Sorting is not categorization
A critique of the claim that Brazilians have
fuzzy racial categories
Francisco J. Gil-White
Solomon Asch Center
for the Study of Ethnopolitical Conflict
University of Pennsylvania
3815 Walnut Street
Philadelphia Pennsylvania, 19104-6196
(215) 573-2622
Abstract:
As a result of a spate of studies geared to investigating Brazilian racial
categories, it is now believed by many that Brazilians reason about race in a
manner quite different to that of Americans. This paper will argue that this
conclusion is premature, as the studies in question have not, in fact,
investigated Brazilian categories. What they have done is elicit sorting tasks
on the basis of appearances, but the cognitive models of respondents have not
been investigated in order to determine what are the boundaries of their
concepts. Sorting based on appearances is not sufficient to infer the
boundaries of concepts whenever appearance is not a defining criterion for the
concepts in question, as the case appears to be for racial and ethnic
categories. Following a critique of the methods used, I review a terminological
and theoretical confusion concerning the use of the terms ‘emic’ and ‘etic’ in
anthropology that appears directly responsible for the failure so far to choose
methods appropriate to parsing the conceptual domain of ‘race’ in Brazil.
I. Introduction
Do people use
the same kinds of cognitive resources when thinking about ‘race’ in different
parts of the world? The claim has been made that Brazilians have quite
different racial categories from those familiar to Americans, even though the
phenotypic contrasts in Brazil are very similar to those in the US. Many social
scientists now consider this to have been established, and the notion itself
approaches that of a cultural common-sense. This paper will examine the
evidence on which the claim rests, and will argue that it is not very good. This paper will also
make the Moreover,
it will make an argument that we perhaps we should not
expect Brazilian concepts of race to be very different from American ones oncewhen the evidence,
properly gathered, is finally in. A growing body of evidence suggests that humans are
essentialist when it comes to racial and ethnic categories, and that the bias
to essentialize such categories may be a universal aspect of human cognition.
Such evidence is being
gathered at the intersection of evolutionary psychology and the strands of
cognitive anthropology that intersect with it (e.g. Tooby &
Cosmides 1992; Cosmides & Tooby 1994; Atran 1998, 1990; Hirschfeld 1996;
Sperber 1996; Boyer 1994; Brown 1991). These intersections have focused on what is universal to humans, rather than
on the variation, and in so doing have brought at least the possibility of more
balance to the conceptualization of humans from the anthropological
perspective, which so far has been in the tight grip of the assumptions
inherited from the Boasian tradition: human universals are unlikely, and
the importance of cultural differences is great (cf. Brown 1991).
Significantly, the cognitive turn has made students of culture who are
informed by these theoretical turns of events more careful about apparent
differences and similarities that may exist between different cultures.
The main new bias, however, is certainly to ignore differences in order to
focus on what appears general. Perhaps predictably, this risks becoming a
new dogma that fails to do justice to the reality and importance of
cultural diversity (Gil-White 2001c; Richerson & Boyd 2000)-a grave danger
because what may sometimes appear as trifling differences can turn out to
have profound effects for the historical trajectories and system behavior
of different societies (e.g. Nisbett & Cohen 1996; Kelly 1985). But, that
danger aside, there is no question that the cognitive revolution has
provided a needed counterbalance, as well as better methods. Cognitive
anthropologists have moved beyond mere elicitations or demands that
informants introspect about their (often unconscious) mental processes. In
their stead has grown a concern for procedures that will make people use
their cognitive models, allowing us to infer their content and structure.
Evolutionary psychology and the strands of
cognitive anthropology that intersect with it have in the tight,
even suffocating, grip
of the assumptions inherited from the in
which the Boasian tradition: human universals are unlikely, and the importance of
cultural differences is great (cf. Brown 1991). S
has held much of cultural anthropology
(e.g. Tooby & Cosmides 1992; Cosmides & Tooby 1994; Atran 1998, 1990;
Hirschfeld 1996; Sperber 1996; Boyer 1994; Brown 1991). These
two disciplines (and their myriad intersections) have
focused on what is universal to humans, rather than on the variation, and in so
doing have brought at least the possibility of more balance to the
conceptualization of humans. Significantly,
the cognitive turn has made students of culture who are informed by these
theoretical turns of events more careful about apparent differences and
similarities that may exist between different cultures. The main new bias,
however, is certainly to ignore surface differences
in order to focus on what is appears general. Perhaps predictably, Tthis risks
becoming a new dogma that fails to do justice to the reality and importance of
cultural diversity (Gil-White 2001c; Richerson & Boyd 2000)—a grave danger
because ; what may sometimes appear as
trifling differences can turn out to have profound effects for the historical
trajectories and system behavior of different societies (e.g. Nisbett &
Cohen 1996; Kelly 1985). But, that danger aside, there is no question that the
cognitive revolution has provided a needed counterbalance, as well as better
methods. equipped us with a new way of formulating and
asking questions. Cognitive anthropologists have moved beyond mere
elicitations or demands that informants introspect about their (often
unconscious) mental processes. In their stead has grown a concern for
procedures that will make people use their cognitive models, allowing us
to infer their content and structure.
In a cumulative science, improved perspectives and
methods call for a reassessment of earlier conclusions. Given
these new perspectives and methods now at our disposal, it is imperative to
make reassessments of the conventional wisdoms that may have emerged from
studies based on prior methodologies. One such
area that may call for reexamination is the literature on the cognitive models
of race in Brazil (Harris et al. 1995, 1993; Byrne et al. 1995; Harris 1970;
Sanjek 1971; Harris & Kottack 1963). The consensus that has emerged from
this research is that racial categories in Brazil are fluid (interpersonally,
and perhaps also inter-temporally variable), fuzzy (without clear boundaries), and
ambiguous (admitting of different, imperfectly overlapping labels). To me, this
was surprising, because I found ethnic categories in Mongolia to be crisp
classical categories with necessary and sufficient conditions of membership,
and a very strong overlap across most individuals. Specifically, I found that
for the overwhelming majority of respondents, membership in an ethnie was a
matter of patrilineal biological descent from another member of the category.
It could be that Mongolia and Brazil, being different places in far flung
corners of the world, have developed very different cultural models for
thinking about race and ethnicity. But it could also be that, had I used
Harris’s methods in Mongolia, I would have reached his conclusions, and that,
if I were to use my own methods in Brazil, the conclusions would be very
similar to those I reached in Mongolia. This possibility raises the question of
whether the difference is an artifact of the methods chosen, and so we must
examine and compare in order to decide which set of methods is better at
addressing the central question: what is the native system of categories? That
will be the purpose of this paper.
Before I go further, I must clarify what may appear as a
confusion of domains. Above I have noticed
a disparity between the claims of Harris and colleagues, who were
investigating ‘race’, and my own
investigations, which have to do with ‘ethnicity’. In order for my findings to
be truly discrepant with those of Harris, ‘race’ and ‘ethnicity’ should be synonymous
in some sense, otherwise Harris’s and my own are just different findings
corresponding to different domains of cognition and experience. To clarify this
issue I offer the following distinctions.
‘Race’ in the technical sense. To the
biologist a race is a subspecies, identified by a morphological and
genetic discontinuity between two populations in the same species. The lay
categories of race that humans variously think they see in their own species do
not meet the criteria of a morphological or genetic discontinuity (see Boyd
& Silk 2000; Brown & Armelagos 20010).
There are no human races in the technical, biological sense.
