Handwerker 2002 AA - The Construct Validity of Cultures

W. PENN HANDWERKER

The Construct Validity of Cultures: CulturalDiversity, Culture Theory, and a Method forEthnography

ABSTRACT If we distinguish culture (embodied in individuals) from cultures (embodied in the superorganic properties of groups), we

make it possible to accommodate the observation that individuals vary, make choices, and exert control over their lives with the obser-

vation that those same individuals find themselves constrained by recurrent patterns with properties of things. This shift in perspective

makes the central problem for ethnography the identification and description of the evolving configurations of cognition, emotion, and

behavior at the intersection of evolving, individually unique cultural sets. Principal components analysis of similarities among inform-

ants identifies and describes this intersection and its important variations precisely. Ordinary least squares and logistic multiple regres-

sions test for plausible antecedents of intracultural and intercultural variation, respectively. I illustrate with data on cultural diversity in

the relative importance of components that make up a partnership between parents and teachers in the United States and in how to

organize these components effectively. I demonstrate the existence of a single model of the importance of these components, with

intracultural variation. I demonstrate the existence of two cultural models of how to organize these components into effective par-

ent-teacher partnerships (Separate but Equal and Mutual Decision Makers). Discordance between these cultures and the social identi-

ties of the cultural participants validates Keesing's claim that culture is not bounded in ways many people have long assumed. The shift

in perspective that reconciles Sapir and Kroeber points past culture theory to a theory of culture. As we work out the details, it will also

help us make ethnographic sense out of the contemporary world and give us a better grasp of the past one. [Key words: cultural diver-

sity, culture theory, ethnographic methods, ethnology]

THIS ARTICLE SUGGESTS HOW WE might accom-modate the observation that individuals vary, make

choices, and exert control over their lives with the obser-vation that those same individuals find themselves con-strained by recurrent patterns with the properties of su-perorganic wholes. My proposal draws attention to twolargely overlooked and undervalued implications of Ty-lor's (1871) definition of "cultured (1) the culture thatspecific people use to live their lives constitutes an evolv-ing configuration of cognition, emotion, and behaviorunique to themselves; and (2) cultures consist of evolvingconfigurations of cognition, emotion, and behavior at theintersection of individually unique cultural sets. The cen-tral problem for ethnography thus consists of identifyingand describing that intersection and its important vari-ations.

1 propose a method to identify and describe the Inter-section and Its important variations precisely, which takesas Its point of departure Roinney's cultural consensus the-

ory. Consensus theory contains unrealistic assumptions,however. 1 argue that we approach the identification anddescription of cultures—what people share—as a simpleproblem in construct validation.

1 illustrate the method with data on cultural diversityin the relative importance of components that make up aworking relationship between parents and teachers in theUnited States and in how to organize these componentseffectively. 1 demonstrate

the existence of a single cultural model of the impor-tance of these components, with a specific form of in-tracultural variation;

• the existence of two cultural models (Separate butEqual and Mutual Decision Makers) of how to organ-ize these components into effective parent-teacher re-lationshlps, as well as the existence of a subpopulattonthat works with no shared understanding; and

AMERICAN ANTHROPOLOGIST 104(1) 106-122. COPYWGHT O 2002, AMCHKAN ANTHROPOLOGICAL ASSOCIATION

Handwerker • Construct Validity of Cultures 107

that neither intracultural nor intercultural variationcorresponds with what commonly pass for "culturalgroups."

I conclude by setting this solution and these findings inhistorical context and by advocating a specific directionfor the construction of a theory of culture.

CULTURE AND CONTEMPORARY TRIBES: CLASS.

GENDER. AND ETHNICITY

It has become commonplace to hear people ascribe differ-ences in what people think and do (cultural differences) tolabels for class, gender, and ethnicity, in much the sameway that we once ascribed cultural differences to tribalidentities. I have a very personal sense of comfort in andunderstanding of a West Indian culture that emergedamong people of northwest European and West Africanorigin brought together in a plantation economy. So Ihave no doubt that there exist distinctive cultures legiti-mately called African American, Puerto Rican, or broadlyAmerican. I feel equally sure that there exists a women'sculture and a men's culture. But I feel like the boy wholooked for the emperor's clothes when I look for evidencethat validates their existence.

Historically, we used assumptions, not evidence, toequate cultures with social identities such as Nuer, Navajo,and African American; we also dismissed, overlooked, ordownplayed both cultural variation among people whouse the same social identity and cultural equivalenceamong people who use different social identities. Our his-torical data come to us in labeled bits we usually call socie-ties or cultural groups. But, as Wolf (1982) reminds us, theconstituency of such groups, their labels, and even theirexistence change by means of the interconnected histori-cal processes through which people seek access to re-sources. On the Pepper Coast of West Africa, men who dif-ferentiated themselves as father and son becameundifferentiated members of a regional dako that com-peted with its neighbors for control over land, rights overwomen and children, and trade; men of competing dakobecame undifferentiated "Kru" when they sought work onships plying the commercial shipping lanes off the WestAfrican coast (McEvoy 1977). A few "Kru,1 along with"Yoruba" and despite marked differences in language, cus-toms, and physical attributes, became undifferentiated"slaves" who worked plantations in the West Indies andthe southern United States. Those who found a way toown their own plantations and slaves came to be called"Gens du Couleurs" in Haiti and equivalent names else-where, although their descendants as well as those of theirslaves who now live in the United States became undiffer-entiated "African Americans.''

1 start from Wolf's central claim that "the world of hu-mankind constitutes a manifold, a totality of Intercon-nected processes, and inquiries that disassemble this total-ity Into bits and then fall to reassemble it falsify reality'*(1982:3) and his argument that "by turning names Into

things we create false models of reality. By endowing na-tions, societies, or cultures with the qualities of internallyhomogeneous and externally distinctive and bounded ob-jects, we create a model of the world as a global pool hallin which the entities spin off each other like so many hardand round billiard balls'" (1982:6). But I urge a step further.Cease perpetuating by assumption rather than evidencewhat constitutes the "bits" in which this totality of inter-connected processes can best be understood. We cannotknow precisely what cognitive, emotional, and behavioralconfigurations constitute cultures until we make the real-ity of social groups, the existence of cultures, and the loca-tion of cultural boundaries empirical issues that requireexplicit tests for construct validity.

Indeed, until we identify cultures by reference to spe-cific configurations of cognition, emotion, and behavior,we shall forever find ethnographically incomprehensible aworld in which a successful rap group comes out of Japan;one can watch the movie Out of Africa on a VCR run by akerosene generator in an African bush village a three-hourwalk off the road; and Russians play jazz, compose countrymusic, and turn capitalist—or the ongoing dynamics of so-cial interaction and cultural change in any contemporarycommunity. Indeed, as 1 write this 1 listen to an NPR re-port about typecasting and stereotypes in the entertain-ment industry. A Japanese American explains how hewalked out of a casting session for a Japanese mobster afterhaving been told repeatedly that his accent was wrong forthe part, until he selected one that represented a stereo-typical Cantonese Chinese. A Latina scriptwriter explainsher successful writing career by reference to her non-His-panic professional name. An African American actor anddirector explains how bias enters into the choices made bymost of the people who control the industry because theyhave little or no historical experience with real peoplewho have brown skin tones and minority ethnic labels.The construct of "'cultural group" fits these observationsno better than equivalent observations made centuriesearlier on the culturally diverse populations that came tobe merged into British, German, Italian, French, or Rus-sian identities (e.g., Wolf 1982:379). And to the extentthat we perpetuate by assumption the construct of "cul-tural group" bequeathed to us by structural-functionalistancestors, we beg what may constitute the most importantquestions ethnographers can address in the early 21 st cen-tury: "What groups, by what criteria, where are theirboundaries, and how can we know any of this?"

CULTURE. CULTURES. AND LIFE EXPERIENCES

Culture consists, most simply, of the knowledge peopleuse to live their lives and the way in which they do so.This definition conies to us from Edward B. Tylor, who, inPrimitive Society (1871), first identified culture as that com-plex whole that we acquire by virtue of living our liveswith other people. Ever since, we have wrestled with issuesimplicit in that definition. In what sense is culture a

108 American Anthropologist • Vol. 104, No. 1 • March 2002

"whole," and for whom—individuals, groups, or both? Ifgroups, how can we know them, and what are theirboundaries? How, exactly, do we "acquire" culture by vir-tue of living our lives with other people? Usually, we havetranslated acquire as "learned,1 which raises the questionof what, exactly, this means. Why do we learn the thingswe learn but not learn the things we could have but didnot? Does culture come to us from other people, or do wecreate culture in response to what we experience by virtueof living with others? Because learning "just hap-pens"—our consciousness of having learned comes afterthe fact—how and in what ways (if at all) does learningtake place volitionally? How much of what we learn andwhat we do not is hardwired into us?

