ASSESSING CULTURALANTHROPOLOGY

Robert BorofskyHAWAII PACIFIC UNIVERSITY

New York St. Louis Sm Francisco Aucklmd Bogota Caracas

Lisbon London Madrid Mexico City Milan Montreal New Delhi

Sanjuan Singapore Sydney Tokyo Toronto

11

Methods Belong to All of Us

Respect the nature of the empirical worldand organize a methodological stance to reflect that respect.

Herbert Blumer (1969:60)

The word "method" in anthropology, indeedin all of social science, has at least three mean-mgs.

First, method refers to epistemology, to sets ofassumptions about how we acquire knowl-edge. The phrase "scientific method," for ex-ample, encapsulates a set of such assumptions(as well as some rules of practice). The as-sumptions are: (1) there is a reality "out there"(or "in there" in the case of ideas and emo-tions); (2) it can be apprehended, more or less,by human beings through direct experience(or through some proxy for direct experience);(3) all natural phenomena can be explainedwithout recourse to mysterious forces beyondinvestigation; and (4) though the truth aboutphenomena is never known, we do better andbetter as old explanations are knocked downand are replaced by better ones. Competingepistemologies reject one or more of theseassumptions.

Second, method refers to strategic approachesto the accumulation of actual data. Experi-mentalism and naturalism are two strategic ap-proaches within the scientific method. Experi-mentalism involves the direct manipulation ofvariables under the most controlled conditionspossible. Naturalism involves observing phe-

nomena in their natural environment. Sea-going oceanographers, astronomers, wildlifebiologists, and anthropologists are all natural-ists. Participant observation is a strategic ap-proach to data gathering used by somenaturalists in the social sciences.

And third, method refers to techniques or setsof techniques for collecting and analyzingdata. The survey questionnaire method is a setof techniques (involving sampling, construc-tion of instruments, interviewing, and otherthings) for collecting data. Spot observation isanother set of techniques for gathering data(for example, about how much time peoplespend in various activities). Content analysis,componential analysis, multidimensional scal-ing, and ethnographic decision tree modelingare techniques for analyzing (extracting mean-ing from) data.

Anthropologists receive little formal trainingin any of these three kinds of research meth-ods. Many graduate programs in anthropologyoffer no required course in ethnographicmethods (Trotter 1988; Plattner 1989). I be-lieve that training in all three kinds of meth-ods-epistemology, strategy, and technique-should be a major part of both the undergrad-uate and graduate curriculum in anthropology.

How much training in methods is enough?I'll address this question at the end of the

chapter. First, I discuss method as epistemol-ogy and method as strategy.

One might draw a broad distinction betweentwo epistemologies in cultural anthropology-positivist and interpretivist. I will focus on theformer here and leave others (in this volume)to describe the latter for, in my view, thepositivist perspective has not always beenunderstood. Positivists and interpretivists maydisagree on matters of epistemology. Butwhen we talk about methods at the level ofstrategy and technique, methods belong to allof us.

At its birth, positivism was simple enough.Here is John Stuart Mill in 1866 explaining"Ie positivisme" to an English-speaking audi-ence.

Whoever regards all events as parts of a constantorder, each one being the invariable consequent ofsome antecedent condition, or combination ofconditions, accepts fully the Positive mode ofthought. . . . (15)

All theories in which the ultimate standard ofinstitutions and rules of adions [IS] the happiness ofmankind, and observation and experience the guides... are entitled to the name Positive. (69)

Positivism1 in the mid-nineteenth century wasscience applied to the study of humanity, anddone in the exuberant and optimistic spirit ofservice to human happiness.

The French social philosopher AugusteComte didn't just believe in science as aneffective means for seeking instrumentalknowledge, though. He envisioned a class ofphilosophers who, with support from thestate, would direct all education. They wouldadvise the government, which would becomposed of capitalists "whose dignity andauthority," explained Mill, "are to be in the

ratio of the degree of generality of theirconceptions and operations-bankers at thesurrunit, merchants next, then manufacturers,and agriculturalists at the bottom" (1866:122).Comte attracted admirers who wanted toimplement the master's plans. Mercifully, theyare gone, but the word "positivist" still carriesthe taint of Comte's ego.

The modern version of posItIvIsm wasdeveloped by members of the Vienna Circle, agroup of philosophers, mathematicians,physicists, and social scientists who met from1923 to 1936 in a series of seminars. Themembers of the circle were corrunitted toempiricism and against metaphysics. Hunchescould come from anywhere, but scientificknowledge, they said, is based only on experi-ence, and scientific explanation is based onlyon logical, mathematical principles. Hence thelabel "logical empiricists," by which they wereknown for a while.

