C H A P T E R 5 andMeasurement Holographic0verview Theinterrelatedstepsofconceptualization, opera- tionalization,andmeasurementallow researchersto turnageneralideafor aresearchtopicinto usefuland validmeasurementsin thereal world.An essentialpart ofthisprocess involvestransformingtherelatively I Introduction MeasuringAnythingthat Exists Conceptions, Concepts, and Reality Conceptions as Constructs Conceptualization Indicators and Dimensions The Interchangeability of Indicators Real, Nominal, and Operational Definitions Creating Conceptual Order An Example of Conceptualization: The Concept of Anomie DefinitionsinDescriptive and ExplanatoryStudies OperationalizationChoices Range of Variation Variationsbetween the Extremes IA Note on Dimensions Defining Variables and Attributes vaguetermsofordinarylanguageintoprecise objectsof study withwell-definedandmeasurablemeanings. rn Levels of Measurement Single or Multiple Indicators Some IUustrations of Operationalization Choices Operationalization Goes On and On Criteriaof MeasurementQuality Precision and Accuracy Reliability valldity Who DecidesWhat's Valid? Tension between Reliability and Validity MAIN POINTS KEY TERMS REVIEW QUESTIONS AND EXERCISES ADDITIONAL READINGS SOCIOLOGY WEB SITE 1 INFOTRAC COLLEGE EDITION Introduction This chapterand the next dealwith how researchers move from a general idea about what they want to study to effective and well-definedmeasurements in the real world. This chapter discusses the inter- related processes of conceptualization,operational- ization, and measurement. Chapter 6 builds on this foundationto discuss types of measurements that are more complex. We begin this chapter by confrontingthe hid- den concernpeople sometimes have about whether it's truly possible to measure the stuff of life: love, hate, religiosity,radicalism, alienation. The answer is yes, but it willtake a few pages to see how. Once we establish that researchers can mea- sure anything that exists,we'll turn to the steps in- volved in doingjust that. MeasuringAnflhing That Exists Earlier in thisbook, I said that one of the two pillars of science is observation.Becausethis word can suggest a casual,passive activity, scientistsoften use the term measure~~zentinstead, meaning careful, de- liberate observations of the real world for the pur- pose of desaibing objectsand events in terms of the attributes composing a variable. You may have some reservations about the abil- ity of science to measure the really important as- pects of human social existence.If you've read research reports dealingwith somethinglike liber- alism or religion or prejudice, you may have been dissatisfiedwith the way the researchersmeasured whatever they were studying.You may have felt that they were too superficial,that they missed the aspects that really matter most.Maybe they meas- ured religiosity as the number of times a person went to church, or maybe they measured liberalism by how people voted in a single election.Your dis- satisfaction would surelyhave inaeased if you had foundyourself being misclassifiedby the measure- ment system. Your feeling of dissatisfactionreflects an impor- tant fact about social research:Most of the variables MeasuringAnything That Exists .119 we want to study don't actually exist in the way that rocks exist. Indeed, they are made up. More- over, they seldomhave a single, unambiguous meaning. To see what I mean, suppose we want to study politicalparty affiliation.To measure this variable, we might consult the list of registered voters to note whether the people we were studyingwere registered as Democratsor Republicansand take that as a measure of their party affiliation.But we could also simply ask someone what party they identify with and take their response as our mea- sure.Notice that these two different measurement possibilitiesreflect somewhat different definitions of "politicalparty affiliation."They might even pro- duce Merent results: Someonemay have regis- tered as a Democratyears ago but gravitatedmore and more toward a Republicanphilosophy over time. Or someone who is registered with neither political party may, when asked, say sheis affiliated with the one she feels the most kinship with. Similarpoints apply to religious afiliation. Some- times this variable refers to officialmembership in 4 a particular church; other times it simplymeans whatever religion, if any, you identify yourself with. Perhaps to you it means something else, such as church attendance. The truth is that neither "party affiliation"nor "religious affiliation" has any real meaning, if by "real" we mean corresponding to some objective aspect of reality.These variables do not exist in na- ture. They are merely terms we have made up and assignedspecificmeanings to for some purpose, such as doing socialreseard~. But, you might object, "politicalaffiliation"and "religiousaffiliation"-and a host of other things socialresearchers are interested in, such as preju- dice or compassion-have sonze reality. After all, we make statementsabout them, such as "In Happy- town, 55 percent of the adults affiliate with the Re- publican Party, and 45 percent of them are Episco- palians. Overall, people in Happytown are low in prejudice and high in compassion."Even ordinary people, not just socialresearchers,have been known to make statementslike that. If these things don't exist in reality, what is it that we're measuring and talking about? Určeno pouze pro studijní účely 120 .Chapter 5: Conceptualization,Operationalization,and Measurerner What indeed? Let's take a closer loolcby con- sideringa variable of interest to many socialre- searchers (andmany other people aswell)- prejudice. Conceptions,Concepts, and Reality As you and I wandered down the road of life, we observed a lot of things and knew they were real through our observations, and we heard reports from other people that seemed real. For example: e We personallyheard people say nasty things about minority groups. s We heard people said women were inferior to men. s We read about African Americans being lynched. s We read that women and minorities earnedless for the samework. a We learned about "ethnic cleansing"and wars in which one ethnic group tried to eradicate another. With additional experience,we noticed some- thing more. People who participatedin lynching were also quite likely to call African Americans ugly names. A lot of them, moreover, seemedto want women to "stay in their place." Eventually it dawned on us that these severaltendencies often appeared together in the same people and also had somethingin common. At somepoint, someone had a bright idea: "Let's use the word prejudiced as a shorthand notation for people like that. We can use the term even if they don't do all those things-as long as they're pretty much like that." Being basically agreeable and interested in efficiency,we agreedto go alongwith the system. That's where "prejudice"came from. We never ob- served it. We just agreedto use it as a shortcut, a name that represents a collection of apparently re- latedphenomena that we've each observed in the course of life. In short, we made it up. Here's another clue that prejudice isn't some- thing that exists apart from our rough agreement to use the term in a certain way. Each of us devel- ops our own mental image of what the set of real phenomena we've observed represents in general and what these phenomena have in common. When 1say the word prejudice, it evokes a mental image in your mind, just as it evokes one in mine. It's as though file drawers in our minds contained thousands of sheets of paper, with each sheet of paper labeled in the upper right-hand corner. A sheet of paper in each of our minds has the term prejudice on it. On your sheet are allthe things you've been told about prejudice and everythmg you've observed that seemsto be an example of it. My sheet has what I've been told about it plus all the things I've observed that seem examplesof it- and mine isn't the same as yours. The technicalterm for those mental images, those sheets of paper in our mental file drawers,is co~zception.That is, I have a conception of prejudice, and so do you. We can't communicatethese mental images directly, so we use the terms written in the upper right-hand corner of our own mental sheets of paper as a way of communicatingabout our con- ceptionsand the things we observe that are related to those conceptions. These terms make it possible for us to communicateand eventuallyagree on what we will specificallymean by those terms. In socialresearch,the process of coming to an agree- ment about what terms mean is co~zceptunlizntion, and the result is called a co~zcept. Let's take another example of a conception. Supposethat I'm going to meet someone named Pat, whom you already know. I ask you what Pat is like.Now supposethat you have seen Pat help lost children find their parents and put a tiny bud back in its nest. Pat got you to take turkeys to poor fami- lies on Thanksgivingand to visit a children'shospi- tal on Christmas.You've seen Pat weep through a movie about a mother overcomingadversitiesto save and protect her child.As you search through your mental files, you may Lnd all or most of those phenomena recorded on a singlesheet labeled "compassionate."You loolc over the other entries on the page, and you find they seem to provide an accurate description of Pat. Soyou say, "Pat is compassionate." Now I leaf tlxough my own mental file drawer until I h d a sheet marked "compassionate."I then look over the things written on my sheet, and I say, "Oh,that's nice." I now feel I know what Pat is like, but my expectations reflect the entries on my file sheet,not yours.Later, when I meet Pat, 1happen to b d that my own experiencescorrespondto the entries I have on my "compassionate"fiIe sheet, and I say that you sure were right. But supposemy observationsof Pat contradictthe things I have on my file sheet. I tell you that I don't thinkPat is very compassionate,and we begin to comparenotes. You say, "I once saw Pat weep tlxough