After more than 100 years, increased understanding ofhow the mind works provides at least tentative answers tosome of these questions. These suggest that we reframequestions about culture and change how we go aboutstudying it. Ethnography, as 1 use the word, consists of theprocesses and products of research that documents whatpeople know, feel, and do in a way that situates those phe-nomena at specific times in the history of individual lives,including pertinent global events and processes. Histori-cally, we distinguished ethnography from ethnology. Theformer referred to descriptive accounts of culture, and thelatter, to comparative, explanatory analyses of culture.Rather than look at cultural variability between reified andessentialized social groups that exist only by assumption,we need to pay attention to cultural variability among andbetween individuals (see Barnett 1953; Boster 1987; Kees-ing 1994; Pelto and Pelto 1975; Wallace 1961). Explainingthe intracultural and intercultural variation we find meansextending ethnographic practice to correspond with a lineof ethnological practice (Boas 1894; Driver 1970; Jorgen-sen 1980). If we pay attention to cultural variabilityamong individuals, we make the reality of social groups,the existence of cultures, and the location of culturalboundaries empirical issues that require explicit tests forconstruct validity. We shall also see phenomena we didnot previously recognize.

Cultures as Super-organic Environments

Two sets of observations warrant this shift. First, reifica-tion and its corollaries—treating culture as if it were athing that could be owned and could act—pose significanttheoretical problems that too easily lead to real worlddamage. For example, in the recent Exxon Valdez oil spillcase in Alaska, the assumption that the world consists of"a sum of self-contained societies and cultures" ratherthan "people of diverse origins and social makeup whotake part In the construction of a common world" (Wolf1982:385) warranted relflcatlons on the part of both de-fendants and plaintiffs. Defendants alleged that nativeculture had been "smashed" centuries More the oil spill.Plalntlfls alleged that native culture was "damaged" by thespill. As Joseph Jorflenseii (1995) points out, both consti-

tute irresponsible and indefensible claims that obfuscatethe issues and lead to court decisions grounded on fantasy.The oil spill did not "damage" native culture—it could notbecause cultures are not things—but it significantly dam-aged natives, who experienced real losses of wild re-sources, real damage to the areas in which they gainedtheir livelihoods, real alterations to the manner in whichthey made a living, and real threats to the future genera-tions of animals on which they based their subsistence.Similarly, native culture had not been "smashed" centu-ries earlier—it could not because cultures are notthings—and native and nonnative inhabitants of the oilspill region responded to the spill in culturally charac-teristic ways that reflected the processes of social interac-tion that maintained cultural differences (boundaries)among people in the spill area (Jorgensen 1995:89).

The existence of patterned cultural differences likethese points to a sense in which we might accurately char-acterize cultures as superorganic wholes nonetheless. Cul-tures still do not constitute things. But our minds turnthem into phenomena that exhibit some of the propertiesof things. Recurrent behavioral patterns constitute a majorpart of the environment within which people live theirlives and, thus, constitute a major source of the importantsensory input we process. Random variation in behaviormay remain outside our consciousness and, even if not,usually may be profitably ignored. Recurrent behavioralpatterns, by contrast, exert effects because they establishparameters that cannot be ignored profitably, to whichone must respond. Nonrandom changes in recurrent pat-terns also call for responses. In doing so, they form ourminds—or, at least, the materials our minds use to re-spond to the world of experience.

As D'Andrade (1999) points out, demonstration that aculture exerts effects warrants reification and essenualiza-tion. One must explicitly identify and measure the culturebefore one can demonstrate its effects, however. Themethod 1 propose explicitly identifies and measures cul-tures and thus makes it possible to demonstrate their ef-fects. Some effects may originate most immediately in thepatterned responses of those around us to what we say anddo and in how we make sense of and respond to thosewords and acts in either patterned or individually uniqueways. Dressier and Bindon (2000) and 1 (Handwerker1999b) demonstrate effects that originate most immedi-ately in deprivation relative to specific cultural modelsand the behavioral patterns they rationalize. Some studies(e.g., Handwerker 1993, 1999a) suggest that specific recur-rent behavioral patterns experienced In childhood pro-duce specific, potentially lifelong alterations of how ourminds work.

Some such cultures will correspond to conventionallyIdentified ethnic groups. 1 suspect that the vast majoritywill show only random correspondence to conventionallyIdentified ethnic groups. Some cultures may exhibit spe-cific forms of historical and regional variation (e.g.Dressier and Bindon 2000; Dressier et al. 1991). A broa*»


view may reveal dinal variation over space (e.g., Boas1894; Caulkins 2001; Driver 1970; Jorgensen 1980), ratherthan sharp boundaries. Other cultures may constitute hu-man universals (e.g., Handwerker 1999b). Howeverbroadly shared, all cultures, 1 suspect, will turn out to con-sist of recurrent patterns of behavior rationalized by whatwe otherwise call domain-specific theories, models, or,most generically, schemas (D'Andrade 1995; Kellogg1995).

Culture as the Superorganic Product of IndividualMindsSecond, only individuals learn, and individuals embodyand constitute the only source of cultural data. To makethe world we live in sensible, all of us assemble out of ourindividual sensory experiences ways of thinking aboutwhat we have experienced. We still understand onlyvaguely how we accomplish this, despite the constructionof an increasingly specific characterization of how ourminds work by neuroscientists and cognitive psycholo-gists. One basic point of reference is that the human cen-tral nervous system functions constantly (Barnett 1953;for another application of this observation, see Garro2000). Another is that we construct our understanding ofthe world without being conscious that we are doing so(Bargh and Chartrand 1999; Barnett 1953; Gazzaniga1998). The construction process invariably infuses ostensi-bly intellectual products with often intense and complexemotional associations. Our minds use this information torespond to the world of experience in patterned ways.They also rationalize this information as the conscious un-derstandings of the world of experience that we call theo-ries, models, or schemas. For a specific person, culture, be-cause it is constructed in that individual's mind out of theunique set and sequence of experiences that mark the tra-jectory of the person's life, embodies who that person is asan individual, what he or she knows and does, at specificpoints along that trajectory. As such, the individuallyunique system of mental constructions that allows each ofus to understand and respond to the world of experienceexhibits properties of superorganic wholes, too—a body ofknowledge that, at least metaphorically, we use to live ourlives.

For data collection purposes, these mental construc-tions consist of three phenomena:

labels, names that identify the existence of distinctconfigurations of phenomenal experience;definitions, which, however ambiguous in specificcases, differentiate one thing from another; andintellectual and emotional associations, which givemental constructions distinctive meaning.

In the field, data collection reveals culture as potentiallyephemeral beliefs, feelings, and behavior unique In theirdetails to each individual. No two people can live preciselyidentical life histories. No one person can experience all

things. The unceasing mental processes by which indi-viduals perceive, store, and manipulate information, and,so, create culture, make it physically impossible for anytwo people to hold identical cultural configurations. Theseprocesses make it physically impossible for any one personto hold identical cultural configurations at two points intime. Effective and efficient ethnographic research thushinges on a clear understanding of the implications of twoobservations:

Cultural differences reflect variation in personal expe-riences.Culture evolves, and, because it does, so do cultures.

Our Central Nervous Systems Construct Culture withSensory Input

The set of cognitions, emotions, and behavior thatuniquely identifies each of us as individuals reflects wherewe live and the web of social relations through which wehave lived our lives. This means

• where we were bom and raised;when we were born; andwith whom we have interacted, in what ways, andwhat we experienced at specific places and times overthe course of our lives.

All clinicians share a common set of understandings thatcome from their training in biomedicine, for example, butphysicians work with a body of knowledge that distin-guishes them from, say, nurse practitioners. Similarly,family practice physicians work with a body of knowledgedistinct from that used by surgeons. Variation in life expe-riences leads people to see the world differently and towork with distinctive bodies of knowledge. Older peo-ple—whether physicians, plumbers, or anthropolo-gists—share a distinctive vantage point owing only to age,and older anthropologists typically share a body of knowl-edge that distinguishes them from junior faculty or gradu-ate students, just as older plumbers typically share a bodyof knowledge that distinguishes them from apprentices.The knowledge of men and women the same age—sociolo-gists, developmental psychologists, or airline pilots—maydiffer solely because of gendered experiences. Men andwomen the same age may work with a common bodv ofknowledge merely because they grew up in poverty or ex-perienced the privileges bestowed by wealthy parents. Be-cause they share an ethnic heritage, Puerto Ricans, irre-spective of age, gender, and class, may use a commonbody of knowledge, which may differ significantly from abody of knowledge shared by Mexicans. Fathers—whetherPuerto Rican or Mexican, Eskimo or Navaho, whether phy-sicians or nurse practitioners, whether old or voung, richor poor—may share a body ot knowledge simply becausethey share the experience o\ being fathers Academics,whether they live in China, Russi.1, Nigeria, Mexico, or theUnited States, share a distinctive culture irrespective otother differences.