For these positivists, logic was paramount.They understood that the assumptions ofscience are assertions. They also understoodthat the assumptions of science are not mereassumptions. The members of the ViennaCircle were corrunitted to a scientific, that islogical, mode of thinking and to the benefitsthat this would bring to humanity. This wasthe component of positivism that they sharedwith Comte. It's what motivates many socialscientists today, including me, to count our-selves as positivists.

It is clear, of course, just how dangerous thiscorrunitment can be. My ideas about what isgood may not be the same as yours. This iswhat makes the study of ethics a critical partof method as epistemology.

In the early decades of this century, manyluminaries of cultural anthropology supportedthe collection of both quantitative and qualita-

tive data. They collected texts and countedthings as necessary in their research.

For example, one of Tylor's enduring con-tributions to anthropology was his paper "Ona Method of Investigating the Development ofInstitutions" (1889). In that paper, Tylor de-scribed a numerical method for doing sys-tematic cross-cultural comparisons. One ofBoas's great contributions was a monograph inwhich he demonstrated the relationship of nu-trition to body size among immigrants, ren-dering useless much racist rhetoric of the dayagainst allowing Eastern Europeans into theUnited States. Kroeber's study of 300 years ofwomen's fashions is a landmark in the quanti-tative study of cultural trace materials(Richardson and Kroeber 1940).

When I was a graduate student at the Uni-versity of Illinois thirty years ago, the traditionof using both qualitative and quantitative datawas still strong. Oscar Lewis, for example, ad-vocated the use of survey methods to comple-ment ethnographic field studies.2 Joseph Casa-grande taught cross-cultural hypothesis testingusing the Human Relations Area Files. Both,of course, were ethnographers who had col-lected reams of qualitative data.

There was, to be sure, another tradition inanthropology, one opposed to the use ofquantitative data. The cause against quantifi-cation in anthropology was eloquently takenup by one of Boas's students, Paul Radin, in1933. In The Method and Theory of Ethnology,Radin praised his teacher for criticizing Ty-lor's and Frazer's evolutionary doctrines. Buthe accused Boas of being "naturwissenschaft-lich eingestellt" or science-minded (Radin1933:10), of believing that cultural facts couldbe transformed into physical ones andcounted. He railed at Boas for training a gen-eration of students like Kroeber, Sapir, Lowie,Wissler, and Mead who, Radin said, sought totag cultures and habitats, like animal and plantspecies, and compare them statistically (Radin1933:10).

Radin was right about that crowd, but hehad a different idea of what anthropology

should be about. An ethnographer had to livea long time with the people whom he or shestudied, and had to learn the native languagefluently. Above all, said Radin, the ethnog-rapher must provide readers with the originaltexts, the materials from which he or shemakes observations.

Those texts, and the exegesis of the ethnol-ogist, should show current cultural realities "asseen through the mirror of an actual man'sheart and brain and not through the artificialheart and brain of the marionettes with whichBoas and Sapir and Kroeber operate" (Radin1933:10). Radin didn't mince any words. Todemonstrate his method, Radin presented ameticulous analysis of a series of texts fromthree Winnebagos about the Peyote Cult. Inone of those texts, John Rave discussed hisconversion to the Peyote Cult. Radin drewon his knowledge oflocal history. He showedthat Rave's account reflected acceptance byRave of particular components of Christianfaith and particular components of earlierWinnebago beliefs. Rave failed to achieve avision during his Winnebago puberty rite, andwas denied membership in the MedicineDance. This, said Radin, accounted for Rave'sproselytizing zeal.

You can object to all this as just so muchpop-psych, but Radin was the consummatefieldworker: he knew the language, he spentyears working with the Winnebagos and withthese particular informants, and he had thetexts to back up everything he said. IfKroeberand Boas and Sapir didn't like his arguments,Radin said, "there still remains the documentfor them to interpret better and more pro-foundly" (1933:238).3

Radin would not have wanted to be re-membered as a scientist, but in my terms hewas-and a gifted one, at that. He was empiri-cal in the extreme; he built arguments onlyfrom observations; he was committed to theopen-endedness of truth. With all his texts onthe Peyote Cult, he had the data on which topostulate a theory of revitalization. That heleft this nomothetic exercise to others who

would come later diminishes not one whit thecontribution of Radin's effort.