The lay category of ‘race’. This is a category of people made on the
basis of phenotypic attributes. It is a category of people whose bodies ‘look’
a certain way. By using Since Harris
was using depictions of humans with different kinds of features
and asking his Brazilian respondents to label them, Harris was
attempting to he was investigate ing lay categories of race. But Aa lay category of race is always more than just
a category of people with certain physical traits. Membership in the category
is held to be explanatory or predictive of other things. This is why
‘Brunettes’ are not a ‘race’—there isn’t much in the way of content or
expectations,
—implicitly or explicitly, associated
with a member of the category. It is possible, moreover, that But a lay category of race is more may be even more than just a
phenotypic category with explanatory or predictive beliefs attached; the people
who use it may believe that they are looking at a biologically meaningful
grouping. That is, laypeople may believe that their lay categories
of race are in fact ‘natural’, even if the intuition does not carry all of the
sophistication of the technical understanding described above. Hirschfeld’s
(1996) investigations suggest this is so, and that the intuitions to process
our phenotypic categories as if they cut our the human species
at important natural joints are active at a very young age and may even be
innate. Whether or not in any particular culture some phenotypic categories are
essentialized as natural categories is an empirical
question—one that Harris did not pursue for Brazilian categories.
‘Ethnicity’ in the technical sense. It is not without controversy for me to advance a technical definition of ethnicity because the experts on this topic actually spend much time trying to undermine each other’s definitions. However, I have tried to show before (Gil-White 1999:813) that despite what appear to be lengthy, ink-spilling disagreements between totally opposite camps, all scholars of ethnicity actually implicitly agree with the following definition: “ethnic groups are agglomerations of people who, at a minimum, represent themselves as vertically reproducing historical units implying cultural ‘peoplehood’.”
The lay category of ‘ethnicity’. Since, as noted above, analysts recognize an ‘ethnic group’ when they think they have found a set of laypeople who organize themselves around an ‘ethnic identity’, there is not here, as in race, the possibility of a sharp distinction between the analytic and lay understandings of ethnicity. But at least one can say that the cultural distinctiveness, the purity of vertical reproduction, and the historical depth of the groups may be greatly exaggerated by native participants in an ethnic category system, and the analyst has no need to go along with such exaggerations and/or reifications. As before, a lay category of ethnicity may be, cognitively, more than just the claim of a cultural distinction and its mode of reproduction. Laypeople may essentialize their ethnic categories as corresponding to natural, biological cuts of humanity. This amounts to saying that laypeople may racialize their ethnic boundaries. Again, whether they do or not in any particular locality is an empirical question.
I have argued before that the extant empirical data favors
the interpretation that laypeople generally racialize their ethnic categories
(Gil-White 2001a,
1999). And sub-populations
characterized as representatives of a ‘race’ in a particular times and places so-called
races in particular times and places easily develop all the
trappings of ethnic groups (e.g. the white/black boundary in the US). This
convergence between the two domains is reflected in the past—and sometimes
current—practice whereby ethnic groups were referred to as ‘races’. And English
is far
from being not the only language to uncertainly
apply the same term to both domains (see below). All of this argues for the relevance
of cognitive investigations in the domain of ‘race’ to those which study
‘ethnicity’ and vice-versa. If it seems likely from some evidence that the lay
domains of race and ethnicity are really—from the point of view of the
cognitive processes that underlie them—one single domain, then we must pay
attention to findings that appear to open this to question. My data about the
cognition of ethnicity in Mongolia appear to contradict rather directly
Harris’s data about ‘race’ in Brazil, and therefore finding out whether this is
a genuine contradiction or a methodological obfuscation deserves the attention
I will give it below.
II. The cognition of ethnicity in Mongolia
A. Brief overview
I begin by reviewing briefly the methods I used in Mongolia, the results, and their interpretation. The presentation is quite abridged as my concern here is with broad methodological priorities, but the full details may be found in Gil-White (2001b).
My respondents are Torguud nomads in the district of Bulgan CumBulgan Sum,
XovdHovd
Province, Republic of Mongolia. Torguuds are a relatively small, Mongol ethnic
group, and they share the district of Bulgan CumBulgan Sum
with KazaxKazakhs,
who make up about 30% of the population in the district. Like Torguuds, some KazaxKazakhs
are nomadic herders and others are town-dwellers. Although there are a few KazaxKazakh
families in what is essentially Torguud territory, on the whole Torguuds and KazaxKazakhs
are territorially segregated—even Torguud and KazaxKazakh
town-dwellers live in separate parts of town. Relations between the groups are
outwardly friendly, if for the most part distant. Trade between them is very
limited, as most conduct a subsistence lifestyle and their economies are almost
identical, resulting in an absence of trade efficiencies. Intermarriage is
entirely absent (in a population of 10,000, there was only one reported case of
intermarriage and I was not able to corroborate it). There is very little
day-to-day interaction between Torguuds and KazaxKazakhs,
especially among herders, and they don’t know too much about each other’s
customs and norms.
The purpose of my investigation was to find out what Torguuds considered to be the rule or rules for being a member of an ethnic group, whether their own or another one. Much ethnographic data from around the world suggests that criteria of blood are paramount when making ethnic ascriptions. However, it has become something of a tradition in the social sciences to claim that ethnic membership is fluid, contextual, situational, and even rationally chosen (see Gil-White 1999 for a review). I decided not to pit criteria of blood against rational choices, but against enculturation. If enculturation from birth into a group—absent a tie of blood—will not confer membership, then rational choices are a moot point. This is therefore a very strong test of the ‘primordialist’ hypothesis, which maintains that ethnic actors will require biological descent from a category member in order to consider a given individual a member of the ethnic category.
The procedure consisted in asking a sample of Torguuds the following questions:
Question (1) If the father is KazaxKazakh
and the mother Mongol, what is the ethnicity of the child?
Question (2) The father is KazaxKazakh,
the mother Mongol, but
everybody around the family is Mongol and the child has
never even seen a KazaxKazakh,
outside of the father. The child will learn Mongol customs and language. What
is the ethnicity of this
child?;
Question (3) A KazaxKazakh
couple has a child that they don’t want. They give it in adoption to a Mongol
couple when the child is under a year old. Around the Mongol family there only
are Mongols and the child grows up never meeting or seeing a
single KazaxKazakh.
He is never told of the adoption and thinks that his biological father and
mother are the Mongol adopters. He grows up learning
Mongol customs and language. What is the ethnicity of this child?
Because in the first question my respondents unanimously
answered that the child would be KazaxKazakh,
revealing a patrilineal model for ethnic ascription, the second and third
questions were crafted so as to pit descent against enculturation, with
question 3 presenting the most extreme contrast.
The first time around a sample of 59 Torguuds were asked
these questions in a setup where the child is male, and the questions are
were
asked in sequence. In a second test, a new sample of 41 Torguuds was
asked the same questions but the order was randomized, the ethnicities were
switched (so that the father was now Mongol and enculturation KazaxKazakh),
and the child was made female. A majority in both tests believed that a child’s
ethnicity was that of his biological parents (principally his/her father’s), no
matter what the circumstances of enculturation, or what the child may have been
told. For question 3, this majority was a strong one in the second test (76%,
versus 59% in the first test), which randomized the order of the questions and
therefore did not create an expectation that the researcher, by making the
circumstances of enculturation progressively extreme, was fishing for an
enculturation-based answer.
Although this establishes that one’s ethnic ascription is
a matter of patrilineal descent, this could might be
nothing more than a way to flag one’s community of biological origin. To know
whether the ethnic categories are essentialized one needs to investigate if
respondents attach to ethnic membership the expectation of a behavioral
disposition. A subsequent test asked a sample of those who had answered that
descent always trumped enculturation whether the child in question 3 would
behave exactly like members of its community of adoption, or somewhat
different. A strong majority (74%) replied that the child in question 3 could
not or would not behave exactly like members of its adopting ethnie, but would
instead behave somewhat like the ethnie of its biological parents. This,
despite the child having absolutely no contact with this ethnie, and despite
being unaware of any such ancestry. Ethnographic data further suggest that even
those who gave explicit enculturation-based answers may have implicit intuitions
to match those of respondents who gave explicit descent-based/essentialist
models (Gil-White, 2001a).
B. Implications for research on ethnic and racial categories
The bulk of these findings suggests that ethnic categories
may be ‘naturalized’. Specifically, iIt may even be
that we treat them as ‘folk species’ categories akin to the categories bear
and mouse
(see Gil-White 2001a for a defense of this argument; for a
different interpretation of ethnic and racial essentialism see Hirschfeld 1996,
Gelman & Hirschfeld 1999). Here I restrict myself to commenting on what
this means for a reassessment of past methodology to understand ethnic
cognition.