In short, except for the set of understandings and be-havioral patterns that make each of us unique, no one pos-sesses or participates in a single culture. As Sapir (1932)pointed out, everyone participates in many cultures. Byvirtue of its mode of construction, the cultural configura-tion of cognition, emotion, and behavior unique to indi-viduals necessarily contains elements shared variouslywith others. When you first meet someone, you cannottell very well by that person's age, gender, dress, or skincolor which cultures you share and which you do not.Knowledge about the visual signals of cultural differencescomes only from listening for the right cues—speech orother behavioral patterns that make no sense or do notform part of one's personal cultural repertoire. Variationin historical and regional context provides one set ofchoices to some people and a different set of choices toothers, and it produces different historically and region-ally specific cultures. Both Connecticut Yankees andPuerto Rican migrants to Connecticut, for example, bringto parenthood assumptions constructed from their experi-ences with family members, teachers, friends, and othersocially significant people, which may vary by birth co-hort, gender, and specific social origins. But we cannotknow how or even if these assumptions exhibit patternedvariation without careful examination. And we cannot tellfrom shared historical and regional origins alone whetherConnecticut Yankee natives and Puerto Rican migrantsemploy distinctive cultural models of what constitutesand how to organize a working relationship with theirchild's teacher.

Culture Evolves. Which Means That Cultures Evolve

Over the course of our lives and through variousmeans—listening to news reports, reading, traveling, talk-ing with friends or family members, taking courses or at-tending workshops, or explicit and rigorous research—wecome to think differently about the components of ourworld. We think of new ways to organize activities andnew ways to think about domains of understanding. Wetend to incorporate into our lives the new ways of think-ing about the world that we infer yield better results. Byinteracting with other people—acting and responding towhat we experience of other people's words and acts—weactively participate in an unceasing process that leads tothe evolution of what we know and how we act and pro-duces the shared understandings and patterns of behaviorvisible in temporal and spatial autocorrelation (e.g.,Handwerker and Wozniak 1997). Because culture evolves,it makes a moving target. Ethnographers who have manyyears of experience working with a specific populationmay miss cues that signal Important new cultural differ-ences if they do not keep up with the cultural evolutionthat goes on around them all the time.

And because the culture of Individuals evolves, so dothe cultures to which each of us contribute. The culturethat Puerto Rican migrants bring with them to the main-

land evolves because new cultural environments providenew forms of experience for their minds to process. Simul-taneously, the way in which migrants make sense of andrespond to specific experiences in a new cultural environ-ment itself constitutes a new cultural environment, a non-random change in recurrent behavioral patterns, for theminds of host country natives to process. Migrants thuscontribute to ongoing cultural evolution in their host re-gions. The specific contribution may vary with age, gen-der, education, travel patterns, and the patterns, networks,and character of social networks, including experiencewith people from different ethnic and social class back-grounds, experience with health and education profes-sionals, household composition and caregiving responsi-bilities, or other contingencies. Certain forms of culturalevolution in individuals, which we have yet to specifyvery precisely, produce changes in the recurrent patternedbehavior that constitutes the superorganic environmentsof cultures. One such form, however, may consist ofchanges that shift the balance of power in social relation-ships (e.g., Handwerker 2001a; Wolf 1982).

Naive reification and essentialization of the kind Jor-gensen (1995) describes appreciate the superorganic prop-erties of cultures but fail to grasp that, although culturesmay exert effects, they do so by virtue of eliciting our re-sponse to important forms of sensory information thatcome from recurrent behavioral patterns. Cultures thusexhibit some of the properties of things. But they cannotbe smashed or damaged. They just evolve.

In short, Tylor's definition of "culture" contains twolargely overlooked and undervalued implications: (1) dieculture that specific people use to live their lives consti-tutes an evolving configuration of cognition, emotion,and behavior unique to themselves; and (2) cultures con-sist of an evolving configuration of cognition, emotion,and behavior at the intersection of evolving, individuallyunique cultural sets. The central problem for ethnographythus consists of identifying and describing that intersec-tion and its significant variations. Ethnography becomeethnology when we try to explain by reference to perti-nent life history variables the evolution of specific culturesor the culture of specific people.

Toward the end of the 20th century, global movementand communication placed individuals who embodieddifferent recurrent patterns of thought and behavior (cul-tures) face to face. The challenges of understanding andworking with cultural diversity In the 21st century maycome principally, however, from standing fact to facewith people whom we might think embody different cul-tures but who do not. To make sense of the changed cir-cumstances of our lives, we must pay specific attention tothe domains of cognition, emotion, and behavior perti-nent to Individual lives and explicitly establish which as-pects of these domains one person shares with which spe-cific others. This creates questions bearing on the dettibof the social distribution of cognition, emotion, and be-havior and about the patterns, networks, and chutcKf tf

Handwerker Construct Validity of Cultures 111

social interaction through which culture evolves, locally,regionally, and globally. This requires attention to the de-sign of ethnographic research and specific appreciation ofthe unique point of view ethnography offers—its focus onsimilarities and differences among informants rather thanamong variables.

Standard approaches to numerical analysis do notlend themselves to ethnography because they focus onvariables, not informants. Consensus analysis procedures(e.g., Romney et al. 1986), by contrast, address the analyti-cal issues raised by the question of who agrees with andacts like whom about what and to what degree. However,consensus analysis was designed for analysis of a singleculture. The software implementation of consensus analy-sis can be employed in ways 1 illustrate later—but not eas-ily. More important, cultural consensus theory comes outof a specific branch of learning theory and contains highlyrestrictive assumptions. One central assumption, notoften met in field research, is that there exists a specificbody of knowledge bearing on a specific cultural domainfor which there exist cultural experts who know a lot andothers who know less (see Garro's [2000] excellent review).Where this cultural pool model of the distribution of cul-ture exhibits empirical content (e.g., Weller and Mann1997), consensus analysis answers the analytical questiondirectly.

Under ordinary field conditions and for nearly all eth-nographic goals, however, everyone constitutes a culturalexpert in what she or he knows, feels, and does—likewomen who report on the meaning of stress and socialsupport or parents and teachers who report on what goesinto an effective working relationship between parentsand teachers. Moreover, the culture that rationalizes ourindividual worlds of experience appears to be organizedhierarchically (Garro 2000; see Handwerker 1989). Thesestarting points make the question of who agrees withwhom about what and to what degree bear directly on theempirical problem of identifying many possible culturesand the boundaries between one culture and another. Italso makes the answer the simple outcome of standardconstruct validation procedures, akin to the proceduresBoas (1894) used more than a century ago to evaluate thedistribution of cultural phenomena across space and (im-plicitly) time.

ESTABLISHING THE CONSTRUCT VALIDITY OFCULTURES

Construct validity refers to the observed match between aset of observations and the theoretical construct it pur-ports to measure (Campbell 1970). Our individual cultureprovides the material by which our minds rationalize (In-terpret) sensory input from the world of experience, and avariety of mental processes alter both culture and behaviorin ways that reflect variation in sensory Input. One out-come consists of the assumptions from which we con-struct our understandings of the world. Because our theo-

ries thus constitute cultural constructions, they necessarilycontain biases that mirror everything we have experiencedand everything we have not experienced over the courseof our lives. This makes for a fundamental uncertaintyconcerning the empirical content of mental constructionsand makes it essential to distinguish what is there fromwhat we may put there. Validity consists of a relationshipbetween the definitions of specific mental constructionsand specific observations. One cannot evaluate the valid-ity of one's data or findings without information aboutwhether or not, or the degree to which, specific mentalconstructions correspond with specific observations anddiscriminate those observations from others.

Many critically important variables, such as stress, af-fection, and problem drinking, cannot be measured directly.Indeed, we must posit their existence because we cannotsee them. Measurement of multidimensional constructslike these requires specific observations of a set of items,each of which constitutes an independent and imperfectmeasure of the otherwise unseen, underlying variable. Thefollowing questions, for example, elicit measurements ofdifferent dimensions of problem drinking (Mayfield et al.1974): How often have you tried to cut down your drink-ing? gotten annoyed with people who complain aboutyour drinking? felt guilty about your drinking? or had adrink when you first get up to feel better? As Campbelland Fiske (1959; see Campbell 1970) point out, items thatmeasure the same theoretical construct should correlatehighly. Items that measure a second construct should notcorrelate as highly with the items that measure the first. Ifthere exists such a variable as problem drinking, and if theseitems measure that variable, we will see it as a large inter-section shared by the four items, any one of which meas-ures problem drinking imperfectly. A large shared varianceamong these items means that people who report oftentrying to cut down their drinking also tend to report thatthey often get annoyed, feel guilty, or have a drink whenthey first get up; people who report that they never orrarely try to cut down their drinking also tend to reportthat they never or rarely get annoyed, feel guilty, or have adrink when they first get up.