a movie about a mother overcoming adversity to save and protect her child." I look at my "compassionate sheet"and can't End anything like that. Looking elsewhere in my file, I locate that sort of phenome- non on a sheet labeled "sentimental."I retort, "That's not compassion.That's just sentimentality." To further strengthen my case, I tell you that I saw Pat refuse to givemoney to an organization dedicatedto saving whales from extinction. "That represents a lack of compassion,"I argue. You searchthrough your files and find savingthe whales on two sheets-"environmental activismr' and "cross-speciesdatingn-and you say so. Even- tually, we set about comparingthe entries we have on our respective sheets labeled "compassionate." We then discover that we have many differing mental images correspondingto that term. In the big picture, language and communica- tion work only to the extent that you and Ihave considerableoverlapin the lcinds of entrieswe have on our correspondingmental file sheets.The simi- laritieswe have on those sheets represent the agreementsexisting in our society.As we grow up, we're told approximatelythe same thing when we're first introduced to a particular term. Dictio- naries formalizethe agreements our societyhas about such terms. Each of us, then, shapeshis or her mental images to correspondwith such agree- ments. But because allof us have different experi- ences and observations,no two people end up with exactlythe same set of entries on any sheet in their file systems.If we want to measure "prejudice" or "compassion,"we must first stipulatewhat, exactly, counts as prejudice or compassionfor our purposes. Returningto the assertionmade at the outset of this chapter, we can measure anythmg that's real. Measuring AnythingThat Exists . 121 We can measure, for example, whether Pat actually puts the little bud back in its nest, visits the hospital on Christmas,weeps at the movie, or refuses to contributeto saving the whales.Allof those behav- iors exist, so we can measure them. But is Pat really compassionate?We can't answer that question;we can't measure compassionin any objectivesense, because compassiondoesn't exist the way those thingsIjust describedexist. Compassionexists only in the form of the agreementswe have about how to use the term in communicatingabout things that are real. ConceptsasConstructs If you recall the discussionsof postmodernism in Chapter2, you'll recognize that some people would object to the degree of "reality"I've allowed in the preceding comments.Did Pat "really"visit the hos- pital on Christmas?Does the hospital "really" ex- ist? Does Christmas?Though we aren't going to be radically postmodern in this chapter,I thinkyou'll d recognizethe importance of an intellectuallytough view of what's real and what's not. (Whenthe in- tellectualgoing gets tough, the tough become social scientists.) In this context,Abraham Icaplan (1964) distin- guishesthree classes of things that scientistsmea- sure. The first classis direct observables: those thingswe can observe rather simply and directly, like the color of an apple or the check mark made in a questionnaire.The secondclass, indirect ob- servable~,require "relativelymore subtle, complex, or indirect observations" (1964:55).We note a per- son's check mark beside "female"in a question- naire and have indirectly observed that person's gender. History boolcs or minutes of corporate board meetings provide indirect observations of past social actions.Finally, the third class of observ- ables consists of constructs-theoretical creations that are based on observationsbut that cannot be observed directly or indirectly.A good example is intelligencequotient, or IQ. It is constrzcted mathe- matically fromobservationsof the answers given to a large number of questions on an IQ test. No one can directly or indirectly observeIQ. It is no more a Určeno pouze pro studijní účely 122 . Chapter5: Conceptualization,Operationalization,and Measuremt "real" characteristicof people than is compassionor prejudice. Icaplan (1964:49) definesconceptas a "family of conceptions."A concept is, as Ibplan notes, a construct, somethingwe create. Conceptslike com- passion and prejudice are constructs createdfrom your conceptionof them, my conceptionof them, and the conceptions of allthose who have ever used these terms. They cannot be observed directly or indirectly,because they don't exist.We made them up. To summarize, concepts are constructs derived by mutual agreementfrom mental images (concep- tions).Our conceptions summarize collectionsof seeminglyrelated observationsand experiences.Al- though the observationsand experiencesare real, at least subjectively, conceptions,and the concepts derived from them, are only mental creations.The terms associatedwith concepts are merely devices created for the purposes of filing and communica- tion. A term like prejudice is, objectively speaking, only a collection of letters. It has no intrinsicreality beyond that. Is has only the meaningwe agree to give it. Usually, however, we fallinto the trap of be- lievingthat terms for constructs do have intrinsic meaning, that they name real entitiesin the world. That danger seemsto grow stronger when we be- gin to take terms seriously and attempt to use them precisely.Further, the danger is all the greaterin the presence of experts who appear to know more than we do about what the terms really mean:It's easy to yield to authority in such a situation. Once we assume that terms like prejudice and conzpassio~zhave real me-gs, we begin the tor- tured task of discoveringwhat those real mean- ings are and what constitutes a genuine measure- ment ofthem. Regarding constructs as real is called reification.The reification of concepts in day-to- day life is quite common.In science,we want to be quite clear about what it is we are actually measuring. Does this discussionimply that compassion, prejudice,and similarconstructs can't be measured? Interestingly,the answer is no. (Anda good thing, too, or a lot of us socialresearcher types would be out of work.) I've said that we can measure any- thingthat's real. Constructs aren't real in the way that trees are real, but they do have +other irnpor- tant virtue:They are useful. That is, they help us organize, communicate about, and understand things that are real. They help us make predictions aboutreal things. Some of those predictionseven turn out to be true. Constructs canwork this way because, while not real or observablein them- selves, they have a definite relationshipto things that are real and observable.The bridgs from direct and indirect observablesto useful constructsis the process called conceptualization. Conceptualization As we've seen, day-to-daycommunicationusually occursthrough a system of vague and general agreementsaboutthe use of terms. Although you and I do not agree completely about the use of the term conzpassio~zate,I'm probably safe in assuming that Pat won't pull the wings off flies.A wide range of misunderstandingsand conflict-from the in- terpersonal to the international-is the price we pay forour imprecision, but somehowwe muddle through. Science,however, aims at more than muddling; it cannot operate in a context of such imprecision. The process through which we specify what we mean when we use particular terms in research is called conceptualization.Supposewe want to find out, for example, whether women are more compassionatethan men.I suspect many people assumethis is the case, but it mightbe interesting to find out if it's really so.We can't meaningfully study the question, let alone agree on the answer, without someworking agreements about the meaning of conzpassiolz.They are "worldng" agree- ments in the sense that they allow us to work on the question. We don't need to agree or even pre- tend to agreethat a particular specificationis ulti- 'mately the best one. Conceptualization,then, produces a specific, agreed-uponmeaning for a conceptfor the pur- poses of research. This process of specifyingexact meaninginvolvesdescribingthe indicatorswe'll be using to measure our concept and the different as- pects of the concept, called dimensions. IndicatorsandDimensions Conceptualizationgives definite meaning to a con- cept by speclfylngone or more indicatorsof what we have in mind. An indicatoris a sign of the presence or absence of the concept we're studying. Here's an example. We might agree that visiting children's hospitals during Christmas and Hanukkah is an indicator of compassion.Putting little birds back in their nests might be agreed on as another indicator, and so forth.If the unit of analysisfor our studyis the in- dividualperson, we can then observethe presence or absence of each indicatorfor each person under study. Goingbeyond that, we can add up the num- ber of indicators of compassion observedfor each individual. We might agree on ten specific indica- tors, for example, and find six present in our study of Pat, three for John, nine for Mary, and so forth. Returning to our question about whether men or women are more compassionate, we might cal- culate that the women we studied displayed an average of 6.5 indicatorsof compassion,the men an average of 3.2. On the basis of our quantitative analysis of group difference,we might therefore conclude that women are, on the whole, more com- passionate than men. Usudy, though, it's not that simple.Imagine you're interested in understandinga smallfunda- mentalistreligious cult,particularlytheir harsh views on various groups: gays, nonbelievers,femi- nists, and others.In fact, they suggestthat anyone who refusesto join their group and abide by its teachings will "burn in hell." Inthe context of your interest in compassion,they don't seem to have much. And yet, the group's literature often speaks of their compassionfor others.You want to explore this seemingparadox. To pursue this research interest, you might arrange to interactwith cult members, getting to lcnow them and learningmore about their views. You could tell them you were a socialresearcher