Categorization processes in living kinds have two interesting properties where the importance of ‘appearance’ is concerned: (1) category membership is independent of appearances (Gil-White 2001a; Medin & Atran 1999; Atran 1998, 1990; Keil 1989; Gelman 1996, 1991; Medin & Ortony 1989); but (2) membership in a given category nevertheless carries an expectation of category-specific morphological uniqueness (Markman 1989, 1990).
The first point says that appearance is not the criterion
for categorization, although we may use appearances to guide our first
guess as to where a given object belongs. This is because category membership
is conceived as resulting from possession of the category ‘essence’ (believed
to reside ‘inside’), not a set of external features. It is true that the essence
is believed to typically cause typical surface
features, including appearance, and therefore that common category membership
will lead to the expectation of surface similarity between any two objects. But
since this expectation is not defining, a common appearance is neither
necessary nor sufficient for category inclusion. A dolphin, for example, looks
like a fish, but isn’t. Appearances are not a matter of fuzzy membership
either. A dolphin’s appearance does not make it partly a fish. It is not
a fish at all. And this is also not a question of modern versus old biology
(Atran , 1998). It is our readiness to accept
the statement or discovery by an expert source that dolphins are not, after
all, fish (and our immediate assumption that there must be something intrinsic
to dolphin ‘nature’ that makes this true) that sets living-kind categories
apart—not the actual theory we may have about the taxonomic position of
dolphins.
To see the distinction between sorting (which involves
identification procedures) and categorization (which involves reasoning about
definitional criteria), suppose I present you with a new stimulus and ask you
“what is this?” Assuming I give you nothing to go on beyond what is apparent on
the stimulus itself, you will give me
the best answer you can, with the available information. Using the Given
the appearance of the object as your guide, you will provisionally sort
it into the category that you think it most likely belongs in, and this will be
your answer. That however—and this is crucial—does not mean that the cues you
use to make this initial tentative sorting constitute the actual criteria
for category membership (i.e. the definition of an object inside the category X
is not necessarily equivalent to “possession of the cues a, b, c, ..., which are used to make the tentative
sorting”). If presence of these cues is not coextensive with the definition of
a member of category X, then I can give you additional information about that
object that may lead you to revise the slot that you had put it in at first (for
the difference between identification procedures and concept representation,
see Smith & Medin, 1981:40-43; Armstrong, Gleitman, & Gleitman, 1983)
My data suggests that if Dorj, a Torguud Mongol, were to
encounter a man who was dressed like a KazaxKazakh,
spoke KazaxKazakh,
and behaved in every obvious way like a KazaxKazakh,
Dorj would think to himself: “Lo, a KazaxKazakh.”
However, if told subsequently that this man’s biological parents were both
Mongol and that he had been adopted at an early age into the KazaxKazakh
community, Dorj would then think: “Aha, so this man is really a Mongol!” Dorj at first sorted the man into the KazaxKazakh
slot with the available information. He made the best guess he could make under
the circumstances. But finding that Dorj makes such guesses does not mean that
Dorj thinks “if it talks like a KazaxKazakh
and walks like a KazaxKazakh, then
it is a KazaxKazakh.”
Clearly, Dorj doesn't think this. A sorting guess made on the basis of
appearance, and categorization, are not the same thing.
It is precisely the fact that appearances are insufficient
for categorization, plus the expectation of behavioral regularities resulting merely from
the fact
of descent from Mongols alone (in
the absence of any contact with Mongols), which suggests that we are looking
here at an essentialized ‘natural kind’ category. The fact that the ‘essence’
is taken to be passed on reproductively suggests a species-like category, for i. In such categories, membership is unequivocally established believed to result from possession of the essence,
and is established from those facts as will determine if the the condition s for having the essence has been
ascertained, which condition amounts to the essence. In the case of
species categories, this is biological descent from another category member.
Notwithstanding the fact that appearances are insufficient
for categorization in ‘living kinds’, there is certainly a perceived causal
link between the essence and a given typical appearance; two ‘living kind’
categories will be expected to have
different appearances by virtue of their having two different essences, even if
appearance is not defining for any one individual. This may create a bias
bigger than the baseline category accentuation effect (Hogg, 1988:19) that
obtains generally and which makes us see members of different categories as
more contrasting than they actually are. In Bulgan CumBulgan Sum
I found that my respondents unanimously expected the biological child of
Mongols, if adopted into a KazaxKazakh
community, to look physically different from its adoptive community. However,
when I tested discrimination ability with a set of anonymous photographs, I
found that there was considerable phenotypic overlap between the groups
(Gil-White, 2001b). This suggests that information about category membership
may, in living kinds, greatly exaggerate perceptions of
dissimilarity.
III. Reassessing prior methodologies
The above findings strongly imply that when it comes to
ethnicity, we need to dig deeper than appearances and classifications of
appearances. Given the above-mentioned strong similarities and even common
semantic blurring and confusion between lay categories of ethnicity and race,
as well as experimental data supporting the idea that races are likewise
essentialized (Hirschfeld,
1996), the same admonition applies to racial categories.
There is now a corpus of work on the basis of which at least many some scholars believe that racial
categories in Brazil are quite different from their counterparts in the US
(Harris et al 1995, 1993; Byrne et al 1995; Harris 1970; Sanjek 1971; Harris
& Kottack 1963). The main contention is that racial categories in Brazil
are highly ambiguous with very fuzzy boundaries. In light of my findings, I
shall now argue that this conclusion is premature and, given the methodology
employed so far
by Harris and colleagues, not yet supported by any data. I begin with a
detailed discussion of Harris (1970) since that paper is emblematic of the
others.
Harris (1970) This study elicited
‘racial’ classifications from 100 Brazilian respondents by using
...a deck of 72 full face drawings constructed out of the combination of three skin tone, three hair form, two lip, two nose, and two sex types (...) Responses were initiated without using terms presumed to be part of the domain...In general it was found that a question involving a request for the 'qualidade' or 'tipo' or 'raça' of the person depicted in the drawing was adequate to prime the response process. The word 'cor' was used only as a last resort.
There is either Harris seems to have little concern for
priming the domain of race, or else Harris feels confident that how the question
is put will not really
affect the results much. But to assume that the words
‘qualidade’, ‘tipo’, ‘raça’, and ‘cor’—which translate into English roughly as
‘quality’, ‘kind, ‘race’, and ‘color’[1]—are
interchangeable as stimulants of responses in the same domain is strange. In Bulgan CumBulgan Sum,
for certain, I would not get answers in the same domain if I asked some people
“what is the ündecten (ethnie/race)
of this person? ” and others “what is the öngö
(color) of this person?” It appears that which question was used was ad hoc and depended on the idiosyncratic
comprehension problems of individual participants. Harris does not code his
data according to which responses were elicited by which kind of question, and
he reports no evidence of their interchangeability or a test of their effects
on responses. He obtained quite a wild variation in his responses—restricting
ourselves just to those terms used more than 100 times, these are still a grand
total of 12. However, this much
variation could be at least partly the product of asking four different
questions, not all of them priming the same domain, rather than the product of
a large universe of Brazilian ‘racial’ terms, as he claims.
Harris also finds that there is little intersubjective
agreement about the boundaries of the different categories. Perhaps. But by
showing people pictures and asking them to classify them he is more likely
testing identification procedures rather than cognitive models of category
membership. It is possible for the entire population to be in intersubjective
agreement about what makes a ‘branco’ a ‘branco’, but for different individuals
to have different thresholds—on the basis of appearance—for making a guess that
a person is branco or not. For example, suppose you and I agree that someone is
branco (‘white’) if and only if his/her parents are also branco; however, you
and I have different thresholds of increasing brownness at which we stop
guessing that someone is a branco. Should we be told the facts of descent, you
and I will agree on who is and isn’t branco, but when Harris shows us the
pictures, we will make different guesses. Because Harris’s study does not delve
deeper than the initial appearance-based
sorting guesses on the basis of appearance, the
underlying cognitive models are not really explored, and his data are therefore
silent on the intersubjective agreement that may exist for these. Aggravating
this problem is the fact that using black & white drawings rather than
photographs (or at least more realistic color drawings) is likely to increase
the ambiguity in people’s guesses because people are likely to have individually
varying biases to imaginatively complete the missing information.