Many anthropologists today who shareRadin's views about primary data also want tocontribute to the development of knowledgeabout regularities in human behavior. Theymay have some well-founded doubts aboutwhether all the phenomena in which they areinterested can be quantified accurately. But Ibelieve that what keeps many cultural anthro-pologists (I haven't counted them, but I'mguessing hundreds, not dozens) from conduct-ing quantitative research is that they are justunequipped to do it when they leave graduateschool.

METHOD AS STRATEGY:PARTICIPANT OBSERVATION

Anthropologists are divided on epistemolog-ical issues, but almost all of us use the strategicmethod of participant observation to collectour primary data. Even anthropologists whouse questionnaires in their research engage firstin participant observation in order to build thequestionnaire instrument.

Participant observation is what makes it pos-sible for interpretivists and positivists alike tocollect life-history documents, attend sacredfestivals, talk to people about sensitive topics,map the landholdings of informants, trek witha hunter to count the kill, and interviewwomen traders formally and informally abouthow they cover their losses in the daily mar-ket. Among other things, participant observa-tion helps us build rapport. Rapport is whatmakes it possible for us to observe and talk topeople and record information about theirlives.4

Anthropologists developed the method of par-ticipant observation, and we have seen withsatisfaction its acceptance by researchers inmany other disciplines. But we haven't fol-

lowed up. After decades of experience withthe method of participant observation, we arenot much further along than when we started.We haven't studied participant observationsystematically; we haven't improved it.

At least five things affect the kind and qual-ity of data we can collect as participant ob-servers: (1) personal characteristics, such as ageand gender; (2) language fluency; (3) objec-tivity; (4) informant accuracy; and (5) infor-mant representativeness.

Personal Characteristics. By the 1930s, MargaretMead had already made clear the importanceof gender as a variable in data collection.Gender has at least two consequences: it influ-ences how you perceive others; it limits youraccess to certain information.

In all cultures, you can't ask people certainquestions because you're a [woman][man].You can't go into certain areas and situationsbecause you're a [woman] [man]. You can'twatch this or report on that because you're a[womanHman]. Even the culture of anthro-pologists is affected: your credibility is dimin-ished or enhanced with your colleagues whenyou talk about a certain subject because you'rea [womanHman]. (See Golde 1970; White-head and Conaway 1986; Scheper-Hughes1983b; Altorki and EI-Solh 1988.)

When she worked at the Thule relocationcamp for Japanese-Americans during WorldWar II, Rosalie Wax did not join any of thewomen's groups or organizations. Lookingback after more than forty years, Wax con-eluded that this was just poor judgment.

I was a university student and a researcher. I wasnot yet ready to accept myself as a total person, andthis limited my perspective and my understanding.Those of us who instruct future field workers shouldencourage them to understand and value their fullrange of being, because only then can they copeintelligently with the range of experience they willencounter in thefield. (1986:148)

Besides gender, we have also learned thatbeing old lets you into certain things and shuts

you out of others. Being a parent helps youtalk to people about certain areas of life andget more information than if you were not aparent. Being wealthy lets you talk to certainpeople about certain subjects and makes othersavoid you. Gregarious anthropologists may beunable to talk to shy people.

Being divorced has its costs. NancieGonzalez found that being a divorced motherof two young sons in the Dominican Repub-lic was just too much. "Had I to do it again,"she says, "I would invent widowhood withappropriate rings and photographs" (1986:92).Even height may make a difference: AlanJacobs once told me he thought he did betterfieldwork with the Maasai because he's almost6 1/2 feet tall than he would have if he'dbeen, say, an average-sized 5 feet 10 inches.

Language Fluency. Thirty years ago, RaoulNaroll (1962:89-90) found suggestive statis-tical evidence that anthropologists who speakthe local language are more likely to reportwitchcraft than those who don't. His inter-pretation was that local language fluency im-proves your rapport, and this, in turn increasesthe probability that people will tell you aboutwitchcraft.

Does the credibility of our data depend oncontrol of the local language? In his 1933 dia-tribe against his science-minded peers, Radincomplained that Mead's work on Samoa wassuperficial because she wasn't fluent inSamoan (1933:179). Fifty years later, DerekFreeman (1983) raised questions aboutwhether Mead had been duped by her in-formants, perhaps because she didn't know thelocal language well enough.5

According to Brislin, Lonner, and Thorn-dike (1973:70) Samoa is one of those cultureswhere "it is considered acceptable to deceiveand to 'put on' outsiders. Interviewers arelikely to hear ridiculous answers, not given ina spirit of hostility but rather sport." Brislin etal. call this the "sucker bias," and warn field-workers to watch out for it. Presumably,knowing the local language fluently is one

way to become alert to and avoid this prob-lem.