Use Factor Analysis to Identify the Intersectionamong Sets of VariablesCluster analysis, multidimensional scaling, correspondenceanalysis, and other numerical reasoning tools answerquestions about patterned relationships among many vari-ables. However, only factor or principal componentsanalysis (e.g., Rummel 1970) directly tests the hypothesisthat a specific set of scale items constitutes an inde-pendent set of imperfect measurements of one and onlyone otherwise unseen, underlying variable. Principal com-ponents analysis constructs a small set of variables (factorsor principal components) from additive combinations ofexisting similarities among variables. Each I actor thusIdentifies the existence of otherwise unseen variables that

112 American Anthropologist - Vol. 104, No. 1 • March 2002

lie at the intersection of observed similarities among thevariables measured. The size of the intersection tells us theimportance of the factor. Factor loadings (Pearson's coeffi-cients) measure the size of the intersection. The square ofa loading (r2) tells us how much variance a specific itemshares with the unseen variable identified by each factor.Important measures of this unseen variable show loadingsat or above 0.500 (25 percent shared variance). The sum ofsquared loadings for a factor (its eigenvalue) tells us howmuch variation all our cases or variables share with a fac-tor. A factor's eigenvalue divided by the sum of eigenval-ues for all factors tells us the overall size of the shared in-tersection identified by the factor.

The first factor or principal component identifies thelargest shared intersection among a set of variables. Thesecond factor accounts for the largest shared intersectionthat remains. Subsequent factors account for the largestshared intersections among the variance unaccounted forby previously extracted factors.

Principal components analysis yields one factor forevery variable. Four items will form a 4*4 similarity matrixand will yield four factors, for example. Evidence of a sin-gle valid factor consists of the following:

1. A first factor with an intersection that accounts for50 percent of the variance in the matrix (or more).

2. A sharp scree fall between the eigenvalues for thefirst and second factors. Scree refers to the jumble ofrocks and soil that accumulates at the bottom ofcliffs. A scree plot shows the size of the eigenvaluesfor all factors beginning with factor 1. Just bychance, sample items may show great similarity.Factor analysis of matrices filled with random datawill find some eigenvalues over 1.0 and some highloadings just by chance. But matrices that containno real factors will generate random distributions ofeigenvalues and loadings. One-half of the eigenval-ues will be over 1.00, one-half will be under 1.00,and we will see no perceptible scree. Matrices thatcontain one real factor will exhibit a dramatic screefall between the first and second eigenvalues; thefirst eigenvalue will be approximately three or moretimes larger than the second.

3. The eigenvalue of the second factor lies at the topof the scree.

4. There exist no (or inconsequentially small) negativeloadings on factor 1.

5. There exist no (or inconsequentially small numbersof) high (+/-.50) loadings on factor 2. This last con-dition gives important diagnostic Information. AsRuinmel (1970:373) notes, because the procedurefits the first factor to the data to account for themaximum variance, when there exist two Inde-pendent clusters of Interrelated variables, factor 1may be located between them. All variables mayload highly on factor 1. But the variables will alsoload highly on factor 2, and the variables that con-

stitute each independent cluster will exhibit differ-ent signs on factor 2.

Evidence of these five conditions shows that people wholive dramatically different lives—the descendants of for-mer plantation slaves who make a living in tourist econo-mies in the West Indies and people who continue to hunt,gather, and herd in the Alaskan and Siberian Arctic—agreeabout experiences that make up unitary phenomena, oth-erwise unseen, legitimately called violence and affection(Handwerker 1997). Similar evidence shows that, irrespec-tive of ethnic identity (European American, Latino Ameri-can, African American, and Native American), workingwomen agree about experiences that make up unitary phe-nomena, otherwise unseen, legitimately called stress andsocial support coming from daily social interaction, anddepression (Handwerker 1999b).

For Ethnography, Use Factor Analysis to Identify theIntersection among Sets of Informants

One can easily answer questions like, "Do Inupiat Eskimogive answers or act in ways that differ from West Indians?'Simply divide the replies by identity, count the responsesfor each, and compare the results. As Keesing (1994)points out, however, this procedure yields answers thatbeg the question of the location of cultural boundaries.Questions like these impose cultural differences by as-sumption rather than evidence. To avoid this error and topinpoint cultural boundaries, if they exist, use principalcomponents to establish the construct validity of the cul-ture or cultures in the data. Findings from such an analysiswill tell us whether or not or the extent to which culturalboundaries correspond with social labels like Nuer orNavajo, men or women, or old or young.

One cannot see cultures, however. To see them, onehas to identify what people share, the configuration ofcognition, emotion, and behavior that constitutes the in-tersection of configurations unique in their details to indi-viduals. Cultures thus constitute multidimensional vari-ables, just like variables such as stress, affection, andproblem drinking. To validate this construct, one mustdemonstrate that a specific set of people shares a specificconfiguration of cognition, emotion, or behavior (see, eg.,Handwerker 2001a, 2001b). As Campbell and Fiske (1959;see Campbell 1970) imply, informants who think or actalike provide information that measures the same theo-retical construct (culture). Informants who think or act inother ways provide information that measures a secondconstruct (another culture). The Ideas and behavior of thefirst set of Informants should correlate highly; the ideasand behavior of the second set of informants should notcorrelate highly with the Ideas and behavior of the firstset. If there exists a culture of parent-teacher working rela-tionships, and one has correctly identified Its compo-nents, a principal components analysis of the similaritiesIn the Information provided by Informants will reveal»large intersection shared by those Informants, any one of

Handwerker Construct Validity of Cultures 113

whom provides only imperfect information about the cul-ture. A large shared variance among informants thus tellsus that a culture exists—a shared cognitive model of theresponsibilities and activities that form important compo-nents of parent-teacher relationships, a shared cognitivemodel of how to organize those components effectively,shared recurrent patterns of behavior between parents andteachers, or both one or more cognitive models and one ormore recurrent behavioral patterns.

Establishing the construct validity of a culture thus re-quires that we demonstrate that what some informantstell us or do correlates highly with what other informantstell us or do. Informants who have constructed and par-ticipate in a different culture will say things and act inways that correlate highly among themselves, but whatthey say or do will not correlate as highly with another setof informants. To evaluate the construct validity of cul-tures, carry out a standard construct validity analysis buttranspose the matrix to focus the analysis on similaritiesand differences among informants rather than amongvariables (for practical details, see Handwerker 2001b).

CULTURAL DIVERSITY IN PARENT-TEACHER WORKING

RELATIONSHIPS

Over the last 30 years, U.S. educators and parents haveforged a consensus that effective education requires an ef-fective parent-teacher partnership (Cutler 2000). Al-though there exist many models of how best to imple-ment that relationship (e.g., Comer et al. 1999; Cutler2000; Epstein 1995; Henry 1996), there exists growing rec-ognition that cultural differences bear importantly on ef-fective parent-teacher involvement (e.g., Hidalgo et al.1995; Lareau and Horvat 1999). Forging effective relation-ships between the teachers and parents of Latino students,an increasing proportion of whom speak Spanish as theirfirst language, constitutes a particularly pressing issue. Notonly do Latino students make up the largest minoritygroup in U.S. public schools, they also exhibit both poortest scores and frighteningly high dropout rates, the high-est in the nation.

What makes an effective parent-teacher working rela-tionship, and why some parents were not involved intheir children's education, thus formed a natural studytopic in a graduate seminar on ethnographic methods 1taught in 1996 for teachers working toward graduate de-grees in bilingual education, all of whom worked withPuerto Rican students and many of whom came from thePuerto Rican community. We wanted to find out

• the items or components people thought of as makingup an effective parent-teacher working relationship,the weighting people assigned different components,andhow they organized these elements to construct an ef-fective parent-teacher working relationship.

In short, we aimed, as Hidalgo et al. urge, to look at "howmuch is similar and how much is different between andamong families of different cultural groups" (1995:515).

We did not, however, define cultural groups by refer-ence to ethnicity. We defined a cultural group as a groupof people who share a common way of thinking about orenacting parent-teacher working relationships, and we de-signed our data collection and analysis in ways that al-lowed us to both explicitly identify the cultural group(s) inour data and explicitly assess what may have shaped it(them).

Informants and Data Collection

We anticipated that ethnicity might contribute to a sharedculture bearing on parent-teacher working relation-ships—that Puerto Ricans might share a distinctive way ofthinking about or enacting parent-teacher relationships,which might exhibit systematic, even qualitative, differ-ences from the way, for example, Connecticut Yankee par-ents thought about or enacted the same issues. We alsoanticipated that teachers might share a distinctive way ofthinking about or enacting parent-teacher relationships,which might exhibit systematic, even qualitative, differ-ences from the way parents thought about or enacted thesame issues. We allowed for the possibility that either in-tercultural or intracultural variation might reflect othersets of life experiences, like those shaped by gender, age,number of children in a family, whether or not an inform-ant worked, and whether or not a partner helped. Finally,we wondered if the point of view of parents whom teach-ers evaluated as actively engaged in an effective par-ent-teacher working relationship differed from the pointof view of parents whom teachers evaluated as not so en-gaged.