Another possibility is that category X is understood by everybody to have the same conditions of membership, and that most of the respondents did not even disagree with each other during the identification process, but people name the categories differently. For example, you and I agree on who is a ‘preto’ (‘black’) but when Harris asks us, I say ‘preto’ and you say ‘moreno’, meaning the same category. The looseness of intersubjective variability in naming of the category is not the same as looseness in identification of the category or its membership requirements—especially if one usage is more polite than another and some respondents are trying to be polite. In Mexico, for example, people of African descent are usually referred to as ‘negros’ but, more politely, ‘morenos’ or ‘prietos’ (the two latter words have other usages and that is what makes them polite). Similarly, in the US, it would not be sensible for a researcher to assume intersubjective disagreement if some respondents identified a picture as ‘African American’, others as ‘black’, and yet others as ‘colored’—it is anything but far-fetched to suggest that many such respondents would agree if questioned further that they were talking about the same category. Harris interprets his data as pointing to differences in American and Brazilian racial categories, but a little introspection suggests that applying his methods in the US would result in qualitatively similar data.
Yet another possibility is that some of these words may function sometimes as racial category labels and other times as literal color descriptors. This seems plausible for the word ‘moreno’ ( ‘tanned’ or ‘brown’), and it may account for Harris’s (1963) claim that sometimes full siblings in Brazil receive different ‘racial’ labels. The terms almost certainly don’t function as ‘racial’ labels at all in these cases. It should be obvious that not all phenotypic distinctions are ‘racial’. When we speak of blondes and brunettes we are not, by that token, separating people into different ‘racial’ categories. Likewise, what Harris is referring to almost certainly involves people making non-racial phenotypic distinctions between siblings, and the terms used then function merely as labels to distinguish the surface appearance of siblings naturally considered to be of the same ‘race’.
This is certainly how color labels function within ethnic/racial categories in Mexico, where I grew up. In Mexico the upper-class is overwhelmingly ‘white’. Among whites, I am considered ‘moreno’ because I have dark eyes and hair, and white but usually olive skin due to easy tanning. One of my sisters had blond hair when she was young and still has blue eyes and very white skin. Among whites, she was called, as a girl, a ‘güera’, which is the term used by this group for any blonde. However, to an Indian or a mestizo we are both ‘güeros’ (‘whites’) and that is what Indians and mestizo’s who don’t know me and who want to attract my attention always call me ‘Hey, güerito!’—literally ‘whitey’ (although in Mexico no offense is taken, partly because of the double meaning). The term has this connotation especially for members of Mexico’s lower classes. A similar effect may take place in Harris’s data because his respondents are from different socio-economic strata and may thus, as in Mexico, have class-based usage variations for the same terms.
The residents of Bulgan CumBulgan Sum
are at the geographic center of Asia, and in their history they have migrated
far and wide. As a consequence they display astounding phenotypic diversity—at
least half have either blue or green eyes; there are blondes and brunettes;
some have epicantic folds over their eyes, others don’t; there are very pale
and very dark-skinned people, etc.; and also every conceivable combination of
traits (e.g. very dark-skinned people with green eyes and epicantic folds are
not uncommon). All this naturally leads people to employ, as in Brazil, a quite
rich vocabulary of phenotypic description. Individuals are said to be yellow,
brown, red, white, black… Many a person’s nickname is a diminutive form of
these color terms. But to describe a girl as ‘yellow’ (e.g. ‘Shar Oxio’ or
‘yellow daughter’) and her brother as ‘black’ (e.g. Xaraa or ‘blackie’) hardly
means they put them in different racial or ethnic categories. They are both
Mongol.
For another angle, consider that the residents of Bulgan CumBulgan Sum
are aware through schooling that Africans exist and some of them have even seen
one or two in the flesh because they came to this area as tourists. They refer
to them as ‘blacks’. The fact that some Mongols are also called ‘black’ does
not imply any ambiguity about their ethnoracial categories. Within Mongols, ‘black’ refers to a
person’s hue, not a person’s racial category. This is an elementary observation
in Mongolia, where the contrast between a Mongol referred to as ‘black’ and an
African is obvious. The contrast between Mongols and KazaxKazakhs,
on the other hand, is far from obvious, and corresponds to an objectively fuzzy
continuum comparable to the Brazilian contrast between, say, ‘pretos’ and the
intermediate category ‘morenos’. But since the conceptual distinction between
Mongols and KazaxKazakhs
is not fuzzy, we must consider it at least possible, on the strength of the
Mongolian evidence, that in Brazil objective phenotypic fuzziness in the racial domain
may likewise not imply a fuzzy conceptual boundary.
Finally, when Harris finds that a given drawing was
identified by most as ‘branco’ but by a substantial minority as ‘sarara’, does
this mean that Brazilian racial categories are ambiguous, or that some
respondents are giving the basic-level category and others a more specific
color description? To use a comparison example, if he showed them a picture of
a laptop computer and asked them ‘what is this?’, would finding that some
people answer ‘computer’ and others ‘laptop’ mean that they have different ways
of parsing the world? No. In this case it would simply mean that some people
chose a subordinate category and others a superordinate category, but that they
both agree that it is a computer and, more specifically, a laptop. In the same
way, perhaps, a person can be a ‘branco’ and, more specifically, a ‘sarara’ (in
other words, ‘sarara’ is the kind of ‘branco’ that she is). Which interpretation
is correct, however, cannot be determined unless one looks to see if any of the labels
given stand in inclusion relationships to each other. Harris did not
attempt this.
The above alternative explanations for Harris’s results are certainly not mutually exclusive. A second look at his data will illustrate this.
Figure 1. The numbers on the X-axis represent the descending
rank-order of the drawings in terms of number of ‘branco’ responses (Harris’s
numbers are different because he kept his original numbered identifying labels;
the order of the drawings is the same as Harris’s, however). For each drawing,
the number of ‘branco’ responses can be found on the Y-axis. The branco curve
is a close approximation of Harris’s smoothed line. For the other curves,
Harris provides the actual data points and they are reproduced here.
<<Insert Figure 1 about here>>
The figure above superimposes Harris’s ‘moreno’ responses on the ‘branco’-‘preto’ axis. I will refer to the drawings 1-15 as the ‘exclusive branco’ range because ‘branco’ is the most common response throughout that range. Similarly, I will call the drawings 32-36 as the ‘exclusive preto’ range. The ‘extended branco’ (where ‘branco’ is more common than ‘preto’ but not necessarily more common than ‘moreno’) is the range of drawings 1-21; the ‘extended preto’ is 21-36. The ‘moreno’ range is the set of drawings 15-32. Notice that although the extended branco and preto ranges meet almost at the midpoint, the exclusive preto range is very small because ‘preto’ responses dip very quickly. This is also because on average, there are more ‘moreno’ responses inside the extended preto range than in the extended branco range.