If participant observers run the risk of beingsuckered, then research is called for to deter-mine (1) the conditions under which that'smost/least likely to happen and (2) how torecognize and avoid it. Research is called foron the kinds of data available to "indigenousanthropologists" that are not available to out-siders, and vice versa (see Messerschmidt 1981for papers on doing fieldwork in your ownculture). Research is called for on making par-ticipant observation the best strategic methodit can be.

Objectivity. No one I've ever known in thesocial sciences has seriously thought thathumans could become robotic, completelyobjective field researchers. But just becauseperfect objectivity is impossible, this does notlet us off the hook. The economist RobertSarlow is reported to have observed that,while a perfectly aseptic environment is im-possible, this doesn't mean we might as wellconduct surgery in a sewer (cited by Geertz1973:30). Objectivity clearly varies from per-son to person. Some people achieve more ofit, others less. More is better.

Objectivity means becoming aware of one'sbiases, and transcending them, not the lack ofany biases. Some people do better at tran-scending, some do worse. Striving for objec-tivity is important even if perfect objectivity isunobtainable.

Laurie Krieger, an American woman doingfieldwork in Cairo, studied physical punish-ment against women. She learned that wifebeatings were less violent than she hadimagined, and that the act still sickened her.Her reaction brought out a lot of informationfrom women who were recent recipients oftheir husbands' wrath. "I found out," she says,"that the biased outlook of an Americanwoman and a trained anthropologist was notalways disadvantageous, as long as I was awareof and able to control the expression of mybiases" (1986:120).

The need for objectivity is recognized byeven the most qualitative fieldworkers. ColinTurnbull says that the key to good fieldworkis "to know ourselves more deeply by con-scious subjectivity" because in this way "theultimate goal of objectivity is much morelikely to be reached and our understanding ofother cultures that much more profound"(1986:27).

Objectivity does not mean (and has nevermeant) value neutrality. No one asks CulturalSurvival, Inc., to be neutral in documentingthe violent obscenities against indigenouspeoples of the world. No one asks AmnestyInternational to be neutral in its effort todocument state-sanctioned torture. We recog-nize that the power of the documentation is inits objectivity, in its irrefutability, not in itsneutrality .

Informant Accuracy. In 1940, C. Wright Millswrote that "the central methodological prob-lem of the social sciences" is the fact that whatpeople say and what they do is so different(1940:329). He was right, and the problemremains central today.

Informants lie, of course. But lying appearsto be a small part of the problem. Informants,like all of us, make honest errors ofcommission (they say they did things that theydidn't do) and of omission (they neglect toreport things that they did do) in reportingtheir own behavior.

The problem was articulated clearly byRichard La Pierre in 1934. Accompanied by ayoung Chinese couple, he crisscrossed theUnited States by car. They stayed at and ate at66 hotels and 184 restaurants. Six months afterthe trip was over, La Pierre wrote a letter toall those hotels and restaurants, telling themanagers that he was planning a trip andwould they mind serving Chinese people?More than 90 percent said they would notserve Chinese people. Irwin Deutscher pickedup the theme in 1972, in his book What WeSay, What We Do. He made it clear that theproblem had not been solved, despite the fact

that scholars from many disciplines had readLa Pierre's study and had recognized itsimportance.

My colleagues and I conducted a series ofstudies during the 1970s to find out if peoplecould accurately report their social interac-tions. A lot of social science, after all, is basedon data collected by asking people "who didyou [talk to] [exchange memos with] [interactwith] [call on the telephone] in the last [24hours] [week] [month]?" In 1984, my col-leagues and I reviewed the literature on in-formant accuracy (see Bernard and Killworth1977, for example). A long list of studies, in-cluding our own, showed that a fourth to ahalf of what informants say about their behav-ior is inaccurate.

This finding shows up in studies of whatpeople say they eat, how often they claim tohave gone to the doctor in a given period oftime, and in how much money they say theyhave in their savings accounts. It shows up inthe most unlikely (we would have thought)places: in the 1961 census of Addis Abapa, 23percent of the women underreported thenumber of their children. Apparently, peopledon't "count" babies that die before reachingthe age of two (Pausewang 1973:65).

Progress is being made in explaining in-formant inaccuracy. Thirty years ago, Cancian(1963) showed how informant errors con-formed to expected patterns in prestige rank-ings in a Mexican village. D'Andrade showedin 1974 that there is a general pressure tothink in terms of "what goes with what" evenif this creates errors in reporting factual events.(Also see Shweder and D'Andrade 1980.)More recently, Freeman, Romney, and Free-man (1987) found that, for some behaviors atleast, informants report typical behavior toanthropologists rather than specific behaviorwhen asked to dredge up what they actuallydid and who they actually interacted with (seealso Freeman and Romney 1987; McNabb1990).