interestedin learningabout their group, or perhaps Conceptualization .123 you wouldjust express an interestin learningmore without saying why. In the course of your conversations with group members and perhaps attendanceof religious ser- vices, you would put yourself in situationswhere you could come to understand what the cult mem- bers mean by conzpassioiz.You might learn, for ex- ample,that members of the group were so deeply concerned about sinners burning in hell that they were willing to be aggressive, even violent, to make people changetheir sinful ways. Within their own paradigm, then, cultmembers would see beating up gays, prostitutes, and abortion doctors as acts of compassion. Socialresearchersfocustheir attention on the meanings given to words and actions by the people under study. Doing so can often clanfythe behav- iors observed:At least now you understand how the cult can see violent acts as compassionate. On the other hand, paying attention to what words and actionsmean to the people under study almost always complicatedthe concepts researchers are in- t rested in. (We'll return to this issue when we dis- :cuss the validity of measures, toward the end of this chapter.) Whenever we take our concepts seriously and set about specifyingwhat we mean by them, we discover disagreements and inconsistencies. Not only do you and I disagree,but each of us is likely to find a good deal of muddiness within our own mental images.If you take a moment to look at what you mean by compassion,you'll probably find that your image contains severalkinds of com- passion. That is, the entries on your file sheet can be combined into groups and subgroups, say, com- passion toward friends,co-religionists,humans, and birds. You man also-hd several differentstrate- gies for making combinations. For example,you might group the entriesinto feelingsand actions. The technicalterm for such groupingsis di- mension:a specsable aspect of a concept.For in- stance, we might speak of the feeling dimension of compassionand the action dimensionof compas- sion. In a differentgroupingscheme,we might dis- tinguish compassion for humans from compassion for animals.Or we might see compassion as help- ing people have what we want for them versus Určeno pouze pro studijní účely 124 .Chapter 5: Conceptualization,Operationalization,and Measuremen what they want for themselves. Stilldifferently, we might distinguishcompassion as forgivenessfrom compassionaspity. Thus, we could subdivide compassioninto sev- eral clearly defined dimensions.A complete con- ceptualizationinvolvesboth specifyingdimensions and idenhfyingthe variousindicatorsfor each. Speclfylngthe different dimensionsof a con- cept oftenpaves the way for a more sophisticated understandingof what we're studying.We might observe, for example,that women are more com- passionate in terms of feelings, and men more so in terms of actions-or vice versa. Whichever turned out to be the case, we would not be able to say whether men or women are really more compas- sionate.Our research would have shownthat there is no singleanswerto the question. That alone represents an advancein our understandingof reality. TheInterchangeability ofIndicators There is another way that the notion of indicators can help us in our attempts to understand reality by means of "unreal" constructs. Suppose,for the moment, that you and I have compiled a list of 100indicatorsof compassion and its various di- mensions. Suppose further that we disagreewidely on which indicatorsgive the clearest evidence of compassionor its absence.If we pretty much agree on some indicators, we could focus our attention on those, and we would probably agree on the an- swer they provided. We would then be ableto say that some people are more compassionate than others in some dimension.But supposewe don't really agree on any of the possible indicators. Surprisingly,we can stillreach an agreement on whether men or women are the more compassion- ate.How we do that has to dowith the interchange- ability of indicators. The logicworks like this. If we disagreetotally on the value of the indicators, one solutionwould be to study all of them. Supposethat women turn out to be more compassionate than men on all 100indicators-on allthe indicatorsyou favorand on all of mine. Then we would be able to agree that women are more compassionate than men even though we stilldisagreeon exactlywhat compas- sionmeans in general. The interchangeabiityof indicatorsmeans that if severaldifferent indicators all represent, to some degree, the same concept, then all of them will be- have the same way that the concept wouldbehave if it were real and could be observed. Thus, given a basic agreement aboutwhat "compassion"is, if women are generallymore compassionatethan men, we shouldbe ableto observe that difference by using any reasonable measure of compassion.If, on the other hand, women are more compassion- ate than men on some indicatorsbut not on others, we shouldsee if the two sets of indicators represent different dimensions of compassion. You have now seen the fundamentallogic of conceptualization and measurement.The discus- sions that follow are mainly refinements and ex- tensions of what you've just read. Before turning to a technicalelaborationof measurement,however, we need to iill out the picture of conceptualization by looking at some of the ways socialresearchers provide the meanings of terms with standards, con- sistency,and commonality. Real,Nominal,andOperationalDefinitions As we have seen, the design and execution of social research requires us to clear away the confusion over concepts and reality.To this end, logiciansand scientistshave found it useful to distinguishthree kinds of deilnitions:real, nominal, and operational. The firstof these reflects the reification of terms. As CarlHempel has cautioned, A "real" definition, accordingto traditional logic, is not a stipulationdetermining the meaning of some expressionbut a statement of the "essentialnature" or the "essentialattri- butes" of some entity. The notion of essential nature, however,is so vague asto render this characterizationuseless for the purposes of rig- orous inquiry. (1952 :6) In other words, hying to spec@ the "real" meaning of concepts only leads to a quagmire:It mistakesa constructfor a red entity. The specificationof concepts in scientificin- quiry dependsinstead on nominal and operational dehitions. A ivzoivniivznldejnitionis one that is simply assigned to a term without any claimthat the defi- nition represents a "real" entity. Nominal definitions are arbitrary-I could define conzpassio~~as "pluck- ing feathersoff helpless birds" if I wanted to-but they can be more or less useful. For most purposes, especially communication,that last definition of compassion would be pretty useless. Most nominal definitions representsome consensus,or conven- tion, about how a particular term is to be used. An operationaI defivzition,as you may remember from an earlier chapter, specifiesprecisely how a conceptwillbe measured-that is, the operations we WIII perform. An operational defbition is nomi- nal rather than real, but it has the advantage of achievingmaximum clarityabout what a concept means in the context of a given study. In the midst of disagreement and confusion over what a term "really"means, we can specify a working definition for the purposes of an inquiry. Wishing to examine socioeconomicstatus (SES)in a study, for example, we may simply specifythat we are going to treat SES as a combination of income and educational attainment.In this decision, we rule out other pos- sible aspects of SES:occupationalstatus, money in the b d c , property, lineage, lifestyle, and so forth. Our iindingswillthen be interesting to the extent that our definitionof SES is useful for our purpose. CreatingConceptualOrder The clarScationof conceptsis a continuingpro- cess in social research. Catherine Marshalland Gretchen Rossman (1995:18)speak of a "concep- tual funnel" through which a researcher's interest becomes inaeasingly focused. Thus, a generalin- terest in social activism could narrow to "individu- als who are committedto empowermentand social change" and further focuson discovering"what experiences shaped the development of fully com- mitted socialactivists." This focusingprocess is in- escapablylinked to the languagewe use. In someformsof qualitative research,the clari- ficationof conceptsis a lcey elementinthe collection ofdata. Supposeyou were conductinginterviews Conceptualization .125 and observationsin a radicalpoliticalgroup de- voted to combatingoppressionin U.S. society. Imaginehow the meaning of oppression would shift as you delved more and more deeply into the members' experiences and worldviews.For ex- ample, you might start out thinking of oppression in physicalandperhaps economicterms. The more you learned about the group, however, the more you might appreciate the possibility of psychologi- cal oppression. The samepoint applieseven to contexts where meanings might seem more flxed. In the analysis of textualmaterials, for example, social researchers sometimesspeak of the "hermeneuticcircle," a cyclicalprocess of ever deeper understanding. The understanding of a text takes place through a process in which the meaning of the separate parts is determinedby the global meaning of the text as it is anticipated.The closer determinationof the meaning of the separateparts may eventually change the origi- nalIy anticipatedmeaning of the totality, which 4 again influencesthe meaning of the separate ' parts, and so on. Considerthe concept "prejudice." Supposeyou needed to write a definitionof the term. You might start out thinking about raciallethnicprejudice. At somepoint you would realize you shouldprobably allow for genderprejudice, religious prejudice, anti- gay prejudice, and the like in your deiinition. Ex- aminingeach of these specifictypes of prejudice would affectyour overall understandingof the general concept.As your general understanding changed, however, you would likely