We collected data in two stages. In the first, we con-ducted informal interviews with teachers, students, andparents whom teachers evaluated as being actively en-gaged in effective parent-teacher working relationshipsand two different forms of structured interview's withteachers and actively engaged parents (n = 49). To answerthe first question, we asked teachers, students, and parentsto tell us what they thought was part of an effective par-ent-teacher working relationships and why it was impor-tant. To answer the second question, we asked teachersand actively involved parents to rate (from "not at all im-portant" to "very important") each of the componentsidentified in the informal interviews. To answer the thirdquestion, we asked the same parents and teachers to evalu-ate the relative similarity of these components. We elic-ited these judgments with triads tests using a Lamba-2 de-sign (see Burton and Nerlove 1976). We carried outinterviews in English or Spanish, as appropriate. In theprocess, we collected data on life experience variables totest for their effects on intracultural or intercultural vari-ation. Teachers constituted only IS percent of the basesample of 49 Informants. Most Inlonnants (Sb percent)

114 American Anthropologist • Vol. 104, No. 1 March 2002

identified themselves as Puerto Ricans, and 37 percentlooked at themselves as one or another version of Con-necticut Yankee; the residual consisted of two West Indi-ans, one African American, and one South Asian. Nearly70 percent were women, 76 percent were currently mar-ried, and 84 percent currently worked. Informants rangedin age from 21 to 47 (mean = 35.8 years, s.d. = 6.6). Twoinformants had no children, but one had five (mean = 2.0children, s.d. = 1.2). Two informants had stopped schoolafter the primary grades, but a couple were doctoral stu-dents (mean = 15.5 years of education, s.d. = 3.1).

In the second stage, we conducted triads tests with tenparents whom teachers evaluated as not actively engagedin an effective parent-teacher working relationship. Thismade n = 59 for analysis of the triads tests.

Findings for Question #1: What Items Go into Makingan Effective Parent-Teacher Working Relationship?

Answers to the first question came from the informal eth-nographic interviews. Nearly 30 teachers, parents, andchildren, who spoke from their experience in primary,middle, or high school, identified ten items as potentialcomponents of an effective parent-teacher working rela-tionship:

1. parents who read with their children (READ)2. parents who help their children with homework

(HELP)3. parents who participate actively in school activi-

ties (ACTIVE.SCHL)4. parents who provide the space and materials nec-

essary for their children to carry out homework as-signments (SPACE.MAT)

5. parents who require their children to completehomework assignments (REQ_HMWRK)

6. parents who provide rewards for good grades andconsequences for bad ones (R_C_grades)

7. school policies and programs that reach out toparents (SCHL_REACH)

8. active teacher-parent communication (TPCOM)9. parents who listen to their children to hear their

needs (LISTEN)10. parents who respond to their children's needs (RE-

SPOND)

Findings for Question #2: What Is the RelativeImportance of These Items?

Answers to the second question come from analysis of a49*49 matrix that expresses the degree to which our In-formants rated (from "not at all important" to "very im-portant") the relative importance of these ten Items Iden-tically. Although we collected ratings data with afive-point scale (0-4), many Informants rated all Items as"very Important," which left no variance for the calcula-tion of correlation codtlclents. I created binary variablescoded "1* for ratings of "Important" or "very Important"and "0" otherwise and measured similarities among In-

formants with a simple matching coefficient. Figure 1shows the scree plot from a principal components analysisof the resulting 49*49 informant matrix and illustrates evi-dence for the construct validity of a single cultural group.The first factor's eigenvalue (43.324) was 23 times largerthan the second factor's eigenvalue, and the latter lies onthe top of the scree. The first factor accounted for 88.4 per-cent of the matrix variance; informant loadings averaged.935 (s.d. = .101).

Figure 2 shows two scatterplots of informant loadingson factor 1 by informant loadings on factor 2. The cluster-ing of informants is so tight, 1 added slight amounts ofrandom variation to each informant so that we can seetheir location. Both scatterplots reveal that nearly all casesform a tight cluster centered around a 1.00 loading on fac-tor 1 and a 0.00 loading on factor 2. Figure 2 reveals noevidence of two or more cultures because no informant ex-hibited a high loading (+/-.50) on factor 2 coupled with alow loading on factor 1. On the contrary, this analysis ex-plicitly confirms what a visual examination of the ratingsdata suggests—that nearly every informant rated nearlyevery item as an important or very important componentof an effective parent-teacher working relationship.

Figure 2 reveals intracultural variation, however. Asmall number of informants exhibited a negative loadingon factor 2. An informal (visual) or formal (ordinary leastsquares [OLS] regression) examination of the relationshipbetween informant ratings of different items and the factorloadings (which identify the shared intersection amongthese informants) reveals that this intracultural variationreflects variation in judgments about the importance of a

Scree Plot

10 20 30 40Number of Factors

FIGURE 1. Scatter plot of eigenvalue by factor.

SO


single item—the importance that informants ascribed toproviding rewards for good grades and consequences forbad ones. The scatterplot on the left of Figure 2 plots in-formants with circles of uniform size. The scatterplot onthe right identifies teachers with a T and parents by eth-nicity (P = Puerto Rican, Y = Connecticut Yankee, and O =other ethnicity). This scatterplot also weights symbol sizeby the importance the informant ascribed to providing re-wards for good grades and consequences for bad ones. Theonly demonstrably nonrandom intracultural variation (r =8.322, p < 001) comes from a small number of parentsand one teacher who rated the importance of providing re-wards for good grades and consequences for bad gradeslower than most parents.

Findings for Question #3: How Do People OrganizeThese Items?

Answers to the third question come from analysis of astacked set of 59 matrices. Each matrix contains coeffi-cients that measure a specific informant's judgment aboutthe similarity among the items that go into an effectiveworking relationship between parents and teachers basedon the Lambda-2 triads design. The question answered bythe factor analysis is the extent to which all 59 informantsevaluated the similarity among these ten items identically.Without a close reading, factor analysis output alone maysuggest a single culture: factor 1 accounts for 74.58 percentof matrix variance, and its eigenvalue is 5.3 times largerthan that of the second. However, a large number of lowloadings on factor 1 and several very high loadings on fac-tor 2 exist, and factor 2 accounts for another 15 percent ofmatrix variance. A scree plot would show that the eigen-value of the second factor lies above the scree. Graphicalanalysis thus forms an essential diagnostic tool.

Figure 3 shows the answer produced by the factoranalysis in two scatterplots of informant loadings on fac-tor 1 by informant loadings on factor 2. Where Figure 2 il-lustrated a scatterplot that exhibits evidence of a singleculture, Figure 3 illustrates a scatterplot that exhibits evi-dence of two cultures plus individually unique culturalconfigurations (random data). Informant labels of 1, 2,and 3 identify the two cultures (1,2) and random data (3)in the scatterplot on the left. One culture (1; Separate butEqual—1 describe the culture shortly) consists of 38 in-formants who exhibited high loadings on factor 1 and lowloadings on factor 2. The other culture (2; Mutual DecisionMakers—1 describe the culture shortly) consists of seveninformants who exhibited low loadings on factor 1 andhigh loadings on factor 2. The residual (3) of 14 inform-ants exhibited low loadings on both factor 1 and factor 2.

These findings tell us that 38 informants generallyagreed that the ten items that make up an effective par-ent-teacher working relationship should be organized inone way, that seven informants agreed that these itemsshould be organized in a qualitatively different manner,and that 14 informants exhibited no agreement at all. Thescatterplot on the right identifies teachers with a T, a par-ent who was not actively involved in an effective par-ent-teach working relationship with an n, and a parent ac-tively involved in an effective parent-teacher relationshipby ethnicity. Puerto Ricans (P), Connecticut Yankees (Y),and others (O—West Indian, African American, and SouthAsian) show up in both cultures. One teacher, two PuertoRicans, and two Yankees show up in the set of randomdata (no cultural patterning), but nine of the 14 inform-ants who exhibited no shared agreement were parents notactively involved in an effective parent-teacher working

-1.0 -0.5 0.0 0.5Loading on Factor 1

1.0 -0.5 0.0 0.5Loading on Factor 1

1.0

FIGURE 2. Scatterplot of loadings on factor 2 by loadings on factor 1 for ratings data

116 American Anthropologist Vol. 104, No. 1 • March 2002

relationship. Only one noninvolved parent shows up inone of the cultures.

Once we disaggregate the original matrix into threedistinct sets, we find evidence of the kind illustrated ear-lier for the existence of a single culture. Factor analysis of astacked set of 38 matrices that contain similarity judg-ments about the items that go into an effective workingrelationship between parents and teachers for each of 38informants yields a first factor that accounts for 82 percentof matrix variance, the first eigenvalue of which is 8.5times larger than that of the second, and loadings that av-erage .679 (s.d. = .178). Analysis of a stacked set of sevenmatrices that contain similarity judgments about theitems that go into an effective working relationship be-tween parents and teachers for each of seven informantsyields a first factor that accounts for 62.8 percent of matrixvariance, the first eigenvalue of which is three times largerthan that of the second, and loadings that average .564(s.d. = .122).