People round off, in other words, and reportbehavior according to rules of central ten-

dency. This may go for informants' reports oftheir own behavior as well as for their reportsof the behavior of others. (Cognitive psychol-ogists, of course, have done careful investiga-tions of how people store and retrieve infor-mation.)

Still, the question remains: When peoplereport their behavior to us, do we get inac-curate answers, or answers that are accur-ate proxies for some other question aboutnorms?

Informant Representativeness. Even if infor-mants tell us accurately what they know, thereremains the question of whether what theyknow is representative of the population weare studying.

Since the 1930s, cultural anthropology andsociology have gone their separate wayslargely over this issue. Sociologists havefocused on developing systematic instruments(like questionnaires) and applying those ques-tionnaires to representative samples of re-spondents in a population. This places empha-sis on reliability (whatever the instrumentmeasures today it will measure tomorrow,more or less) and on external validity (what-ever we find out from the sample can be ex-tended to the rest of the population, withinsome known limits of error).

Anthropologists have focused on internalvalidity. What good is it to know that data arereliable or generalizable if they are inaccurate?Every social scientist knows horror stories ofquestionnaires that force people to producenonsensical answers (because the questions areculturally inappropriate, for example).

On the other hand, why would we want toknow what a handful of informants think ordo, if the information cannot be generalizedbeyond those informants? This is one areawhere considerable progress has been made inanthropology. Jeffrey Johnson's recent book(1990) on selecting ethnographic informantssummarizes much useful information in thisarea. Romney, Weller, and Batchelder (1986)offer a technique called consensus modeling

for selecting informants who are experts inparticular cultural domains (like ethnozoology,local medical practice, or whatever). Romneyet al. demonstrate that, when assumptions oftheir model are met, just six informants maybe enough to achieve representative, valid in-formation about a cultural domain. It is anintriguing line of research.

METHODS TRAININGIN ANTHROPOLOGY

I believe that training in methods should be akey part of both the undergraduate and thegraduate curriculum in anthropology. Howmuch is enough? Well, given the three mean-ings of the word "method" (epistemology,strategy, and technique), it takes a lot to beenough.

All anthropologists need a thorough ground-ing in the various approaches to knowledgethat have characterized our discipline. Thismeans exposure of all students, whatever theirinitial predilections, to the philosophical foun-dations of structuralism, symbolism, interpre-tivism, hermeneutics, phenomenology, posi-tivism, and empiricism.

Training in ethics should be part of thetraining in epistemology. Plattner (1989:33)makes the case:

Training in the ethics of research is as important astraining in the techniques of research design, datacollection and data analysis. Research consists of aseries of decisions (where to study, what to study,who to interview, what questions to ask, when toleave the field), and every decision ... has anethical component.

Every anthropologist has stories to tell abouthow his or her ethics were tested by circum-stances in the field. We have accumulated alarge collection of these cases, in the Newsletter

of the American Anthropological Association, forexample, and in Cassell and Jacobs (1987). Weought systematically to use the wisdom theyprovide and not rely on the telling of storiesinformally to disseminate that wisdom. Thecase-study method is used in teaching law andbusiness administration. A case-study courseon ethics should be part of every graduate cur-riculum in anthropology.

All anthropologists need formal trammg inparticipant observation. This means familiaritywith the accumulating literature on the sub-ject. It means studying what we already knowabout how good and bad participant observa-tion is done-about culture shock, about gen-der roles in the field, about violence, disease,and other hazards of field research, aboutmaintaining objectivity without the pretenseof value neutrality.

This kind of training will give anthropolo-gists the skills they need to do better field-work. It will also give them the means tocontribute concretely to improving participantobservation.

This is where it gets tough. College graduatesin the social sciences these days simply mustknow how to use a computer-not just forword processing but for data entry, data man-agement, data retrieval, and data analysis. Stu-dents of anthropology who manage to escapegetting these skills before going to graduateschool will need compensatory educationbefore they can study formal data collectionand analysis.

Once the basics are out of the way, anthro-pologists need training in research design, indata collection (structured and unstructuredinterviewing, for example), and in dataanalysis. Students who hope to do any seriouswork with survey data, for example, need atleast two courses in applied statistics (through

multivariate analysis). Students who want towork with a large corpus of field notes, orwith life history material, need training incomputer-based text management.