see each of the individual formssomewhat differently. The continualrefinement of concepts occurs in all socialresearchmethods. Often you will h d yourself refining the meaning of important con- cepts even as you write up your final report. Although conceptualization is a continuing process, it is vital to addressit specificallyat the be- ginning of any study design, especially rigorously structu~edresearch designs such as surveysand experiments.In a survey, forexampIe, operational- ization results in a commitmentto a specific set of questionnaireitems that will representthe concepts Určeno pouze pro studijní účely 126 .Chapter5: Conceptualization,Operationalization,andMeasuremer under study. Without that commitment,the study could not proceed further. Even in less-structuredresearchmethods, how- ever, it's important to beginwith an initial set of anticipated meanings that can be refined during data collection and interpretation. No one seriously believeswe can observe life with no preconcep- tions; forthis reason, scientific observersmust be conscious of and explicit about these conceptual startingpoints. Let's exploreinitial conceptualizationthe way it applies10structured inquiries such as surveys and experiments.Though specifying nominal defini- tions focuses our observationalstrategy,it does not allow us to observe. As a next stepwe must spec* exactlywhat we are going to observe, how we will do it, and what interpretationswe are going to place on various possible observations.AU these further specificationsmake up the operational definition of the concept. In the example of socioeconomicstatus, we might decide to aslc survey respondentstwo ques- tions, corresponding to the decision to measure SES in terms of income and educationalattainment: 1. What was your total familyincome during the past 12 months? 2. What is the highest level of schoolyou completed? To organize our data, we would probably want to spenfy a systemfor categorizing the answers people give us. For income,we might use categories such as "under $5,000," "$5,000to $10,000,"and so on. Educational attainment might be similarly grouped in categories:less than high school,high school,college, graduate degree. Finally, we would specify the way a person's responses to these two questionswould be combinedin creating a mea- sure of SES. In this way we would create a worldngand worlcable definitionof SES.Although others might disagree with our conceptualizationand opera- tionalization,the definitionwould have one essen- tial scientificvirtue: It would be absolutelyspecific and unambiguous. Even if someone disagreed with our definition, that person would have a good idea how to interpret our research results,because what we meant by SES-reflected in our analysesand conclusions-would be precise and clear. Here is a diagram showingthe progression of measurement stepsfrom our vague sense of what a term means to specificmeasurementsin a fully structured scientijicstudy: Conceptualization .I. Nominal Definition 1 OperationalDefinition 1 Measurementsin the Red World An ExampleofConceptualization: TheConceptofAnomie To bring this discussion of conceptualizationin research together, let's look briefly at the history of a specific socialscientijicconcept. Researchers studyingurban riots are often interested in the part played by feelingsof powerlessness. Socialscientists sometimesuse the word anonzie in this context.This term was First introducedinto social scienceby Ernile Durkheim, the great French sociologist,in his classic 1897study, Stiicide. Using only governmentpublications on suicide rates in differentregions and countries, Durlrheim produced a work of analyticalgenius. To determine the effectsof religion on suicide, he comparedthe suicide rates ofpredominantlyProtestant countries with those of predominantly Catholicones, Protes- tant regions of Catholic countrieswith Catholic regions of Protestantcountries, and so forth. To determine the possible effects of the weather, he compared suicide rates in northern and southern countriesand regions, and he examined the differ- ent suicide rates acrossthe months and seasons of the year. Thus, he could draw conclusionsabout a supremelyindividualistic and personal act without having any data about the individualsengagingin it. At a more general level, Durkheim suggested that suicide also reflects the extent to which a soci- ety's agreements are clear and stable.Noting that times of socialupheaval and change often present individuals with grave uncertaintiesabout what is expected of them, Durkheim suggested that such uncertaintiescause confusion, anxiety, and even self-destruction.To describe this societalcondition of norrnlessness,Durkheim chose the term anomie. Durkheim did not make this word up. Used inboth German and French, it literallymeant "without law." The English term aizoiny had been used for at least three centuriesbeforeDurlheirn to mean dis- regard for divinelaw. However, Durkheim created the social scientific concept of anomie. In the years that have followed the publication of Suicide, social scientists have found anomie a use- ful concept, and many have expanded on Durlc- heim's use. Robert Merton, in a classicarticle entitled "Social Structure and Anomie" (1938), concludedthat anomie results from a disparitybe- tween the goals and means prescribedby a society. Monetary success, for example, is a widely shared goal in our society, yet not all individualshave the resourcesto achieveit through acceptable means. An emphasison the goal itself,Merton suggested, produces norrnlessness,because those denied the traditional avenues to wealth go about getting it through illegitimatemeans. Merton's discussion, then, couldbe considereda further conceptualiza- tion of the concept of anomie. Although Durlrheim originallyused the con- cept of anomie as a characteristicof societies, as did Merton after him,other social scientistshave used it to describe individuals.To clarify this distinction, some scholars have chosen to use anomie in refer- ence to its original, societalmeaning and to use the term anomia in reference to the individual charac- teristic. In a given society, then, some individuals experienceanomia, and others do not. Elwin Pow- ell,writing 20 years after Merton, provided the fol- lowing conceptualizationof anomia (thoughusing the term arzornie)as a characteristicof individuals: When the ends of actionbecome contradictory, inaccessibleor insignificant, a condition of anomie arises. Characterized by a generalloss of orientationand accompaniedby feelings of "emptiness" and apathy, anomie can be simply conceived as meaninglessness. (1957:132) Conceptualization .127 Powell went on to suggestthere were two dis- tinct kinds of anomia and to examine how the two rose out of different occupationalexperiences to result at times insuicide.In his study, how- ever, Powell did not measure anomiaper se; he studied the relationshipbetween suicideand oc- cupation,maldng inferencesabout the two ldnds of anomia. Thus, the study did not provide an operational deiinitionof anomia, only a further conceptualization. Although many researchers have offered oper- ational definitionsof anomia, one name stands out over all. Two years beforePowell's article appeared, Leo Srole (1956)published a set of questionnaire items that he said provided a good measure of anomia as experienced by individuals.It consists of five statementsthat subjectswere asked to agree or disagree with: 1. In spite of what some people say, the lot of the average man is getting worse. 2. It's hardly fair to bring childreninto the world with the way things loolcfor the future. s' 3 , Nowadays a person has to live pretty much for today and let tomorrow talce care of itself. 4. These days a person doesn't really know who he can count on. 5. There's little use writing to public officialsbe- cause they aren't really interestedinthe prob- lems of the average man. (1956: 713) In the decades followingits publication, the Srole scale has become a research staple for social scientists.You'll likely h d this particular opera- tionalizationof anomia used in many of the re- searchprojects reported in academicjournals. Srole touches on this in the accompanyingbox, "The Origins of Anomia,"which he prepared for this boolcbefore his death. This abbreviatedhistory of anomie and anomia as socialscientific concepts illustrates severalpoints. First, it is a good example of the process through which general concepts become operationalized measurements.This is not to say that the issue of how to operationalizeanomielanomiahas been re- solvedonce and for all. Scholarswill surelycontinue Určeno pouze pro studijní účely 128 .Chapter5: Conceptualization,Operationalization,and Measurement Definitions in Descriptiveand ExplanatoryStudies .129 the other hand,Iwas moved by Durkheim's bonds.We neededto work expeditiously, its macro-socialmeaningandto sharplyseg- unswerving preoccupationwith the moral Centerfor Geriatricsand Gerontology, deem~hasizingproliferation of macro-level regate itfrom its individual manifestations. ColumbiaUniversity force of the interpersonalties that bind us to theory in favor of adirect exploratory Forthe latter purpose,the cognate but hith- our time, place, and past,and also his insights encounterwith individuals,using newly erto obsoleteGreekterm, anomie, readily aboutthe lethalconsequencesthat canfol- developedstate-of-the-art survey research suggesteditself. Mycareer-longfixation on anomie began low from shrinkage and decay in thoseties. methodology.Such research,Ialso felt, Ifirst publishedthe anomieconstructin with readingDurkheim'sLeSuicideas a Har- MYinterestin anomie receivedan eyewit- shouldfocus on a broaderspectrumof be- a 1956articlein theAmericanSocio/ogica/ vard undergraduate.Later,as agraduate nessjolt at the finale of WorldWar II,when I havioral pathologiesthan suicide. Review* describingways of operationalizing student at Chicago, Istudied undertwo servedwith the United Nations Reliefand Re- My initial investigation~were a diverse