How can we explain the above distributions? One interpretation could go as follows. If the term ‘moreno’ sometimes refers to a racial category intermediate between ‘branco’ and ‘preto’ we should see a moreno range in the middle where ‘moreno’ responses swamp both ‘branco’ and ‘preto’. This is apparent above. If ‘moreno’, within a racial category, simply means ‘intermediate’ for that category, then we should see a ‘moreno’ spike in the middle of the branco range, and another one in the middle of the preto range. The first is apparent in the figure above (drawings 8-9). The second might be the spike at 31-32, except that it is not really occurring within the exclusive preto range whereas the first one occurred in the middle of the exclusive branco range. This might be because ‘moreno’ is sometimes used politely for people who are considered racially ‘preto’. So the spike and its base are both raised, the ‘preto’ curve lowered, and as a result the size of the exclusive preto range is cut shorter. This interpretation is bolstered by the fact that even with the most extreme drawing on the preto side (#36) there appears to be a reluctance to use the term ‘preto’ (32 instances against 50 of ‘branco’ for the comparison drawing on the branco side). Harris claimed that there was no relationship that he could find for the moreno responses on the ‘branco’-‘preto’ axis. But superimposing them as I do above (which Harris doesn’t do for his readers) suggests this interpretation as a possibility, though it is not one admitting of a statistical test. More data needed to be gathered about the respondents’ cognitions for such tests to be meaningful. There is simply too much missing information, and too much potential noise, in this data set.
The main reason for the noise is that Harris has asked his
respondents to sort a lot of perceptual stimuli. But sorting is not
categorization—that is, not necessarily. To find out if what emerges from a
sorting task of perceptual stimuli corresponds to the racial categories we need
to investigate how people think about the resulting ‘lumps’. If we don’t do
this all we have is data about categorization sorting guesses
that result from identification procedures—these are , not
necessarily the categories in people’s heads.
Harris wants us to accept the data as evidence of a racial categorization
system that, by comparison to, say, the American, is very wild. But this is
premature. We need to know what the racial categories are. These cannot be
inferred willy nilly from appearance categories—it is not written anywhere that
racial categories and first-guess piles due to sorting on the basis of
appearance should match exactly.
In a later study, Harris et al (1993), with a different methodology, again fail to rigorously rule out the confounding of
color descriptors and racial labels. They have two different but roughly
comparable samples in terms of phenotype. People in both samples are first
asked to self-identify freely (the authors do not report the question used).
Then one sample is asked to self-identify again in terms of just the four forced
choices of ‘branco’, ‘preto’, ‘amarelo’, and ‘pardo’ (the ‘pardo option’),
while the other sample does the same with the forced choices of ‘branco’,
‘preto’, ‘amarelo’, and ‘moreno’ (the ‘moreno option’). The question is: of
those who initially freely self-identify as ‘moreno’ how many will reidentify
as ‘pardo’ in the ‘pardo option’?
Harris et al find s that
only 37.4% of those who at first freely-identify as ‘moreno’ later reidentify
as ‘pardo’ in the ‘pardo option’. They His conclusion conclude is that
the racial categories are ambiguous, and he criticize s an
earlier academic effort (Silva, , 1988)
that led to the Brazilian census categories for having disregarded this by
discarding the ambiguous term ‘moreno’— and
adopting ‘pardo’ instead—as
the label for the intermediate racial category in the list of forced choices
that appear on census forms.
Harris’sThe conclusion
is rather startling. However, In the first
place, Silva (1988) argued that ‘moreno’ sometimes acts as
a racial label and sometimes as a literal color descriptor (this is a
recognition of ambiguity, but of a different kind), and he was trying to avoid
confounding the two. Because he thought that ‘pardo’ primed the racial domain
better by avoiding this confusion, the recommendation for the census became to
use ‘pardo’ instead of ‘moreno’ for the intermediate racial category. One may
argue that Silva did not rigorously establish that ‘pardo’ was the best choice,
but there is no question that he was concerned with distinguishing racial labels
from color descriptors. Harris again seems to have no concern for this, and it
leads him to his perplexing practice of treating any color terms applied to
people as ‘racial’ labels without any need for further evidence. He thus seems
unaware of
the fact that his own data in fact support
Silva’s contention that ‘moreno’ is sometimes a racial label and sometimes
a color descriptor (Telles, 1995 makes the same point), and that ‘pardo’
narrowly primes the racial domain. It is precisely under such
circumstances the latter circumstance that we should
expect a substantial proportion of freely self-identifying ‘morenos’ to later
reidentify as ‘branco’ or ‘preto’ in the pardo option, as Harris found.
And what about the moreno option? It is plausible that asking people to freely self-identify will cause many to answer in terms of color rather than race (especially if there is not much rigor in how the question is posed), but that being presented with Harris’s four forced choices (even when one of them is ‘moreno’) will better prime the basic-level domain of racial categories. This hypothesis would be supported if there was a noticeable decrease in the number of people who reidentify as ‘moreno’ even in the moreno option. However, Harris et al do not even report this crucial evidence.
Some suggestions for fixing the problems with the above methodologies are as follows. First, if the investigation is about race, one must be careful to elicit data in such a way that other domains are not primed. Sticking to the word ‘race’ is a good idea, but even better than that would be a pretest of how different words may lead to different kinds of elicitations. The use of color photographs, rather than black-and-white drawings, should be considered; but if drawings are used, the more information in them the better—realistic color is best. The drawings should be used merely as an instrument for respondents to express, if they can, a set of mutually exclusive categories that fully specify the domain. Proceeding this way is likely to establish the set of mutually exclusive categories at the ‘basic level’, for if respondents are asked to sort all photographs into categories they will almost certainly privilege the basic level (for discussions of basic-level categories see Lakoff, 1987; Smith & Medin, 1981; Rosch, 1978). This however, should not be taken for granted; one should proceed with further investigations. For example, if more than one clump received the same or similar names, or if a set of photographs is called by one name by some respondents, and a different name by others, respondents should be challenged to explain. Following that, one can ask if it is possible to name this or that category with other labels in some contexts, and what would these be. Careful attention should be paid to possible inclusion relationships to see which of the resulting categories parse the domain at the same taxonomic rank-level. Then one can ask about membership conditions to these categories—a very important question that helps distinguish appearance-based from ‘essentialized’ categories, and determines whether or not the categories obtained are mutually exclusive. Finally, one can compare the results of many individuals to see how much intersubjective variability there is in basic-level categories found and membership conditions given. Whether or not the categories show intersubjective agreement in the phenomenological boundaries is much less interesting than whether the cognitive models of respondents agree, and whether they would resolve borderline cases in the same way should they be given the same background information about the individuals that constitute the stimuli.
It is true that one does not find in the US as varied and prolific a vocabulary for describing people’s phenotypes as one does in Brazil. It is also true that the US system features hypodescent, which prevents the emergence of intermediate racial categories. These are quite hoary observations of the differences between the US and Brazil. However, it is possible that this is the extent of the difference. Harris simply assumed that all of the terms he got referred to ‘racial’ categories and did not bother to examine people’s cognitive models, limiting his study to an elicitation of labels. (This posture explains Harris’s remarkable claim that in Brazil full siblings may be classed in different ‘racial’ categories, an interpretation that stretches thin not only the data he has, but also the concept of ‘race’ as usually understood.) It seems quite possible that the differences in terms of cognitive models of race between Brazil and the US may be smaller than Harris thinks.
IV. Emic and etic: The theoretical pitfalls of having bad terms of art
The problems alluded to here appear to result at least in part from Harris’s reinterpretation of Kenneth Pike’s original distinction between emic and etic. Harris keeps the terms but his definitions in fact are quite orthogonal to Pike’s approach, have a different theoretical purpose, and importantly influence the methodological road taken. Revisiting these concepts is helpful both in terms of understanding Harris’s approach and its pitfalls, and to clarifying the useful work that cognitive anthropology can do to understand alien category systems. The words emic and etic, unfortunately, have come to mean almost anything to almost anyone (a sad commentary on the triumph of rhetoric over science in much anthropology). Headland (1990:20-22) reviews an astounding plethora of uses with almost nothing in common. In this discussion, however, I will focus on the original meaning, and on Harris’s early departure from it.