All anthropologists need formal training inthe collection and analysis of both qualitative andquantitative data. Photography, videography,audio taping, stenography, and sketching aremethods for collecting qualitative data. Cod-ing, counting, and measuring produce quanti-tative data. People's responses to open-endedquestionnaire items are qualitative. Codingand attaching numbers to the responses createquantitative data. Observing people andrecording their behavior in words producequalitative data. Counting the behaviorscreates quantitativr data. Transcribed texts arequalitative data. Counting the number oftimes people use a particular word or theme ina text creates quantitative data.

Pawing through field notes is qualitativedata processing. Putting the notes on a com-puter and using a text management programto paw through them is still data processing.Doing a chi-square test with a hand-heldcalculator is quantitative data processing. So isputting thousands of numbers into a computerand having the machine calculate the chi-square value.

Much of what is called statistical analysis is,in my vocabulary, data processing, not dataanalysis. Thinking about the themes in a text,like field notes, is data analysis. Thinkingabout what a chi-square (or any statistical)value means is also data analysis. All analysis isultimately qualitative. But we can't even getto that stage unless we've collected solid(valid, credible) data.

Besides learning to use research methods,we should be improving them. Every field tripprovides information on how personal charac-teristics and language fluency affect data col-lection. Every field trip is an opportunity totest the relationship between what people sayand what they do. We need to monitor andpublish the results of all these naturally occur-ring experiments.6

In the past, cultural anthropologists were moreconcerned with description than with explan-ation and prediction. A good description is stillits own analysis;being right about what causesa problem is still the best contribution anyonecan make to solving it; and good writing-telling a good story-is still the best methodany anthropologist can acquire.

Many anthropologists today, however, areinterested in research questions that demandexplanation and prediction, questions like:Why are women in nearly all industrialsocieties, socialist and capitalist alike, paid lessthan men for the same work? Why is medicalcare so hard to get in some(societies that pro-duce plenty of it? Does cultural pluralism sup-port or undermine the stability of states (as inCanada, Belgium, India, Azerbaijan, Yugo-slavia, or Kenya, for example)?

Answering such questions demands greatersophistication in research methods than is nowcustomary in anthropology. It requires skills incomparative statistical research and in gather-ing data from published sources including datathat are only available on computer tapes.Anthropologists who are involved in multi-disciplinary team research on complex humanproblems-in agricultural development, fer-tility control, health care delivery, and educa-tion-need more methods training than hasbeen the norm to date.

Knowing about methods for collectingqualitative and quantitative data is more diffi-cult in some ways today than it was forty yearsago. There are simply more methods nowthan there were then? In some ways, though,learning research techniques is easier now.Microcomputers and efficient software maketext management, statistics, and cognitive do-main analysis less intimidating.8 No one cancontrol all research methods. But those whocontrol a certain fraction will be unintimi-dated about learning new methods that be-come available and that seem appropriate toparticular research projects.

As I said at the beginning of this essay,anthropologists may disagree, even vehe-mently, about method as epistemology. Butthey share a common interest in and concernfor method at the level of strategy and tech-nique. Controlling many different methods forcollecting and analyzing qualitative and quan-titative data liberates us to investigate what-ever theoretical questions attract us. And con-versely: limited training in methods limits ourability to think about and solve many impor-tant theoretical questions. Knowing more isbetter than knowing less.

1. Mill felt that the French word positillisme was notsuited to English, but decided in the end to use Comte'sword, properly anglicized to refer to Comte's concept.I've long wished Mill had decided otherwise.

2.. See Lewis's contribution to the 1953 volume Anthro-pology Today, edited by Alfred Kroeber. See also Lewis'sarticle on family studies in the 1950 volume of theAmerican Journal of Sociology. In those days, family studieswas still the province of sociologists. When Lewis wrotefor anthropologists, he stressed the need for surveyresearch as a complement to ethnography. When hewrote for sociologists, he stressed the need for ethno-graphic research as a complement to survey data.

3. I was inspired as a graduate by Radin's work on theproduction and presentation of native texts. Radinfollowed Boas's lead in native ethnography of indigenousNorth American peoples. Boas trained George Hunt, aKwakiutl, to write ethnographic descriptions ofKwakiutl culture. Hunt produced over 5,000 pages ofnotes that served as the basis for much of Boas's pub-lished work on Kwakiutl life. Radin trained CrashingThunder to write in Winnebago, and Crashing Thunderwrote his classicautobiography of the same name.