itland presentingthe resultsof its initial field Durkheimiananthropologists:WilliamLloyd habilitation~dministration,helpingto re- effort-In1950tforexample, Iwas able to in- . applicationresearch.By 1982,the Science warner and Alfred Radcliffe-Brown.Radcliffe- build awar-torn Europe.At the Naziconcen- terview asampleof 401 bus riders in Spring- Citation Indexand socialScience Citation Brown had carried on alively correspon- tration camp of Dachau, Isaw first-handthe field, Mass.Fouryears later,the Midtown Index had listed some400 publicationsin po- dent- with Durkheim,making meacollateral depths of dehumanizationthat macro-social Mi3~hattanMental HealthStudy provided a litical science, psychology, social work, and "desCendant"ofthegreat Frenchsociologist. forces, such as thosethat engaged Durkheimf much larger population reach.These and Sociologyjournals here and abroad that had For mettheearly impact of Durkheim's could producein individualslike Hitler, Eich- field projects gave mescopeto expand cited useof that article's instrumentsor find- work on suicidewas mixedbut permanent. mann,and the others servingtheir dictates at and refinemy ~~leasurementsof that quality ings,warranting the American Institutefor onthe one hand, 1hadserious reservations all levels inthe Nazideathfactories. in individualswhich reflectedthe macro- Scientific Informationto designateit al'cita- about hisstrenuous, ingenious,and often Returningfrom my UNRRA post, Ifelt social quality Durkheimhad called anomie. tion classic." awkwardeffortsto force the crude, bureau- most urgentlythat the time was long Over- While Ibegan by using Durkheim's term cratic recordson suicide ratesto fit with his due to cometo an understandingof the C in my Own I decided that it was f Leo SrCIl~,%~ialbtegrationandceMin &PIoF unidirectionalsociologicaldeterminism.On dynamics underlyingdisintegrated Social necessary limitthe use of that conceptto dtory~tudy,"~rnericansociological Review21 (1956):709-16. to reconceptualize and reoperationalizethese con- cepts for years to come, continuallyseelhgmore useful measures. The Srole scale illustratesanother important point. Letting conceptualizationand operationaliza- tion be open-ended does not necessarilyproduce anarchy and chaos, as you might expect. Order often emerges. For one thing, although we could define anomia any way we chose-in terms of, say, shoe size-we're likely to defbe it in ways not too different from other people's mental images.If you were to use a really offbeat definition, people would probably ignore you. A second source of order is that, as researchers discover the utility of a particular conceptualiza- tion and operationalizationof a concept, they're likely to adopt it, which leads to standardized definitions of concepts. Besides the Srole scale, examples includeIQ tests and a host of demo- graphic and economicmeasures developed by the U.S. CensusBureau. Using such established meas- ures has two advantages:They have been exten- sivelypretested and debugged, and studies using the same scales can be compared.If you and I do separate studies of two different groups and use the Srole scale,we can compare our two groups on the basis of anomia. Socialscirntists, then, can measure anything that's real; through conceptualization and opera- tionkation, they can even do a pretty goodjob of measuringthings that aren't. Grantingthat such concepts as socioeconomicstatus,prejudice, com- passion, and anomia aren't ultimatelyreal, social scientists can create order in handling them. It is an orderbased on utility, however, not on ultimate truth. DefinitionsinDescriptive and ExplanatoryStudies As you recall from Chapter 4, two general purposes of research are description and explanation.The distinctionbetween them has important implica- tions for definitionand measurement. If it seems that descriptionis a simplertask than is explana- tion, you may be surprisedto learn that definitions are more problematic for descriptiveresearch than for explanatoryresearch.Beforewe turn to other aspects of measurement, you'll need a basic under- standingof why this is so (we'll discussthis point more fullyin Part 4). It's easy to see the importance of clear and pre- cise definition~for descriptiveresearch.If we want to describe and report the unemployment rate in a city, our definition of being unemployed is obvi- ously critical.That definition will depend on our definition of another term: the labor force. If it seemspatently absurdto regard a three-year-old child as being unemployed,it is because such a childis not considered a member of the labor force. Thus, we might followthe U.S. Census Bureau's convention and exclude all people under 14years of age from the Iabor force. This conventionalone, however, would not give us a satisfactorydefinition,because it wouId count as unemployed such people as high school students,the retired, the disabled, and homemak- ers. We might follow the census convention further by de&g the labor force as "allpersons 14years of age and over who are employed, looldng for work, or waiting to be called back to ajob from which they have been laid off or furlo~ghed.~If Určeno pouze pro studijní účely 130 .Chapter5: Conceptualization,Operationalization,and Measurement Definitionsin Descriptiveand ExplanatoryStudies .131 vwvbyPatriciaFirher GraduateSchoolofPlanning, UniversityofTennessee operationalization is one of those things that's easier saidthan done.It is quite simple to explainto someonethe purpose and im- portanceof operationaldefinitions for vari- ables, and evento describe how operational- izationtypically takes place.However,until you've tried to operationalizea rather com- plexvariable,you may not appreciatesome of the subtle difficultiesinvolved.Of consid- erable importanceto the operationalization effort is the particularnamethat you have chosenfor avariable.Let's consideran ex- amplefrom the field of UrbanPlanning. A variable of interestto planners is citizen participation. Plannersare convincedthat participation inthe planning processby citi- zens is importantto the success of planim- plementation.Citizen participation is an aid to planners'understanding of the realand perceivedneedsof acommunity,and such involvementby citizens tendsto enhance their cooperationwith and supportfor plan- ning efforts.Although many different con- ceptualdefinitions might be offered by different plannersthere would be little mis- understandingover what is meant by citizen participation. The nameof the variableseems adequate. However,if we askeddifferent plannersto providevery simpleoperational measuresfor citizenparticipation,weare likelytofind ava- rietyamongtheir responsesthat doesgener- ateconfusion.Oneplannermight keepatally of attendanceby privatecitizens at city com- missionand other localgovernmentmeet- a student, homemaker, or retired person is not looking for work, such a person would not be in- cluded inthe labor force.Unemployed people, then, would be those members of the labor force, as de- fined, who are not employed. But what does "looldngfor work" mean? Must a person register with the state employment service or go from door to door asking for employment? Or would it be sufficient to want ajob or be open to an offer of employment? Conventionally,"look- ing for work" is defined operationallyas saying yes in response to an interviewer's asking "Haveyou been looking for ajob during the past seven days?" (Sevendaysis the period most often specified,but for some research purposes it might malce more . sense to shorten or lengthen it.) As you can see, the conclusion of a descriptive study about the unemployment rate depends di- rectly on how each issue of definitionis resolved. Increasingthe period during which people are counted as looking for work would add more un- employedpeople to the labor force as defined, thereby increasing the reported unemployment rate. If we follow another convention and speak of the civilian labor force and the civilian unemploy- ment rate, we are excluding military personnel; that, too, increasesthe reported unemployment rate, because military personnel would be em- ployed-by definition. Thusthe descriptivestate- ment that the unemployment rate in a city is 3 per- cent, or 9 percent, or whatever it might be, depends directly on the operational definitions used. This exampleis relatively clearbecause there are severalaccepted conventionsrelating to the la- bor force and unemployment. Now, consider how d3i3cultit would be to get agreement about the definitions you would need in order to say, "Forty- five percent of the students at this institution are ings;another might maintainarecordofthe differenttopicsaddressed by privatecitizens at similar meetings;while athird mightrecord the numberof localgovernment meeting attendees, letters, and phonecalls received bythe mayorand other publicofficials,and meetingsheld byspecialinterestgroups during a particulartime period.8~skilled re- searchers,we can readilyseethat each plan- nerwould be measuring (inavery simplistic fashion)adifferent dimension of citizen par- ticipation:extent ofcitizen participation,is- sues promptingcitizenparticipation,and form of citizen participation.Therefore,the originalnaming of our variable,citizenpartici- pation, which was quite satisfactoryfrom a conceptualpoint of view, provedinadequate for purposesofoperationalization. The preciseand exact naming of variables is important in research.It is both essentialto and aresultof good operationalization.Vari- able names quite often evolvefrom an itera- tive processof forming aconceptual defini- tion, then an operational definition,then renamingthe conceptto better matchwhat can or will be measured.This looping process continues(ourexampleaboveillustratesonly one iteration),resulting in agradual refine- ment of the variable nameand its measure- ment until a reasonablefit is obtained.Some- times the concept of the variable that you end up with is a bit different from the original one that you started with, but at leastyou are measuringwhat you