Before I do, an important caveat. In anthropology it is customary to spill much ink trying to wrest control of the meaning of terms of art from other anthropologists, as if defining a word were equivalent to providing an explanation or making an intellectual contribution. Many proceed as if the meaning of a word—itself apparently of profound importance—was something to be discovered through careful application of quasi-philosophical reflection, and the anthropologist was the professional whose task was to discover it (Headland et al, 1990 is an egregious example of this sort of thing). But this exercise does not advance science and it is not my purpose here. I am not advocating that the terms ‘emic’ and ‘etic’ be used in any particular way, and I don’t think the words themselves have any mysterious power. I merely want to show (1) that Pike and Harris are talking about completely different things despite the fact that they use the same terms, and (2) that Harris, though he defines the terms differently, nevertheless expects ‘emic’ and ‘etic’ to do the same work for him that they do for Pike. This causes serious confusion about what needs to be done in order to explicate an alien category system, and is directly responsible for the methodological problems outlined above. In a nutshell, I will argue that Harris uses emic/etic to distinguish between a ‘folk’ and an ‘analytic’ category , whereas Pike uses emic/etic exclusively to analyze the boundaries of folk categories.
Kenneth Pike’s original innovation was the result of applying his understanding of one system of categories, phonemes, to other systems of categories to see if this could generate any insights. A phoneme is not a sound but a collection of sounds. The ‘p’ in ‘pit’ and the ‘p’ in ‘spit’ are not the same sound, and they are produced by a different combination of vocal articulations. However, to an English speaker they both count as the same phoneme /p/. We therefore have that there is a collection of different sounds (phonetics) that a speaker will parse into a smaller number of functional/grammatical units (phonemics). The suffixes were used by Pike as technical terms to distinguish the variation in any domain—etic phenomena—that are parsed by people in a given culture into the same categorical slot—or emic unit. Thus, Pike was extending an insight of linguistics—the explication of the grammatical properties of the smallest units of speech, which resulted from understanding how a collection of different sounds constituted the same category of grammar—to other cultural ‘grammars’.
In all cultural domains, there is a lumping of objects,
performances, or events —which
are phenomenally and measurably different —into
the same category. A useful illustration offered by Pike (1990:28-29) is a
‘strike’ in baseball, as there are many different kinds of performances that
all count as ‘a strike’.
Individual emic units may have variants without losing their
status as the 'same' emic unit for a particular person or subculture.—Pike
(1990:28-29)
This insight may not appear as a very deep one until one
realizes that there may be nothing obvious at all about the actual parsing
processes and the categories themselves. For example, to many most native English
speakers, the contention that the ‘p’ in ‘pit’ and the ‘p’ in ‘spit’ are two
different sounds will come as a surprise. Thus, for Pike, the
‘insider/outsider’ distinction is directly tied to the fact that category
systems may be opaque to those who use them daily.
Insiders act but mostly don't understand what they know. I use and categorize sounds spoken by English speakers correctly into the phonemes of English. This is because I am a competent English speaker, an insider of the English language. But I don't have any insight into this tacit knowledge until I analyze my behavior in the manner of an outsider, and treat what I implicitly know as something requiring explanation.—Pike (1990:33-34)
These observations were important to anthropologists because the ethnographer, typically coming from a different culture than the one she studies, is an outsider, and therefore in a good position to find alien category systems as strange and demanding explanation, and also to see at work the categorization processes that go unnoticed by natives. However, there is a danger that the anthropologist will miss native categories if she superimposes her own. Thus, a determination of the native parsing of the world is indispensable to a proper understanding of an alien culture.
Again, think of phonemes. For English speakers, the sounds
which are orthographically glossed as ‘r’, and those which are glossed as ‘l’,
belong to two different phonemes. However, to a native Japanese speaker these sets are both
included in the same phoneme (it is
exactly like the ‘p’ in ‘pit’ and the ‘p’ in ‘spit’ to for a native English
speaker). This explains why Japanese learners of English so often get their
‘r’s and ‘l’s confused, and why this is so saliently strange and amusing to
native English speakers, but so hard to notice for the
Japanese. Similarly, one should expect that people from different cultures will
often parse the world of objects, events, and performances in different ways
that will surprise and confound an ethnographer. It is also possible that some
native categories that should surprise
the ethnographer never will because they go unnoticed—unless the ethnographer
has a method for discovering how natives parse the world. Therefore, it is
important, in the effort to understand an alien culture, to establish what
people will call ‘the same’. But it is equally important, because much of this
may not be conscious, not to trust the natives’ own reports of what is the same because, as insiders, natives may not
always be able to explain their own categories. Cognitive tasks that make
people reveal the underlying and perhaps cryptic categories by using them
are a much better methodological route than asking informants to introspect on
the meanings of things.
To sum up, Pike points out the importance of finding out what objects, performances, or events are lumped as ‘the same’ into one category. To speak of this problem he refers to the measurable variation as etic phenomena, and to the categories in which this variation is lumped by natives as emic units.
I now turn to Harris’s use of ‘emic’ and ‘etic’, which clearly departs from Pike. Below is Harris’s view, consistent over the years (Harris, 1990:53-54; my emphasis):
Grammatical rules...have the same epistemological status as phonemes, since the test of their validity, no matter how abstract the formulation, is whether they generate utterances that native speakers recognize as meaningful and appropriate. Such tests, however, are completely irrelevant to etic analyses, which stand or fall on their contribution to predictive or retrodictive nomothetic theories about the evolution of sociocultural differences and similarities.
Do not be fooled by the reference to phonemes—we are no longer talking about the same distinction that Pike has drawn. Harris distinguishes between what makes sense to the native, and what makes sense to the scientist; a distinction between two different languages, rather than two aspects of the same language (which is Pike’s distinction). ‘Emic’ and ‘etic’, for Harris, stand for two entirely separate ways of looking at the world. Emic is the native’s parsing of the world, and its measurement; etic is the scientist’s parsing of the world (dictated by her hypothesis), and its measurement. This is evident in the fact that Harris considers emic analyses “completely irrelevant” to etic analyses.
Could Pike ever say that emic and etic phenomena are irrelevant to each other? No. In Pike’s definitions ‘emic’ refers to a native category the boundaries of which are determined by looking at the ‘etic’ variation which it encloses. Remember, he starts with an analysis of phonemes: the phonetic variation is the collection of different variants (different along some domain that we can measure) that native categorizers lump together in the same concept (a given phoneme). Thus, ‘etic’, for Pike, does not refer to an alternative categorization with entirely different purposes that a scientist may have, but rather to the ‘stuff’ that the native emic unit is made of. Pike’s project is to find out what natives will consider ‘the same’. The measurement of the (Pikean) ‘etic’ variation is therefore indispensable—rather than completely irrelevant—to finding out which and how many (Pikean) ‘emic’ units natives will parse this variation into.
It is apparent that what Harris has really seized on is
Pike’s reference to an outsider/insider distinction. But notice the difference.
By the outsider/insider distinction Pike merely means to observe that the
outsider (the scientist) can measure the etic variation, but the native cannot
perform this analysis because she is trapped in the perception of the world
that her categories impose. Harris, on the other hand, has taken the words
‘etic’ and ‘emic’ to stand for the perspectives or categorical worldviews
of the outsider and the insider, respectively. For Harris , the
scientist is the
(outsider), and she has has
a hypothesis that dictates one choice of categories, and a system
of measurement that determines one set of claims; , while the
native is the (insiderr), with a separate has
her own—perhaps quite orthogonal—system of categories, and an—also
perhaps quite orthogonal—value system and set of priorities that may lead to
very different claims. This interpretation becomes even more evident as Harris
(1990:53-54; my emphases and brackets) continues:
This does not mean that etic analysis necessarily results in descriptions that are at variance with participants' sense of appropriateness and historical truth. In many domains, but especially technological processes, emic versions of cultural practices and behavior stream events [the native’s claims] correspond closely to etic versions of the same [the scientist’s measurements]. As shown by Allen Johnson in a study of Brazilian peasant farmers (1974), elicited rules for planting crops on particular kinds of lands and elicited descriptions of past planting activity sometimes corresponds closely with etically observed behavior. But as Johnson emphasized, the facts of correspondence or noncorrespondence raise equally important questions: Why are some rules followed while others are broken? Why do some individuals follow rules while others break them? Why are some rules and concepts widely shared while others vary from individual to individual? (Johnson, 1974:100).
