pulau Langkawi. American Ethnologist 22, 223-41.
------ 1997. The heat of the hearth: the process of kinship in a Malay fishing community. Oxford:
Clarendon Press.
Csordas, T. 1994. Introduction: the body as representation and being-in-the-world. In Embodi-
ment and experience: the existential ground of culture and self (ed.) T. Csordas, 1-24. Cambridge:
University Press.
Firth, R. 1985. Degrees of intelligibility. In Reason and morality (ed.) J. Overing, 29-46. London:
Tavistock Publications.
Gell, A. 1998. Art and agency. Oxford: Clarendon Press.
Hirschfeld, L.A. 1996. Race in the making: cognition, culture and the child's construction of human
kinds. Cambridge, Mass: MIT Press.
Ingold, T. 1991. Becoming persons: consciousness and sociality in human evolution. Cultural
Dynamics 4, 355-78.
Lambek, M. 1998. Body and mind in mind, body and mind in body: some anthropo-
logical interventions in a long conversation. In Bodies and persons: comparative perspectives
from Africa and Melanesia (eds) M. Lambek & A. Strathern, 103-23. Cambridge: University
Press.
Latour, B. 1993. We have never been modern. London: Harvester Wheatsheaf.
Lutz, C.A. 1988. Unnatural emotions: everyday sentiments on a Micronesian atoll and their challenge
to Western theory. Chicago: University Press.
Marriott, M. 1976. Hindu transactions: diversity without dualism. In Transaction and meaning
(ed.) B. Kapferer, 109-42. Philadelphia: ISHI Publications.
Middleton, K. 2000. How Karembola men become mothers. In Cultures of relatedness: new
approaches to the study of kinship (ed.) J. Carsten, 104-27. Cambridge: University Press.
Ortner, S.B. 1996. Making gender. Boston: Beacon Press.
Scheper-Hughes, N. & M. Lock 1987. The mindful body: a prolegomenon to future work in
medical anthropology. Medical Anthropology Quarterly 1, 6-41.
Schneider, D.M. 1984. A critique of the study of kinship. Ann Arbor: University of Michigan
Press.
Solomon, G.E.A., S.C. Johnson, D. Zaitchik & S. Carey 1996. Like father, like son: young
children's understanding of how and why offspring resemble their parents. Child Development
67, 151-71.
Sperber, D. 1985. On anthropological knowledge. Cambridge: University Press.
------ 1997. Intuitive and reflective beliefs. Mind and Language 12, 67-83.
Springer, K. 1992. Children's awareness of the biological implications of kinship. Child Devel-
opment 63, 950-9.
------ & F.C. Keil 1989. On the development of biologically specific beliefs: the case of
inheritance. Child Development 60, 637-48.
------ & ------ 1991. Early differentiation of causal mechanisms appropriate to biological
and nonbiological kinds. Child Development 62, 767-81.
Strathern, M. 1988. The gender of the gift. Berkeley: University of California Press.
------ 1996. Cutting the network. Journal of the Royal Anthropological Institute (N.S.) 2, 517-
35.
Toren, C. 1990. Making sense of hierarchy: cognition as a social process in Fiji. (London School of
Economics Monographs on Social Anthropology 61.) London: Athlone Press.
------ 1999a. Compassion for one another: constituting kinship as intentionality in Fiji. Journal
of the Royal Anthropological Institute (N.S.) 5, 265-80.
446 RITA ASTUTI

------ 1999b. Mind, materiality and history: explorations in Fijian ethnography. London:
Routledge.
Yanagisako, S.J. & J.F. Collier 1987.Towards a unified analysis of gender and kinship. In Gender
and kinship (eds) J.F. Collier & S.J.Yanagisako, 14-50. New York: Routledge.
Sommes-nous tous dualistes par nature? Une démarche
cognitive portant sur le développement
Résumé
Cet article prend  parti la position anthropologique bien établie selon laquelle les peuples
non-occidentaux ne sont pas piégés par la pensée dualiste. Bien que les informants Vezo de
Madagascar tiennent des propos qui pourraient servir  soutenir une telle position, l'appli-
cation de procédés expérimentaux  leurs raisonnements par inférence révle qu'ils font une
différence systématique entre l'esprit et le corps, entre les phénomnes biologiques qui déter-
minent l'organisme et les phénomnes sociaux qui donnent forme aux personnes. Cela donne
 penser qu'il y a une différence notable entre les déclarations linguistiques explicites des
gens et leur savoir théorique implicite. De plus, les données sur le développement démon-
trent que de telles présuppositions théoriques sont essentielles pour la production et la trans-
mission du savoir culturel.Ainsi les enfantsVezo qui ne différencient pas encore le mécanisme
biologique de la naissance et le mécanisme social par lequel ils sont élevés sont encore inca-
pables de comprendre un aspect dominant de la culture Vezo,  savoir l'ensemble des idées,
reliées entre elles par des relations de cause  effet, selon lesquelles les adultes classifient le
monde social. Ces découvertes ont des implications théoriques et méthodologiques consid-
érables en ce qui concerne la constitution du savoir anthropologique.
Dept. of Social Anthropology, London School of Economics & Political Science, Houghton Street, London
WC2A 2AE. R.Astuti@lse.ac.uk
RITA ASTUTI 447