aretalking about,if only becauseyou aretalking about what you are measuring! politically conservative."Lilce the unemployment rate, this percentagewould depend directly on the definition of what is being measured-in this case, political conservatism.A different definitionmight result in the conclusion "Five percent of the stu- dent body are politicallyconservative." Ironically, definitions are less problematicin the case of explanatoryresearch.Let's suppose we're interested in explainingpolitical conser- vatism. Why are somepeople conservativeand othersnot? More speciiically,let's supposewe're interestedin whether conservatismincreases with age.What if you and I have 25 different operational definitionsof consewative,and we can't agree on which definitionis best? As we saw in the discus- sion of indicators, this is not necessarily an insur- mountable obstacleto our research. Supposewe found old people to be more conservativethan young people in terms of all 25 definitions. Clearly, the exact definition would be of small conse- quence. Supposewe found old people to be more conservativethan young people by every reason- able definition of conservatismwe could think of. It wouldn't matter what our definitionwas. We would conclude that old people are generallymore conservativethan young people-even though we couldn't agree about exactlywhat coizservative means. In practice, explanatoryresearch seldomre- sultsin findings quite as unambiguous as this ex- ample suggests;nonetheless, the general pattern is quite common in actual research. There are con- sistentpatterns of relationships in human social life that result in consistent research findings. How- ever, such consistencydoes not appear in a de- scriptive situation. Changing definitions almost inevitably result in different descriptiveconclu- sions. The box "TheImportance of Variable Names" Určeno pouze pro studijní účely 132 .Chapter 5: Conceptualization,Operationalization,andMeasurer Operationalization Choices . 133 exploresthis issue in connectionwith the variable citizelz participatiolz. Operationalization Choices In discussingconceptualization,I frequentlyhave referred to operationalization,for the two are inti- mately linked.To recap: Conceptualizationis the rehement and spe~cationof abstract concepts, and operationalizationis the developmentof spe- cificresearchprocedures (operations)that wdl re- sultin empiricalobservationsrepresentingthose conceptsin the real world. As with the methods of data collection, social researchers have a variety of choiceswhen opera- tionalizing a concept. Although the severalchoices are intimately interconnected, I've separated them for the sake of discussion. Realize,though,that op- erationalizationdoes not proceed through a sys- tematic checklist. RangeofVariation In operationdizingany concept, researchers must be clear about the range of variation that interests them. The question is, to what extent are we will- ing to combine attributesin fairly gross categories? Let's supposeyou want to measure people's incomesin a studyby collectingthe information from either records or interviews.The highest an- nual incomespeople receive run into the millons of dollars,but not many people get that much. Un- less you're studyingthe very rich, it probably won't add much to your study to keep track of extremely high categories.Depending on whom you study, you'll probably want to establisha highest in- come category with a much lower floor-maybe $100,000or more. Although this decisionWLU lead you to throw together people who earn a tril- lion dollars a year with paupers earning a mere $100,000,they'll surviveit, and that mixingproba- bly won't hurt your research any, either.The same decisionfaces you at the other end of the income spectrum.In studies of the generalU.S. popula- tion, a bottom categoryof $5,000or less usually works fine. Instudies of attitudesand orientations, the question of range of variationhas another dimen- sion. Unless you're careful, you may end up meas- uring only half an attitude without really meaning to. Here's an example of what I mean. Suppose you're interestedin people's attitudes toward expanding the use of nuclearpower gener- ators.You'd anticipatethat somepeople consider nuclear power the greatestthing sincethe wheel, whereas other people have absolutelyno interest in it. Given that anticipation,it would seem to make senseto aslcpeople how much they favor expand- ing the use of nuclear energy and to give them an- swer categoriesrangingfrom "Favorit very much" to "Don't favorit at d." This operationalization,however, conceals half the attitudinalspectrum regarding nuclear energy. Many people have feelingsthat go beyond simply not favoringit:They are, with greater or lesser de- grees ofintensity, activelyopposedto it. In this in- stance, there is considerable variation on the left side of zero. Some opposeit a little, some quite a bit, and others a great deal. To measure the full range of variation, then, you'd want to operational- ize attitudes toward nuclear energywith a range fromfavoringit very much, through no feelings one way or the other, to opposingit very much. This considerationappliesto many of the vari- ables social scientists study.Virtually any public is- sue involvesboth supportand opposition,each in varying degrees.Politicalorientations range from very liberal to very conservative, and dependingon the people you're studying, you may want to allow for radicalson one or both ends. Similarly, people are not just more or less religious; some are posi- tively antireligious. Thepoint is not that you must measure the full range of variation in every case. You should,how- ever, consider whether you need to, givenyour particular research purpose. If the difference be- tween not religious and antireligiousisn't relevant to your research, forgetit. Someone has defined pragmatism as "any differencethat makes no differ- ence is no difference."Be pragmatic. Finally, decisionson the range of variation shouldhe governedby the expected distribution of attributesamong the subjects of the study.In a studyof college professors' attitudestoward the value of higher education, you could probably stop at no value and not worry about those who might consider higher educationdangerousto students' health. (If you were studying students, however. ..) Variationsbetween theExtremes Degree of precision is a second considerationin op- erationalizingvariables. What it boils down to is how fineyou will make distinctionsamong the various possible attributescomposing a given vari- able. Does it matter for your purposes whether a person is 17or 18years old, or could you conduct your inquiryby throwing them together in a group labeled 10to 19years old? Don't answer too quickly.If you wanted to studyrates of voter regis- tration and participation, you'd definitely want to know whether the people you studiedwere old enough to vote.In general, if you're going to mea- sure age, you must look at the purpose and proce- dures of your study and decide whether fine or gross differencesin age are important to you. In a survey,you'll need to make these decisions in order to design an appropriatequestionnaire.In the case of in-depthinterviews, these decisionswill condi- tion the extent to which you probe for details. The samething appliesto other variables. If you measurepolitical &ation, willit matter to your inquirywhether a person is a conservativeDemo- crat rather than a liberalDemocrat, or will it be sufficientto know the party? In measuring reli- gious affiliation, is it enough to know that a person is a Protestant, or do you need to know the denom- ination? Do you simply need to know whether or not a person is married, or will it make a difference to know if he or she has never married or is sepa- rated, widowed, or divorced? There is, of course, no generalanswer to such questions.The answers come out of the purpose of a given study or why we are making a particular measurement.I can give you a useful guideline, though. Whenever you're not sure how much detail to pursue in a measuement, get too much rather than too little. When a subjectin an in-depth interview volunteersthat she is 37 years old, record "37"in your notes, not "inher thirties." When you're analyzing the data, you can always combine precise attributesinto more generalcategories,but you can never separate any variationsyou lumped together during observationand measurement. A Note onDimensions We've already discussed dimensionsas a character- istic of concepts.When researchers get down to the business of creatingoperationalmeasures of vari- ables, they oftendiscover-or worse, never no- tice-that they're not exactly clear about which di- mensions of a variable they're really interestedin. Here's an example. Let's supposeyou're studyingpeople's attitudes toward government, and you want to includean examination of how people feel about corruption. Here arejust a few of the dimensionsyou might examine: e Do people thinkthere is corruptionin government? 