Again, notice the contrast. ‘Emic’, in the above quote,
refers to how the native sees the world, and the language she employs. ‘Etic’
refers to how the scientist sees the world, and the language she
employs. For Harris ‘emic’ = what
people say they believe and do, and ‘etic’ = what people actually do as measured by an outside observer.
So, for Harris, if a farmer says he believes rule X about planting crops, that
is ‘emic’; the farmer’s actual behavior, when, for example, he is observed by
a scientist chooses (for example) to
violate that rule, is ‘etic’. But for the question of what rule X itself is
(and this is Pike’s question, and the raison d’être for his emic/etic
distinction), it matters not one whit whether people in fact go about breaking
rule X or not
because, for. Pike, ‘etic’ refers not to what people do but to how
people count what they do. An etic instance of rule X is anything that counts
as satisfying it according to the natives. A native who steals
may obviously count what she did as ‘stealing’ and therefore as breaking the
rule “thou shalt not steal.”Whether or
not I, in point of fact, steal property has no bearing on what kinds of
behaviors I consider to count as ‘stealing
property’. Thus, if Pike was
interested in the native emic unit ‘crop planting rule X’, in what sense would
a violation be ‘etic’? For Pike, an etic instance of rule X is anything that
counts as satisfying it (and violations, by definition, do not satisfy a rule). Her behavior is
irrelevant to Pike’s analysis; what matters is how she counts it.
The point is not that recording what people do, or whether they follow their own expressed rules, is unimportant or uninteresting. The point is simply that this is not what Pike’s emic/etic distinction is about. We may be interested in the degree to which people follow their rules, or to which they behave consistent with their salient concepts (Harris’s question), but this is not the same thing as having established what they understand their rules to be, or what the boundaries of their concepts are (Pike’s question).[2] I hope to have shown that we are faced here with two very different research problems, both of which quite unfortunately use the same two terms of art.
Now, if one wishes to conduct a study of how native
Brazilians parse the domain of ‘race’, one could understandably begin by
thinking that the Pikean emic/etic distinction will be useful for framing the
work ahead. The project will be to look at the Pikean etic variation, and find
out what tokens within that variation natives will call ‘the same’, in order to
determine what the Pikean emic units are. However, if one frames the problem
merely by using the words ‘emic’ and ‘etic’, while foisting a
different distinction on them, the problem will not be framed in the same way
at all. If the distinction used is Harris’s, one will say that etic is
the scientist’s perspectivesystem of categories,
having no necessary relation to the native’s system of categories. ,
while emic is the worldview of
the natives. This distinction may be handy for some purposes, but
it does not usefully frame the methodological task ahead for someone who is
about to engage in the decoding the boundaries of
a native system of categories (which was, in fact, Harris’s project). On
the contrary, if ‘emic’ is to ‘etic’ as the native’s claims are to the worldview
is to the scientist’s measurements, one may conclude that, since
we are after the emic (to the exclusion of the etic!), the proper
methodological route is to privilege the first answer the native gives
without challenge or further investigation. Who is the scientist, after all, to
impose his own ‘etic’ understanding on the native, if the native does not share
it? This stance seems to be directly responsible for Harris’s methodological
pitfalls.
Here are Harris et al (1993:453) explaining the application of his emic/etic distinction to the study of race:
In conducting a racial census, researchers can follow an etic or emic strategy. The most rigorous etic approach would employ biologically valid categories based on clusters of alleles as revealed through genetic testing (e.g. blood types, mitochondrial DNA). (...) In a more practical if less rigorous etic approach, racial categories, can be defined by clusters of phenotypical traits (e.g. skin color, hair form, nose width) by trained observers.
In an emic approach, on the other hand, racial identity
is established by eliciting respondents' categorizations of themselves or of
others using terms that respondents regard as appropriate, whether or not
observers find them to be ambiguous or contradictory.—Harris
For Harris et al an ‘etic’ approach to ‘race’ entails a form of analysis where the natives are not consulted, whereas ‘emic’ is a different form of analysis that requires the researcher to ask the natives about ‘race’, and privilege their answers.[3] The difference between Pike and Harris becomes starkly clear in this passage. Harris calls the studies I have criticized above ‘emic’ because he is recording peoples’ answers, rather than taking some other kind of measurement. However, Pike would say that what Harris obtained is a sample of ‘etic’ variation and, since Harris has not tried to establish where the conceptual cleavages and boundaries are in that variation, he has not yet found a single ‘emic’ unit. That is, in Pike’s terms, Harris has elicited many (etic) behavioral responses, but has not found—yet—any (emic) categories. To see this, it suffices to consider merely that Harris has not sought to establish whether any of the myriad terms his respondents gave are synonyms, or whether any exist in taxonomic inclusion relationships to each other, only one rank-level of which people consider to be ‘race’. Pike’s project, remember, is to find out what people will call ‘the same’, and thus leaving the possibility of synonyms and inclusion relationships unexplored reveals that Harris’s methods do not even attempt Pike’s project, even though, ostensibly, Harris claims to be after the native system of categories.
The above interpretation is bolstered by the fact
that Now, it is interesting that on at
least two recent occasions (Harris et al , 1993;
Byrne & Harris et al,
1995) Harris defines emic/etic for his readers with the historical link to phonetics. He does not cite
Pike despite the fact that it is Pike who made the derivation from phonetics,
and despite having had a public debate with him about the use of the terms
(Headland et al, 1990). Moreover, Harris uses the
definition he has used over the years rather than keep it in line with the
linguistic analogy. If Harris is talking about something else, why keep the
same terms? And why keep an alleged link to phonetics? Probably because the
historical link to phonetics creates the impression that the emic/etic
distinction is useful for doing work that explicates alien category systems (since
that was Pike’s explicit program), and Harris is ostensibly trying to
understand one such system.
Perhaps this is why we fight over words in anthropology:
we hope that words alone will do theory for us. But they can’t and don’t. One
cannot base a study of an alien categorical system on Pike’s terms (much less
only one of them). One has to base it on his linked definitions.
The lexemes ‘etic’ and ‘emic’ are themselves immaterial and have no special
power. It is what they stand for, in Pike’s exposition, that is useful.
Harris’s methodology elicited labels on the basis of appearance, but this is
not enough for establishing the nature or the boundaries of the categories,
nor is it sufficient to determine whether the responses really are all in the
same domain, or at the same taxonomic level. The reason is that it fails to
investigate whether and unless it seeks to establish whether and which which
of those elicited terms count as ‘the same’ in people’s minds.
Anthropologists should be wary of applying their own (Pikean)
emic system—that is, their own culturally-based native system of
categories—to an alien culture, but they should be equally wary of making the
facile assumption that the native system must be different than from their own.
Harris may have been too ready to believe that Brazilian racial categories are
radically different from American ones, and his studies y make s little
effort to prove him wrong.
Cognitive anthropologists need methods that go
beyond mere elicitation. Especially in domains such as natural categories where
appearances underlie first-guess identification rather than conceptual parsing,
elicitations of labels based on appearances (no matter how rigorous and careful
the preparation of elicitation materials) will fall short in our attempts to
map native categorical systems.
V. Conclusion
I found in Bulgan CumBulgan Sum
that people believe ethnic membership to be, first, strongly a question of
biological origin; second, predictive of behavior despite the fact that a
member may have no exposure to his ethnic group and may not even know himself
to be a member. This suggests that ethnic categories may be essentialized in
the manner of species categories and, like them, not defined by appearances.
These latter serve merely to assist first-guess identification, and therefore ethnic
categories they cannot be studied with methods
that rely solely
or even primarily on identification procedures without delving deeper
into the cognitive models of respondents. Due to the close empirical and theoretical overlap
between racial and ethnic categories, and to the plausibility that from the
cognitive point of view they are indistinguishable, the same admonition applies
to racial categories.