The third set of informants who exhibit only randomsimilarities and differences, most of whom teachers evalu-ated as not engaged in effective parent-teacher relation-ships, suggests the existence of a subpopulation that sim-ply may not know how to effectively organize theelements of parental involvement, even if they show fa-miliarity with some or all of its elements. Effective behav-ior presupposes planning, organization, and coordinationof clearly identified activities. But not everyone has putthe same effort into thinking about these issues, and noteveryone has the same degree of experience carrying outthese activities. Moreover, variation in life experiencesmean that people see the world differently. They may seedifferent things. Or they may not see what other peopledo. The core issue for parents who remain uninvolved in

their children's education may be simply that they do notknow how to go about it. They work with no organizedmap of the things that go into the construction of an ef-fective working relationship between parents and teach-ers, such as those shown in Figure 4. Some informantswho belong to this group may have worked out an indi-vidually unique way to organize a parent-teacher relation-ship. However, the absence of a shared model of par-ent-teacher relationships probably makes putting togetheran effective one difficult to achieve.

Figure 4 shows the differences between the two cul-tures as multidimensional scaling maps of the aggregatesimilarities among the ten items. Multidimensional scal-ing (MDS) transforms a matrix of similarities or differencesinto a map (e.g., Kruskal and Wish 1978). MDS transformsa matrix of distances among cities in the United States(city block dissimilarities) into map coordinates of theUnited States, for example. MDS coordinates thus expressstrong similarities (small dissimilarities) as spatial close-ness and weak similarities (great dissimilarities) as spatialdistance. Like the Cartesian coordinates that mapmakersuse, the axes of an MDS picture are arbitrary and uninter-pretable. The only thing that matters is the relative place-ment of cases or variables on the MDS map. A value calledstress measures the extent to which mapping distorts thedistance between the mapped items. Stress varies between0 and 1. A stress value of 0.0 means that no distortion tookplace. Stress values of .083 and .082 for the map on the leftand the map on the right, respectively, constitute evi-dence of an excellent fit between the original matrix ofsimilarities and the mapped coordinates (see Sturrock andRocha 2000).

The Separate but Equal culture organizes an effectiveworking relationship between parents and teachers into

CM 0 . 5 -o73n>u.

oO)

Io-OS-

-1.0 -0.5 0.0 0.5Loading on Factor 1

1.0 -0.5 0.0 0.5Loading on Factor 1

FIGURE 3. Scatter plot of loadings on factor 2 by loadings on factor 1 for similarity data.


one set of activities focused on the school (school policiesand programs that reach out to parents, being active in theschool, and parent-teacher communication) and a distinctset of home activities arranged along a continuum fromimpersonal activities (providing space and materials sochildren can complete homework assignments) to highlypersonal activities (listening for and responding to chil-dren's needs). In this culture, rewarding good grades andproviding consequences for bad ones constitute a schoolactivity carried out by parents. Listening for and respond-ing to children's needs constitute activities on the periph-ery of an effective working relationship between parentsand teachers, ones that may but because a child's needs gofar beyond school performance, do not necessarily con-tribute to the parent-teacher relationship.

The Mutual Decision Makers culture organizes an ef-fective working relationship between parents and teachersqualitatively differently. In this culture, one set of activi-ties focuses on active engagement in a child's education(responding to their needs, requiring the completion ofhomework, and rewarding good grades and providing con-sequences for bad ones). The other set integrates the mate-rial infrastructure for a child's education (on one end, pa-rental provision of space and materials and, on the other,school provision of policies and programs that reach outto parents) with parental activities that entail listening forchildren's needs, reading with them, helping with home-work, and discussing home and school issues with achild's teacher.

The MDS maps thus identify cultural differences whoseeffects may include conflict, disappointment, and frustra-

tion—or a comfortable working relationship between ateacher and a parent. For example, the first cultural modelof parent-teacher relationships identifies a clear separa-tion of school and home activities. Parents have one set ofresponsibilities and activities to perform. Teachers haveanother. Parents and teachers complement each other, buttheir domains of responsibilities and activities remainseparate. By contrast, the second cultural model of par-ent-teacher relationships implies mutual and overlappingdomains of responsibilities and activities. Requiring that achild complete homework assignments and providingrewards for good grades and consequences for bad onesremain solely a parental responsibility. But the continuumthis culture employs to organize the other components ofa parent-teacher relationship draws no boundary betweenhome and school. One mother explained: "You want tocreate a support system for your child that extends fromhome to school and back again." A parent who employsthis model may expect suggestions and direction from thechild's teacher for activities like reading to children andhelping with homework. The same parent may expect toprovide an active voice in classroom and school activities.As one such parent commented, "I know what's best formy child!" A teacher who employs this model may givesuggestions and direction for home activities and expectparents to provide an active voice in classroom and schoolactivities. Parents (teachers) who try to create effective work-ing relationships based on the Separate but Equal modelwith teachers (parents) who employ the Mutual DecisionMakers model set themselves up for the classic symptomsof clashing cultures—misunderstandings, disappointments,

Separate but Equal Mutual Decision Makers

1 -CMco

"55§

-2

1

-

CTIVE_SCHLO

iCHL.REACHO

TPCOMO

-

t

1 1

OSPACE_MAT -

OREQ_HMWRK

OR.Cjndet

OHELP

OREAD

L1STENO ORESPOND

1 1

- 2 - 1 0 1Dimension 1

egc

. 0"tocQ)

5

1 -

o-

-1 -

-2

I

OR_C_gradM

RESPONDO

REQ_HMWRKO

-

I

I 1

OSPACE_MAT

OL1STEN OREAD

OHELP

OTPCOM

OACTIVE.SCML

OSCHL_REACH

1 1

-2 -1 0 1Dimension 1

FIGURE 4. Multidimensional scaling map for two cultural models of parent-teacher relationships.


frustration—and ineffective if not hostile parent-teacherrelationships. A mother who employed the Mutual Deci-sion Makers model but continually ran into teachers whoemployed the Separate but Equal model explained: "1 feelsupmtrongly about this. I've given talks about it [to par-ent-teacher groups]! 1 really hated the ones [teachers]when you went in to talk about your child and all theytalked about was what they were doing in the classroom' 1hated them to the degree that they eventually hated me. 1just wouldn't give up."

Thus, there exist one cultural model of the compo-nents that go into the construction of an effective par-ent-teacher working relationship, with some minor in-tracultural variation, and two cultures about how toorganize that relationship. The pertinent "cultural group1"consists of the parents and teachers who share the Sepa-rate but Equal or the Mutual Decision Makers culturalmodel. OLS regression provides a tool to test for the effectof one or another life experience on intracultural vari-ation, including measuring the extent to which that vari-ation corresponds to the difference between parent andteacher or to the differences between Puerto Rican andConnecticut Yankee. As a matter of course, one wouldcarry out an OLS regression analysis of the intraculturalvariation identified by variance in the loadings on bothfactors. Here 1 report only the analysis of factor 2 loadings,which, as Figure 2 shows, exhibited the only importantform of intracultural variation. Logistic regression pro-vides an equivalent tool to test for the effect of plausibleantecedents of different cultures.

The top part of Table 1 shows OLS regression findingsproduced by SYSTAT software which reveal that Connecti-cut teachers and parents in their twenties, thirties, and for-ties, whether men or women, married or not, or currentlyemployed or not, and irrespective of educational attain-ment and social identity (Puerto Rican or ConnecticutYankee), assessed the importance of the components of aworking relationship between parents and teachers inways that differ only by chance. The maximum conditionindex of 26.669 confirms the absence of multicolinearitydisturbances. As Figure 2 reveals, some parents and teach-ers downplayed the relative importance of assigning re-wards for good grades and consequences for bad ones. Butthe variables we thought might contribute to cultural di-versity clearly do not account for what we found, and theanalysis leaves open the question of the sources of this in-tracultural variation.

The bottom part of Table 1 shows logistic regressionfindings produced by LogXact software which reveal thatConnecticut teachers and parents In their twenties, thir-ties, ami forties, whether men or women, married or not,or currently employed or not, and Irrespective of educa-tional attainment and social identity (Puerto Rican orConnecticut Yankee), exhibited odds to employ the Sepa-rate but Equal or Mutual Decision Makers model of how toorganize a paii'iit-teadu? relationship that differ only bychance. The variables we thought might contribute to cul-

tural diversity clearly do not account for what we found,and this analysis, too, leaves open the question of thesources of intercultural variation.

Discordance between these cultures and the socialidentities of the cultural participants validates Keesing's(1994) claim that culture is not bounded in ways manypeople have long assumed. This highlights the conceptualerror implicit in Hidalgo et al.'s otherwise laudable admo-nition to look at "how much is similar and how much isdifferent between and among families of different culturalgroups" (1995:515). The pertinent "cultural groups" arenot Puerto Ricans and Connecticut Yankees. They are thepeople who constructed the Separate but Equal and Mu-tual Decision Makers cultural models. This also highlightsthe methods error implicit in designing research to look atdifferences between Connecticut Yankees, Puerto Ricans,Chinese Americans, or African Americans. If we divide thereplies by identity (class, gender, ethnicity, age, ad nau-seam), count the responses for each, and compare the re-sults, we impose cultural differences by assumption, notevidence. To avoid this error, ask who shares what withwhom; look for similarities and differences among yourinformants.