Native ethnography has long held great promise as ameans to develop an authentic emic data base aboutIndian and other indigenous cultures. I can only surmisethat the physical difficulties involved must have keptnative ethnography from becoming a major method incultural anthropology.

In recent years, I have found that microcomputersmake native ethnography much easier to do. I simplyprogram a word processor to handle whatever specialcharacters are needed in a particular language, teach

native speakers of previously nonwritten languages to usethe equipment, and hand the equipment over to them.

For many years, my partner in this effort hasbeen Jesus Salinas Pedraza, a Nahfiu (Otom!) Indianfrom the Mezquital Valley of Mexico. Salinashas writtenan extensive ethnography of the Nahfiu, which Itranslated and annotated (Bernard and Salinas 1989).

Since 1987, Salinas and Josefa Gonza lez Ventura, aNuu Saavi (Mixtec) Indian from Oaxaca, Mexico, haverun the Native Literacy Center at the Centro de Inves-tigaciones y Estudios Superiores en Antropologia Socialin Oaxaca City. They have now trained 75 other Indiansfrom Mexico, Peru, Bolivia, Argentina, Ecuador, andChile to use microcomputers to write books in theirrespective native languages.

There are many social and ethical issues raised by thiseffort. I deal with these, and the history of the project, inBernard 1992.

4. In this sense, participant observation, like everydaylife, is also an exercise in impression management (seeBerreman 1962:11).

5. Was Mead duped? Whether she was or not hasnothing to do with her overwhelming contributions toanthropology. All ethnographers run the same risk.

6. How much, for example, can we rely on informantsto tell us about their income? The size of their land-holdings? Where their grown children are? Data on thenature and causes of informant inaccuracy would helpilluminate the relation between cognition and behavior,between the internal and the external worlds of humanbeings around the world.

7. Methods do come and go, however. No one uses so-called "culture free" tests anymore to investigate cogni-

tion cross-culturally. Those tests were discredited andabandoned by most researchers thirty years ago (Brislin,Lonner, and Thorndike 1973:109).

8. For handling relatively small amounts of text material(up to say, a thousand pages of field notes, transcribedinterviews, etc.) GoFer is an excellent product. Write toMicrolytics, Inc., Two Tobey Village Office Park,Pittsford, NY 14534. For handling large amounts of textmaterial, ZylNDEX is highly recommended. Write toZyLAB, Information Dimensions, Inc., 100 LexingtonDrive, Buffalo Grove, IL 60089. For coding and analyz-ing text, consider the following: (1) The Text Handler,by Gery Ryan. This package works with WordPerfect.Write to Gery Ryan, Dept. of Anthropology, Universityof Florida, Gainesville, FL 32611. (2) TALLY 3.0, byJeffrey W. Bowyer. This program works with ASCIIftIes. Write to Wm. C Brown Publishers, 2460 KerperBoulevard, Dubuque, IA 52001. (3) DtSEARCH, byDavid Thede. DtSEARCH works with all of the topword processors, as well as with ASCII flies. Write toDT Software, Inc., 2101 Crystal Plaza Arcade, Suite231 Arlington, VA 22202. (4) THE ETHNOGRPH,by John Seidel. This program works with text frommajor word processors. Write to Qualis ResearchAssociates, P.O. Box 2240, Corvalis, OR 97339. (5)ANTHROPAC, by Stephen Borgatti. This programhelps researchers collect and analyze numerical data. It isparticularly useful for multidimensional scaling, hier-archical clustering, consensus modeling and othermethods that operate on similarity matrices. Write toAnalytic Technologies, 306 South Walker Street,Columbia, SC 29205.

All of the programs mentioned here have beenreviewed in Cultural Anthropology Methods from 1989 to1993.

H. Russell Bernard is Professor of Anthro-pology at the University of Florida. He hasdone field research in Greece, Mexico, andthe United States and on ships at sea. Bernardwas editor of Human Organization (1976-1981)and of the American Anthropologist (1981-1989). Recently Bernard has been workingwith Jesus Salinas and other Indian colleaguesin Oaxaca, Mexico, to establish a centerwhere native peoples can publish books intheir own languages. Bernard's best-knowncontributions are Research Methods in CulturalAnthropology (1988); Native Ethnography (with

Jesus Salinas Pedraza, 1989); Technology andSocial Change (edited with Pertti Pelto, 2ndedition, 1987); and a series of articles on socialnetwork analysis (with Peter Killworth andothers).

In the summer of 1959, as a junior at Queens Col-lege, I went to Mexico to study Spanish and cameback knowing that I wanted to be an anthropolo-gist. As an undergraduate, I studied with ErnestineFriedl, Hortense Powdermaker, and Mariam Slater.Late in my senior year, Powdermaker told meabout a new Ph.D. program just opening at the

University of Illinois. Perhaps I could get in there,she said.