I
have critiqued here what I see as problems with previous research that
attempted to investigate the cognitive
models of race
in Brazil ethnicity by eliciting behavioral
responses to ‘appearance’ stimuli, and have suggested alternative methodology
that is more likely to succeed in the effort to map the conceptual boundaries
in naturalized or essentialized domains. I have also sought to trace the
historical source of the methodological road taken by Harris and colleagues,
which consists in its essence springs from of a theoretical misconception
resulting from using the terms emic/etic with different meanings than those
originally proposed by Kenneth Pike, while at the same time expecting them to
do the same work that, with the original meanings, they did for Pike in terms
of addressing the work that needs to be done to understand an alien category
system.
The most important lesson here is that the cognitive anthropologist needs methods that go beyond mere elicitation. Especially in domains such as natural categories where appearances underlie first-guess identification rather than conceptual parsing, elicitations of labels based on appearances (no matter how rigorous and careful the preparation of elicitation materials) will fall short in our attempts to map native categorical systems.
References Cited
Armstrong, S., Gleitman, L., & Gleitman, H. (1983). What some concepts might not be. Cognition, 13, 263-308.
Atran, S. (1990). Cognitive foundations of natural history. New York: Cambridge University Press.
Atran, S. (1998). Folk-biology and the anthropology of science: Cognitive universals and cultural particulars. Behavioral and brain sciences, 21, 547-609.
Boyer, P. (1994). The naturalness of religious ideas. Berkeley: University of California Press.
Brown, D. E. (1991). Human
Universals.: McGraw Hill.
Brown, R. A., and G. J. Armelagos. (2001). Apportionment of racial diversity: A review. Evolutionary anthropology 10, 34–40.
Byrne, B., Harris, M., Consorte, J. G., & Lang, J. (1995). What's in a name? The consequences of violating Brazilian emic color-race categories in estimates of social well-being. Journal of anthropological research, 51, 389-397.
Cosmides, L., & Tooby, J. (1994). Origins of domain-specificity: The evolution of functional organization. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain-specificity in cognition and culture (pp. 85-116). New York: Cambridge University Press.
Gelman, S. A. (1988). The development of induction within natural kin and artifact categories. Cognitive psychology,, 20, 65-95.
Gelman, S. A. (1996). Concepts and theories. In R. Gelman & T. Kit-Fong (Eds.), Perceptual and cognitive development (2 ed., pp. 117-150). San Diego: Academic Press, Inc.
Gelman, S. A., & Hirschfeld, L. A. (1999). How biological is essentialism? In D. Medin & S. Atran (Eds.), Folkbiology.: MIT Press.
Gelman, S. A., & Wellman, H. (1991). Insides and essences: Early understandings of the non-obvious. Cognition(38), 213-244.
Gil-White, F. J. (1999). How thick is blood? The plot thickens...: If ethnic actors are primordialists, what remains of the circumstantialist/primordialist controversy? Ethnic and Racial Studies, 22(5), 789-820.
Gil-White, F. J. (2001a). Are ethnic groups 'species' to the human brain?: Essentialism in our cognition of some social categories. In press, Current Anthropology.
Gil-White, F. J. (2001b). The cognition of ethnicity: Native category systems under the field-experimental microscope. In press, Field Methods.
Gil-White, F. J. (2001c). L'evolution culturelle a-t-elle des règles? La Rechérche, Hors-Série No. 5(Avril).
Harris, M. (1970). Referential ambiguity in the calculus of Brazilian racial terms. In N. E. Whitten (Ed.), Afro-American anthropology (pp. 75-85). New York: Free Press.
Harris, M. (1990). Emics and etics revisited. In T. N. Headland & T. N. Pike & M. Harris (Eds.), Emics and etics: The insider/outsider debate (Vol. 7, pp. 48-61). Newbury Park and London: Sage.
Harris, M., Gomes Consorte, J., Lang, J., & Byrne, B. (1993). Who are the whites? Imposed census categories and the racial demography of Brazil. Social forces, 72(2), 451-462.
Harris, M., Gomes Consorte, J., Lang, J., & Byrne, B. (1995). A reply to Telles. Social Forces, 73(4), 1609-1611.
Harris, M., & Kottack, C. (1963). The structural significance of Brazilian racial categories. Sociologia(25).
Headland, T. N. (1990). A dialogue between Kenneth Pike and Marvin Harris on emics and etics. In T. N. Headland & K. L. Pike & M. Harris (Eds.), Emics and etics: The insider/outsider debate (Vol. 7).
Headland, T. N., Pike, K. L., & Harris, M. (Eds.). (1990). Emics and etics: The insider/outsider debate ( Vol. 7).
Hirschfeld, L. (1996). Race in the making: Cognition, culture, and the child's construction of human kinds. Cambridge MA: The MIT Press.
Hogg, M. A., & Abrams, D. (1988). Social motivation, self-esteem and social identity. In D. Abrams & M. A. Hogg (Eds.), Social identity theory: Constructive and critical advances (pp. 28-47). New York: Harvester Wheatsheaf.
Johnson, A. (1974). Ethnoecology and planting practices in a swidden agricultural system. American ethnologist, 1, 87-101.
Kelly, R. (1985). The Nuer conquest: Structure and development of an expansionist system. Ann Arbor: University of Michigan Press.
Lakoff, G. (1987). Women, fire, and dangerous things : what categories reveal about the mind. Chicago: University of Chicago Press.
Markman, E. M. (1989). Categorization and naming in children: Problems of induction. Cambridge, MA: MIT Press, Bradford Books.
Markman, E. M. (1990). Constraints children place on word meanings. Cognitive science, 14(1), 57-77.
Medin, D., & Atran, S. (1999). Folkbiology.: MIT Press.
Medin, D. L., & Ortony, A. (1989). Psychological essentialism. In S. Vosniadou & A. Ortony (Eds.), Similarity and analogical reasoning. Cambridge: Cambridge University Press.
Nisbett, R. E., & Cohen, D. (1996). Culture of honor: The psychology of violence in the South. Boulder, CO: Westview Press.
Pike, T. N. (1990). On the emics and etics of Pike and Harris. In T. N. Headland & T. N. Pike & M. Harris (Eds.), Emics and etics: The insider/outsider debate (Vol. 7, pp. 28-47). Newbury Park and London: Sage.
Richerson, P. J., & Boyd, R. (2000). Climate, culture, and the evolution of cognition. In C. Heyes & L. Huber (Eds.), The evolution of cognition. Cambridge, MA: MIT Press.
Rosch, E. (1978). Principles of categorization. In E. Rosch & B. B. Lloyd (Eds.), Cognition and categorization. Hillsdale, NJ: Lawrence Erlbaum Associates.
Sanjek, R. (1971). Brazilian racial terms: Some aspects of meaning and learning. American anthropologist, 73, 1126-1143.
Silva, N. d. V. (1988). Cor e processo de relizacão socio-economica. In C. Hassenbalg & N. d. V. Silva (Eds.), Estrutura social, mobilidade, e raça.: Editora Revista Dos Tribunais.
Smith, E. E., & Medin, D. L. (1981). Categories and concepts. Cambridge, MA
London, England: Harvard University Press.
Sperber, D. (1996). Explaining culture: A naturalistic approach. Oxford: Blackwell.
Telles, E. E. (1995). Who are the morenas? Social Forces, 73(4), 1609-1611.
Tooby, J., & Cosmides, L. (1992). The psychological foundations of culture. In J. H. Barkow & L. Cosmides & J. Tooby (Eds.), The adapted mind: Evolutionary psychology and the generation of culture. (pp. 627-637). New York and Oxford: Oxford University Press.
Figure 1. The
numbers on the X-axis represent the descen-ding rank-order of the drawings in
terms of number of ‘branco’ responses (Harris’s numbers are different because
he kept his original numbered identifying labels; the order of the drawings is
the same as Harris’s, however). For each drawing, the number of ‘branco’
responses can be found on the Y-axis. The branco curve is a close approximation
of Harris’s smoothed line. For the other curves, Harris provides the actual
data points and they are reproduced here.