PAST AND FUTURE

Ethnography is partially about learning from teachersrather than studying subjects. It is also about determiningthe similarities and differences among our teachers. In thelate 1800s, Tylor and Boas made the point that similaritiesand differences among people exhibit recurrent patternsof thought and behavior over space and time. During thefirst half of the 20th century, Kroeber (e.g., 1948) andWhite (e.g., 1949) explored the notion that these recur-rent patterns constitute superorganic wholes, sui generisphenomena, while Sapir (e.g., 1949) has pointed out thatthe foundations for these patterns reside in the minds ofspecific people and find expression in specific social inter-actions. During the second half of the 20th century, wewrestled with the problem of how (or if) what our individ-ual brains store and process can find expression in recur-rent patterns with the properties of superorganic wholes,how we might accommodate the observation that indi-viduals vary, make choices, and exert control over theirlives with the observation that those same individuals findthemselves constrained by recurrent patterns with theproperties of superorganic wholes. Toward the end of the20th century, global movement and communicationtrashed the notion that region and history constrain theserecurrent patterns by placing individuals who embodieddifferent recurrent patterns of thought and behavior faceto face. We are still trying to make sense of the observa-tion that we find standing face-to-face with us peoplewhom we might think embody different recurrent patternsof thought and behavior but who do not.

This article takes as Its central problem this last obser-vation. My way to make sense of It rests on a solution fa

the more fundamental question of how we might accom-modate the observation that individuals vary, makechoices, and exert control over their lives with the obser-vation that those same individuals find themselves con-strained by recurrent patterns with the properties of su-perorganic wholes. Stated concisely:

• The ways in which our brains store and process sen-sory input mean that individually unique life trajecto-ries yield individually unique people whose choicesexert control over their lives.

• Critically important sensory input comes to us in theform of other people's behavior.

• Our cognitive, emotional, and behavioral response tothat input reflects our prior life history and the per-sonal configuration of culture (cognition, emotion,and behavior) that our brains have constructed fromthat history of experience.

• Our behavioral responses elicit equivalent cognitive,emotional, and behavioral responses from others. Wesummarize these conditional responses to responses associal interaction.

• Other people influence us, and so constrain what wethink and do, by means of their behavior—by whatthey do or do not do, by the circumstances of theirlives, and by their immediate responses to our re-sponses; we influence others likewise. By virtue of thesensory input they generate, social interaction and liv-ing our lives in the presence of others thus produce theevolution of the personal cultures we use to live ourlives. We cannot now characterize how that takesplace with any specificity. Evolution in the recurrent,

TABLE 1, Tests for sources of intracultural and intercultural variation.


patterned behavior that exhibits properties of su-perorganic wholes, however, may come from theevolution of a personal or shared culture that entailsa shift in the balance of power in social relations.

• Recurrent, patterned behavior exhibits the propertiesof a thing because recurrent behavior constitutes anenvironment in which we carry out daily activities,which elicits cognitive, emotional, and behavioralresponses. In eliciting these responses, recurrent,patterned behavior thus elicits evolution in the per-sonal configuration of culture (cognition, emotion,and behavior) our minds use to produce personalcognitive, emotional, and behavioral responses tofuture sensory input. Certain forms of recurrent,patterned behavior may induce specific lifelongchanges in how our minds work and in the behav-ioral trajectories of our lives.

• Recurrent, patterned behavior means that people withwhom we interact make patterned choices fromamong the alternatives they see.

• Patterned choices come from the application of spe-cific criteria to the choice alternatives provided bysensory input.

• Our minds rationalize recurrent, patterned responses inthe form of domain-specific theories, models, orschemas, which consist of assumptions about thenature, components, and organization of the worldof sensory experience.

A theory of culture as cognitive elements and struc-ture now dominates ethnographic research (D'Andrade1999; Dressier and Bindon 2000). But definitions of culture

Ordinary Least Squares Regression Test for IntraculturalVariation—Loadings on Factor 2 (R2 = .069, maximumcondition index = 26.669)

Variable Beta P(2-tail)

Ethnicity—Yankee compared with otherEthnicity—Puerto Rican compared with otherTeacher compared with parentMen compared with womenEducationInformants with partners compared with those withoutEmployed informant compared with those unemployedNumber of childrenAge

Logistic Regression Test for Intercultural Variation-Separate but Equal Culture (1) and Mutual DecisionMakers Culture (0)

Variable

Ethnicity—Yankee compared with otherEthnicity—Puerto Rican compared with otherTeacher compared with parentMen compared with womenEducationInformants with partners compared with those withoutEmployed informant compared with those unemployedNumber of childrenAge

.040-.042.009

-.122.005.002

-.057.019.007

0.393-0.4580.113-1.8650.3580.017-1.2930.6151.199

0.6970.6490.9910.0700.7220.9860.2040.5420.238

Odds Ratio2.7265.7740.0540.1080.9489.6651.1591.84St.127

P (2-tail)

0.6100.53h0.2230.1700.9020.22^0.9630.6090.05S


may focus on mental phenomena merely because mentalmodels provide our only means to make behavior sensible.Despite the difference between cognition and behavior,especially the troublesome and ambiguous relationshipbetween the two, there exist demonstrable patterns of be-havior that correspond with specific cognitive models.More importantly, perhaps, what counts in the real worldis what people do. Cognition and emotion respond to sen-sory input, and critically important sensory input comesto us in the form of what other people do. Who cares howparents and teachers think about organizing a parent-teacher working relationship? What counts is whether ornot it works. It does not work unless people do somethingabout it.

In the process of making it work, the culture of spe-cific parents and teachers will evolve. As the culture of spe-cific parents and teachers evolves, they will bring to theirinteraction with others new forms of sensory informationthat the minds of the people with whom they interact willturn into additional cultural evolution. Cultural evolutionin individuals that shifts the balance of power in social re-lationships, and other forms we have yet to specify veryprecisely, produce changes in the recurrent, patterned be-havior that constitutes the superorganic environments ofcultures (e.g., Handwerker 2001a; Wolf 1982). We will notmake sense of what people think, feel, and do, 1 suspect,without conceptualizing the interdependencies of all threecomponents. A more sensible definition of culture thusmight focus on the relationship between ideas and behav-ior, not on a single component. From this point of view,cultures consist of recurrent, patterned behavior rational-ized by domain-specific theories, models, or schemas.

The problematical relationship between ideas and be-havior calls for much rethinking. Until the 1950s, gener-ally we maintained the definition of culture as encompass-ing both ideas and behavior. D'Andrade (1999) points outthat the shift in our thinking took place in an attempt toexplain what people do by reference to their ideas. Thisdid not work. For one, it produced tautologies. Once weexplain a particular pattern of behavior by reference to aparticular configuration of ideas, we rule out the possibil-ity that that pattern of behavior comes out of another con-figuration of ideas—if people really used that other con-figuration of ideas, they would have to produce a differentpattern of behavior. For another, the assumption that par-ticular configurations of ideas produce specific patterns ofbehavior could not account for the observation that indi-viduals vary, make choices, and exert control over their lives.

The response that sought to resolve this dilemma withmore nuanced interpretations of cultural variation re-tained the notion that Ideas explain behavior. This failedto reiognlze that all mental model explanations of behav-ior suffer fatal flaws. Human minds rationalize what we doby reierence to reasons, Intentions, and dispositions. Butwe can never rule out the possibility that, In accountingfor behavior, we or our Informants have not forgottensomething, mls|udged the relative Importance of some-

thing, ignored one or more multiple mental states thatmight go into the production of behavior, lied (to our-selves or others), or merely produced an ad hoc or posthoc fantasy. Most importantly, explanations of behaviorby reference to mental states garnered from our inform-ants fail to comprehend unconscious mental processesand the observation that consciousness of having donesomething or having thought something occurs after thedeed is done.

We face important ambiguities in the relationship be-tween behavior and ideas. When we rely on our personalor informant reports, we can find that people can do thesame thing for different reasons, can do different thingsfor the same reasons, and can claim to think one way andact in other, contradictory ways. Most tellingly, perhaps,changes in behavior can yield subsequent changes inideas.

The reasoning summarized above suggests that by fo-cusing solely on products of the mind—ideas and behav-ior—we created this dilemma ourselves. Certainly, thereexists an underlying logic that makes culture more than astatistical artifact, but it does not consist of ideas that pro-duce behavior. Perhaps, in reifying the ideas of individualsas the knowledge they use to produce behavior, our mindstricked us into attributing causal properties to a phenome-non that does not exhibit them? If so, the resolution tothe ambiguous and problematical relationship betweenideas and behavior may come from asking what producesboth.