4 How much corruptiondo they think there is? o How certain are they in theirjudgment of how much corruption there is? a How do they feel about corruption in govern- ment as a problem in society? e What do they thinkcauses it? e Do they think it's inevitable? e What do they feel should be done about it? e What are they willing to do personally to elirni- nate corruptionin government? o How certain are they that they wouldbe will- ing to dowhat they say they would do? The list could go on and on-how people feel about corruption in governmenthas many dimen- sions.It's essential to be clear about which ones are important in our inquiry; otherwise, you may mea- sure how peoplefeel about corruptionwhen you really wanted to know how much they think there is, or vice versa. Once you have determinedhow you're going to collectyour data (forexample, survey, field re- search)and have decided on the relevant range of variation,the degree of precisionneeded between Určeno pouze pro studijní účely 134 .Chapter 5: Conceptualization,Operationalization,andMeasurein1 the extremes of variation, and the specific dimen- sions of the variables that interestyou, you may have another choice: a mathematical-logicalone. That is, you may need to decide what level of meas- urement to use. To discuss this point, we need to take another look at attributesand their relation- shipto variables. Defining IlariablaandAffributer An attribute, you'll recall, is a characteristicor qual- ity of something. "Female"is an example. Sois "old or "student." Variables, on the other hand, are logical sets of attributes. Thus, gender is a vari- able composed of the attributesfemale and male. The conceptualizationand operationalization processes canbe seen as the spec%cation of vari- ables and the attributescomposingthem. Thus, in the context of a study of unemployment,employ- ment stattls is a variable having the attributesem- ployed and unemployed; the list of attributescould alsobe expanded to include the other possibilities discussed earlier,such as homemaker. Every variable must have two important quali- ties. First, the attributescomposingit shouldbe exhaustive.For the variable to have any utility in research, we must be able to class* every observa- tion in terms of one of the attributescomposing the variable. We'll run into trouble if we conceptualize the variablepolitical party afiliation in terms of the attributesRepublican and Democrat, because some of the people we set out to studywill identifywith the Green Party, the Reform Party, or some other organization, and some (oftena largepercentage) willtell us they have no party affiliation.We could make the list of attributesexhaustiveby adding "other"and "no fiation." Whatever we do, we must be able to classify every observation. At the same time, attributescomposing a vari- able must be mutually exclusive.Every observation must be able to be classifiedin terms of one and only one attribute.For example,we need to define "employed"and "unemployed in such a way that nobody can be both at the same time. That means being able to class@ the person who is working at ajob but is also lookingfor work. (Wemight run across a fully employed mud wrestler who is look- tnt ing forthe glamour and excitement of being a so- cialresearcher.)In this case, we might d e k e the attributes so that employed takesprecedence over unemployed, and anyone working at a job is em- ployed regardless of whether he or sheis looking for somethingbetter. LevelsofMeasurement Attributes operationalized as mutually exclusive and exhaustivemay be related in other ways as well. For example, the attributes composingvari- ables may representdifferent levels of measure- ment. In this section, we'll examine four levels of measurement:nominal, ordinal,interval, and ratio. NominalMeasures Variables whose attributeshave only the character- istics of exhaustivenessand mutual exclusiveness are nominalmeasures.Examples include gender, religious fiation, politicalparty fiation, birth- place, collegemajor, and hair color. Although the attributes composingeach of these variables-as male and female composethe variablegender-are distinct fromone another (andexhaustthe possi- bilities of gender among people),they have no ad- ditional structures.Nominal measuresmerely offer names or labelsfor characteristics. Imagine a group of people characterizedin terms of one such variable and physically grouped by the applicable attributes.For example, say we've asked a large gathering of people to stand together in groups accordingto the statesin which they were born: all those born in Vermont in one group, those born in Californiain another, and so forth. The variable is place of birth; the attributesare born in California,born in Vermont, and so on. All the people standingin a given group have at least one thing in common and differ from the people in all other groups in that same regard.Where the indi- vidual groups form, how close they are to one an- other, or how the groups are arranged in the room is irrelevant. All that matters is that all the mem- bers of a given group share the same state of birth and that each group has a different shared state of birth.All we can say about two people in terms of a nominal variable is that they are eitherthe same or different. OrdinalMeasures Variableswith attributeswe can logicallyrank- order are ordinalmeasures.The different attrib- utes of ordinalvariables representrelatively more or less of the variable. Variables of this type are so- cial class, conservatism, alienation, prejudice, intellectual sophisticatiorz, and the like.In additionto saying whether two people are the same or different in terms of an ordinal variable, you can also say one is "more"than the other-that is, more conserva- tive, more religious, older, and so forth. In the physical sciences, hardness is the most frequently cited example of an ordinalmeasure. We may say that one material (forexample, dia- mond) is harder than another (say,glass)if the for- mer can scratch the latter and not vice versa. By attemptingto scratch various materialswith other materials, we might eventually be able to arrange severalmaterialsin a row, rangirlgfrom the softest to the hardest. We could never say how hard a given material was in absolute terms; we could only say how hard in relative terms-which mate- rials it is harder than and which softerthan. Let's pursue the earlier example of grouping the people at a social gathering. This time imagine that we ask all the people who have graduated from collegeto standin one group, alI those with only a high school diploma to stand in another group, and all those who have not graduated from high schoolto stand in a third group. This manner of groupingpeople satisfies the requirements for exhaustivenessand mutual exclusivenessdiscussed earlier.In addition, however, we might logically arrange the three groupsin terms of the relative amount of formal education (theshared attribute) each had. We might arrangethe three groups in a row, ranging from most to least formal education. This arrangementwould provide a physical repre- sentationof an ordinalmeasure.If we knew which groups two individuals were in, we could deter- mine that one had more, less, or the same formal . educationas the other. Notice in this example that it is irrelevant how close or far apart the educationalgroupsare from Operationalization Choices . 135 one another. The college and high schoolgroups might be 5 feet apart, and the less-than-high- school group 500 feet farther down the line. These actual distancesdon't have any meaning. The high schoolgroup,however, shouldbe between the less- than-high-school group and the college group, or else the rank order will be incorrect. IntervalMeasures For the attributes composing some variables, the actual distance separatingthose attributesdoeshave meaning. Suchvariablesare intervalmeasures. For these, the logicaldistance between attributes canbe expressedinmeaningfulstandard intervals. For example, in the Fahrenheittemperature scale, the difference,or distance, between 80 de- grees and 90 degreesis the same as that between 40 degreesand 50 degrees.However, 80 degrees Fahrenheit is not twice as hot as 40 degrees, be- cause the zero points in the Fahrenheit (andCel- sius)scales are arbitrary;zero degrees does not re- mean lack of heat. Similarly,minus 30 degrees on either scale doesn't represent 30 degrees less than no heat. (Incontrast,the ICelvin scale is based on an absolutezero, which does mean a complete lack of heat.) About the onlyintervalmeasures commonly used in social scientificresearch are constructed measures such as standardizedintelligencetests that have been more or less accepted.The inter- val separatingIQ scores of 100and 110may be regarded as the same as the interval separating scores of 110and 120by virtue of the distribution of observed scoresobtained by many thousands of people who have taken the tests over the years. But it would be incorrect to infer that someone with an IQ of 150is 50 percent more intelligent than someonewith an IQ of 100. (Aperson who received a score of 0 on a standard IQ test could not be regarded, strictly speaking, as having no intelli- gence, althoughwe might feel he or she was un- suitedto be a collegeprofessor or even a college student.But perhaps a dean . ..?) When comparing two people in terms of an in- tervalvariable, we can say they are different from one another (nominal),and that one is more than Určeno pouze pro studijní účely Operationalization Choices . 137 another (ordinal).In addition, we can say "how much"more. RatioMeasures Most of the social scientificvariables meeting the minimum requirements for intervalmeasures also meet the requirementsfor ratio measures.In ratio measures,the attributes composinga variable, be- sides having allthe structural characteristicsmen- tioned previously, are based on a true zero point. The Kelvin temperature scaleis one such measure. Examples from social scientificresearch include age, length of residencein a givenplace, number of organizationsbelonged to, number of timesattend- ing church during a particularperiod of time, num- ber of times married, and number of Arab friends. Returning to the illustration of methodological party games, we might ask a gatheringof people to group themselvesby age. AU the one-year-olds would stand (orsit or lie) together, the two-year- olds together,the thee-year-olds, and so forth. The fact that members of a single group share the same age and that each differentgroup has a different shared age satisfiesthe minimum requirementsfor a nominal measure. Arranging the severalgroups in a line from youngest to oldest meets the addi- tional requirements of an ordinalmeasure and lets us determine if one person is older than, younger than, or the same age as another. If we spacethe groups equally far apart, we satisfy the additional requirements of an internalmeasure and will be able to say how much older one person is than another. Finally,because one of the attributesin- cludedin age represents a true