Illinois in 1961 was an intense, intellectual envi-ronment. I studied with Kenneth Hale and DuaneMetzger for my M.A. in anthropological linguis-tics, and then with Edward Bruner, Oscar Lewis,Julian Steward, Dimitri Shimkin, and JosephCasagrande for the Ph.D.

Metzger was part of the (then) new ethnosciencecamp. The goal was to write the grammar of a cul-ture--to learn what a native speaker of a languageknows about, say, ordering a drink and to lay thatknowledge out clearly.

Making cultural grammars turned out to beharder than anyone imagined. Metzger offered ahands-on seminar. With a few other students, Ispent a semester working with one Japanese house-wife, learning and mapping the implicit rules sheused for deciding how to cut and arrange vegeta-bles on a plate.

It was an enormous effort just to keep track ofthe data. One of the other students got the com-puter to sort and print the whole corpus every timewe learned a new rule. The people over at thecomputer center thought this was pretty quaint,but this systematic approach to data gathering andthe idea of using computers to make light work ofcomplex data-management tasks have stayed withme ever smce.

Ken Hale was Carl Voegelin's student. Like Carl(and like ~oas and his early students before him),Ken worked closely and collaboratively withIndian colleagues. The model was to help Indiancolleagues produce their own texts, in their ownlanguages, and then to use the texts for linguisticanalysis and for cultural exegesis. Ken's example,and the tradition it represented, led to my lifelongcollaboration with Jesus Salinas, a Nahiiu Indianfrom the Mezquital Valley in Mexico.

Jesus was my informant in 1962 when I did theresearch for my M.A. thesis on the tone patterns ofOtom! (called Nahiiu in those days). In 1971, Ibecame Jesus's informant, teaching him to write inNahiiu and, in the 1980s, to use a Nahiiu wordprocessor. I'm still working with Jesus, who nowheads the Native Literacy Center in Oaxaca, whereIndians from around Latin America train in usingcomputers to write and print books in their ownlanguages.

Much of my career, then, was shaped by mywork at the M.A. level. I learned from that ex-

perience how important it is for students to be-come involved in research projects early and often.

During my Ph.D. studies, Julian Steward,Dimitri Shimkin, Joe Casagrande, and Kris Leh-man encouraged me to pursue my interests inquantitative data analysis. In Casagrande's seminaron cross-cultural research, I first learned to use theHuman Relations Area Files and to test hypothesesusing cultures as units of analysis.

I don't recall anyone labeling all this "positivism"in those days, or worrying about whether my in-terest in scientific, quantitative research was un-healthy. I read works by Tylor, Boas, Kroeber,Driver, Wissler, Murdock, and Roberts andnoticed that all of them did quantitative work andpublished reams of ethnographic work as well. Ifound this mix of qualitative and quantitativemethods to be very sensible.

My major doctoral professor was Ed Bruner. Edbecame identified with symbolic anthropology andI went in a different direction. But Ed taught meto write, and to understand that seeking knowledgewas only half the battle. You have to be able to tellothers what you have learned, to engage theirattention, and to keep them from closing the bookbefore you have finished your argument. This maybe one of the few things that positivists and inter-pretivists fully agree on; but for my money, it's themost important thing of all.

In 1972, I spent a year at the Scripps Institutionof Oceanography, where I met Peter Killworth, anocean physicist. We decided to study problemstogether that (1) neither of us could tackle alone,(2) both of use agreed were sheer fun, and (3) werenot in the mainstream of research in either of ourdisciplines. We also agreed that we would not letour joint projects get in the way of our separateresearch careers. (He is in ocean modeling.)

We did, in fact, have a great time doing a seriesof papers on informant accuracy, and we are hav-ing just as much fun now testing a network modelfor estimating the size of populations that you can-not count (like the number of rape victims in acity). Peter has taught me a lot about data analysis.

I have also benefited greatly from my associationwith Pertti Pelto. We began the National ScienceFoundation Summer Institute on Research Meth-ods in Cultural Anthropology in 1987. StephenBorgatti joined the teaching team of the summerinstitute in 1988, and I have learned a lot from himabout new analytic methods.

My intellectual biography is still being written. Ican look back and see the influences of my profes-sors clearly. But just as clearly, I see the influence

of contemporaries, of junior colleagues, and of stu-dents. This is what makes anthropology so excitingfor me. The learning never has to slow down.