Ideas and behavior reflect or constitute choices, so asolution to this dilemma-of-our-own-making may comefrom a theory of choice alternatives. Currently, econo-mists and behavioral ecologists work with a theory ofchoices, but the research agenda of neither concerns itselfwith the full range of pertinent questions. Core researchquestions stem from the observation that our central nerv-ous systems construct both the mental and the behavioralcomponents of culture. If recurrent, patterned behaviorexerts effects because it requires our response, somethingabout the specific properties of those recurrent patternsevokes our attention. What are these properties? Why dothey evoke our attention? How and why do we see alterna-tives or understand that alternatives even exist? By whatcriteria do we differentiate alternatives? Why do we em-ploy those criteria? What are the properties of the mentalmechanism or mechanisms that produce these effects?How and why did they become part of our central nervoussystems? Answers to these questions will yield a compre-hensive theory of choice alternatives. Such a theory willtake us beyond culture theory because It will imply a the-ory of culture. It will also help us make ethnographic senseout of the contemporary world and give us a better graspof the past one.

W. PENN HANDWERKER Department of Anthropology, Uni-

versity ot Connecticut, Storrs, CT 06269


NOTESAcknowledgments. The following graduate students in the doctoralprogram in bilingual education at the University of Connecticutmade significant contributions to research design and data collec-tion: Mayra E. Cruz, Madeline Delgado, Orlando Hernandez, Ther-e$e R- Horn, Yolanda Lopez, Gladys Menendez, Michael Packevicz,Madeline Ramos, and Robert A. Toth. Jim Boster pointed out thediagnostic utility of the second unrotated factor. Rose Jones andRobin Harwood let me play with their data (on women's mam-mography concerns and mothers' child-rearing goals, respectively)to work out issues with the method outlined here. They also pro-vided challenging commentary that proved of immense impor-tance. Russ Bernard, Bill Dressier, Rich Sosis, and AA referees pro-vided commentary, asides, and suggestions that helped make thisversion several orders of magnitude better than the original. Ithank you all.

REFERENCES CITEDBargh, John A., and Tanya L Chartrand

1999 The Unbearable Automaticity of Bei ng. American Psycholo-gist 54:462-479.

Bainett,H.G.1953 Innovation. New York: McGraw-Hill.

Boas, Franz1894 Indianische Sagen von der Nord-Pacifischen Kuste Amerikas.

Berlin: Asher.Boster, James S.

1987 Introduction. American Behavioral Scientist 31:150-162.Burton, Michael L, and Sarah Beth Nerlove

1976 Balanced Designs for Triad Tests. Social Science Research27:73-84.

Campbell, D. T.1970 Natural Selection as an Epistemological Model. In A Hand-

book of Methods in Cultural Anthropology. Raoul Naroll andRonald Cohen, eds. Pp. 51-85. Garden City, NY: Natural HistoryPress.

Campbell, D. T., and D. W. Fiske1959 Convergent and Discriminant Validity by the Multi-Trait,

Multimethod Matrix. Psychological Bulletin 56:81-105.Caulkins, Douglas

2001 Consensus, Clines, and Edges in Celtic Cultures. Cross-Cul-tural Research 35:109-126.

Comer, James P., Michael Ben-Avie, Norris M. Haynes, and EdwardT. Joyner, eds.

1999 Child by Child. New York: Teachers College Press.Cutler, William W., Ill

2000 Parents and Schools: The 150-Year Struggle for Control inAmerican Education. Chicago: University of Chicago Press.

D'Andtade, Roy1995 The Development of Cognitive Anthropology. Cambridge:

Cambridge University Press.1999 Comment. Current Anthropology 4O:S16-S17.

Dressier, William W., and James R. Bindon2000 The Health Consequences of Cultural Consonance. Ameri-

can Anthropologist 102:244-260.Dressier ,William W., F. E. Viteri, A. Chavez, G. A. C. Grell, and J. E.Dos Santos

1991 Comparative Research in Social Epidemiology: Measure-ment Issues. Ethnicity and Disease 1:379-393.

Driver, Harold E.1970 Statistical Studies of Continuous Geographical Distribu-

tions. In Handbook on Methods in Anthropology. Raoul Narolland Ronald Cohen, eds. Pp. 620-639. Garden City, NY: NaturalHistory Press.

Epstein, Joyce L.1995 School/FamIly/Communlty Partnerships: Caring f<>rthe

Children We Share. Phi Delta Kappan 76:701-712.Garro, Linda C.

2000 Remembering What One Knows and the Construction ofthe Past: A Comparison of Cultural Consensus Theory and Cul-tural Schema Theory. Ethos 28:275-319.

Gazzanlga, Michael S.1998 The Mind's Pasi. Berkeley: University of California Press.

Handwerker, W. Penn1989 The Origins and Evolution of Culture. American Anthropolo-

gist 91:313-326.1993 Gender Power Differences between Parentsand High Risk

Sexual Behavior by Their Children: AIDS/STD Risk Factors Extendto a Prior Generation. Journal of Women's Health 2:301-316.

1997 Universal Human Rights and the Problem of UnboundedCultural Meanings. American Anthropologist 99:799-809.

1999a Childhood Origins of Depression: Evidence from Nativeand Nonnative Women in Alaska and the Russian Far East. Jour-nal of Women's Health 8:87-94.

1999b Cultural Diversity, Stress, and Depression: WorkingWomen in the Americas. Journal of Women's Health and Gen-der-Based Medicine 8:1303-1311.

2001a Child Abuse and the Balance of Power in Parental Relation-ships: An Evolved Domain-Independent Mental MechanismThat Accounts for Behavioral Variation. American Journal of Hu-man Biology 13:679-689.

2001b Quick Ethnography. Walnut Creek, CA: AltaMira Press.Handwerker, W. Penn, and Danielle F. Womiak

1997 Sampling Strategies for the Collection of Cultural Data. Cur-rent Anthropology 38:869-875.

Henry, Mary E.1996 Parent-School Collaboration: Feminist Organizational Struc-

tures and School Leadership. Albany: State University of NewYork Press.

Hidalgo, Nitza M., Sau-Fong Siu, Josephine A. Bright, Susan M.Swap, and Joyce L. Epstein

1995 Research on Families, Schools, and Communities: A Mul-ticultural Perspective. In Handbook of Research on MulticulturalEducation. James A. Banks, ed. Pp. 498-524. New York: Simonand Schuster Macmillan.

Jorgensen, Joseph G.1980 Western Indians: Comparative Environments, Languages,

and Cultures of 172 Western American Indian Tribes. San Fran-cisco: W. H. Freeman.

1995 Ethnicity, Not Culture? Obfuscating Social Science in theExxon Valdez Oil Spill Case. American Indian Culture and Re-search Journal 19:1-124.

Keesing, Roger1994 Theories of Culture Revisited. In Assessing Cultural Anthro-

pology. Robert Borofsky, ed. Pp. 301-312. New York: McGraw-Hill.

Kellogg, Ronald T.1995 Cognitive Psychology. Thousand Oaks, CA: Sage.

Kroeber, A. L.1948 Anthropology. New York: Harcourt, Brace and World.

Kruskal, J. B., and M. Wish1978 Multidimensional Scaling. Newbury Park, CA: Sage.

Lareau, Annette, and Erin McNamara Horvat1999 Moments of Social Inclusion and Exclusion: Race, Class, and

Cultural Capital in Family-School Relationships. Sociology ofEducation 72:37-53.

Mayfield, D., G. McLeod, and P. Hull1974 The CAGEQuestionnaire: Validation of a New Alcoholism

Screening Instrument. American Journal of Psychiatry131:1121-1123.

McEvoy, F.1977 Understanding Ethnic Realities among the ilrebo and Kru of

West Africa. Africa 47:h2-78.Pelto, P., and G. Pelto

1975 lntracultural Diversity: Some Theoretical Issues. AmericanEthnologist 2:1-18.

Romney, A. Kimball, Susan C. Weller, and William H BatcheMcr1986 Culture as Consensus. American Anthropologist

88:313-338.Rummcl, R.J.

1970 Applied Factor Analysis. Evanston. 1L; Northwestern Univer-sity Press.

Saplr, Edward1932 Cultural Anthropology and Psychiatry. Journal ot Abnormal

and Social Psychology 27:229-242.1949 Selected Writing ot Edward Saplr. IXivul Mandelbauin, ed.

Berkeley: University oi Ollfornla l*revv


Sturrock, Kenneth, and Jorge Rocha Weller, Susan C , and N. C. Mann2000 A Multidimensional Scaling Stress Evaluation Table. Field 1997 AssesslngRaterPerformancewithouta"GoldStandanrU$.

Methods 12:49-60. ing Consensus Theory. Medical Decision Making 17:71-79.Tylor, Edward B. White, Leslie A.

1871 Primitive Society. London: J. Murray. 1949 The Science of Culture. New York: Farrar, Straus.Wallace, A. F. C. Wolf, Eric R.

1961 Culture and Personality. New York: Random House. 1982 Europe and the People without History. Berkeley: Universityof California Press.

Handwerker 2002 AA - The Construct Validity of Cultures

Documents

Transcript of Handwerker 2002 AA - The Construct Validity of Cultures