zero (babiescarried by women about to give birlh), the phalanx of haplessparty goers also meets the requirements of a ratio measure, permittingus to say that one person is twice as old as another. (Rememberthis in case you're asked about it in a workbook assign- ment.) Another example of a ratio measure is in- come,which extendsfrom an absolute zero to approximatelyW t y , if you happen to be the founder of Microsoft. Comparingtwo people in terms of a ratio vari- able, then, allowsus to conclude (1)they are differ- ent (orthe same), (2) one is more than the other, (3)how much they differ, and (4)the ratio of one to another. Figure 5-1summarizesthis discussion by presenting a graphicillustration of the four lev- els of measurement. implicationsofLevelsofMeasurement Because it's unlikely that you'll undertake the physicalgrouping of peoplejust described (tryit once, and you won't be invited to many parties),I should drawyour attention to some of the practical implications ofthe differencesthat have been dis- tinguished. These implications appearprimarily in the analysis of data (discussedin Part 4),but you need to anticipate such implicationswhen you're structuringany research project. Certain quantitativeanalysis techniques re- quire variables that meet certainminimum levels ofmeasurement.To the extent that the variables to be examinedin a research project are limited to a particular level of measurement-say, ordinal- you shouldplan your analyticaltechniques accord- ingly.More precisely, you shouldanticipatedraw- ing research conclusionsappropriate to the levels of measurement used in you variables. For example, you might reasonablyplan to determine and report the mean age of a populationunder study (addup all the individualages and divide by the number of people),but you should not plan to report the mean religious mation, because that is a nominal variable, and the mean requiresratio-leveldata. (Youcouldreport the modal-the most common- religious affiliation.) At the same time, you can treat some variables as representingdifferentlevels of measurement. Ratio measures are the highest level, descending through interval and ordinalto nominal, the lowest level of measurement.Avariable representinga higher level of measurement-say, ratio-may also be treated as representing a lower level of measure- ment-say, ordinal. Recall,for example, that age is a ratio measure.If you wished to examine only the relationshipbetween age andsome ordinal-level variable-say, self-perceivedreligiosity:high, medium, and low-you might chooseto treat age as an ordinal-levelvariable as well. You might char- acterizethe subjectsof your study as being young, middle-aged,and old, specifyingwhat age range composed each of these groupings.Finally, age FIGURE 5-1 LevelsofMeasurement Nominal Measure Example: Gender I Female Male II Ordinal Measure Example: Religiosity "Howimportant is religion to you?" I Low > High Interval Measure Example: IQ Ratio Measure Example: Income might be used as a nominal-levelvariable for cer- tain research purposes. People might be grouped as being born during the depression of the 1930s or not. Another nominal measui-ement,based on birth date rather than just age, would be the group- ing of people by astrologicalsigns. The level of measurenlent you'll seek, then, is determined by the analyticaluses you've planned fora given variable, keepingin mind that some variables are inherently limited to a certainIevel. If a variable is to be used in a variety of ways, re- quiring differentlevels of measurement, the study should be designed to achieve the highest level required.Por example,if the subjects in a study are aslced their exact ages,they can later be organized into ordinal or nominal groupings. You need not necessarily measure variables at their highest level of measurement,however.If you're sure to have no need for ages ofpeople at higher than the ordinal level of measurement, you Určeno pouze pro studijní účely 138 .Chapter5: Conceptualization,Operationalization,andMeasurement may simply askpeople to indicate their age range, such as 2 0 to 29, 30 to 39, and so forth. In a study of the wealth of corporations, rather than seek more precise information,you may use Dun S. ~radstreetratings to rank corporations.Whenever your researchpurposes are not altogetherclear, however, seek the highest level of measurement possible.Again, although ratio measures can later be reduced to ordinalones, you cannot convert an ordinal measure to a ratio one. More generally, you cannot convert a lower-levelmeasure to a higher-level one. That is a one-way streetworth remembering. SingleorMultipleIndicators With so many alternatives for operationdizbg socialscientiilcvariables, you may h d yourself worrying about making the right choices.TO counter this feeling,let me add a momentary dash of certaintyand stability. Many socialresearch variables have fairly obvi- ous, straightforwardmeasures. No matter how you cut it, gender usually turns out to be a matter of male or female:a nominal-levelvariablethat can be measuredby a single observation-either look- ing (well,not always) or asldng a question (usu- ally).In a studyinvolvingthe size of families,you'll want to think about adopted and foster children, as well asblended families,but it's usuallypretty easy to find out how many children a family has. For most researchpurposes, the resident population of a country is the residentpopulation of that coun- try-you canlook it up in an almanacand know the answer.A great many variables, then, have ob- vious singleindicators.If you can get one piece of information, you have what you need. Sometimes,however,there is no singleindi- cator that will give you the measure of a variable you really want. As discussed earlier in this chap- ter, many conceptsare subject to varying inter- pretations-each with severalpossible indicators. In these cases, you'll want to make several obser- vations for a given variable.You can then com- bine the severalpieces of informationyou've col- lected to create a compositemeasurement of the variable in question. Chapter 6 is devoted to ways of doing that, so here let's just consider one simple illustration. Considerthe concept "collegeperformance." Allof us have noticed that some studentsperform well in college courses and others don't perform well. In studying these differences,we might ask what characteristics and experiencesare related to high levels of performance (manyresearchers have donejust that).How shouldwe measure overall performance? Each gradein any single course is a potential indicator of collegeperformance, but it also may not typify the student's generalperfor- mance. The solutionto this problem is so bmly established that it is, of course, obvious:the grade point average (GPA).We assignnumericalscores to eachletter grade,total the points earned by a given student, and divide by the number of coursestaken to obtain a composite measure. (If the coursesvary in number of credits,we adjust the point valuesac- cordingly.)It is often appropriate to create such composite measures in social research. SomeIllustrationsof OperationalizationChoices To bring together allthe operationalizationchoices availableto the socialresearcher and to show the potential in those possibilities, let's loolcat some of the distinctways you might addressvarious research problems.The alternativeways of opera- tionalizing the variables in each case should dem- onstratethe opportunitiesthat social research can present to our ingenuity and imaginations.To sim- pMy matters, I have not attemptedto describe all the research conditionsthat would make one alter- native superior to the others,though in a given sit- uation they would not allbe equally appropriate. Here are specificresearch questions, then, and some of the ways you could address them. We'll be- gin with an example discussed earlierin the chap- ter. It has the added advantagethat one of the vari- ables is straightforwardto operationalize. 1. Are women more compassionatethan men? a. Select a group of subjectsfor study, with equalnumbers of men and women. Present them with hypotheticalsituationsthat in- volve someone's being in trouble. Ask them what they would do if they were con- fronted with that situation.What would they do, for example,ifthey came aaoss a small child who was lost and crying for his or her parents? Consider any answerthat involveshelping or comfortingthe child as an indicator of compassion. See whether men or women are more likely to indicate they would be compassionate. b. Set up an experiment in which you pay a smallchild to pretend that lle or she is lost. Put the child to work on a busy side- wall~and observewhether men or women are more likely to offer assistance. Also be sure to count the totalnumber of men and women who wallc by, because there may be more of one than the other. If that's the case, simply calculatethe percent- age ofmen and the percentage of women who help. c. Select a sample of people and do a survey inwhich you ask them what organizations theybelong to. Calculatewhether women or men are more likely to belong to those that seemto refiect compassionatefeelings. To talceaccount of men who belongto more organizationsthan do women in general- or vice versa-do this: For each person you study,calculatethe percentageof his or her organizationalmemberships that reflect compassion. Seeif men or women have a higher averagepercentage. 2. Are sociology students or accountingstudents better informed about world affairs? a. Prepare a short quiz on world affairs and arrangeto administer it to the studentsin a sociology class and in an accounting class at a comparablelevel.If you want to compare sociology and accountingmajors, be sure to ask students what they are majoringin. b. Get the instructor of a course in world affairs to giveyou the averagegrades of sociology and accounting students in the course. c. Take a petitionto sociology and accounting classes that urges that "the United Nations OperationalizationChoices . 139 headquartersbe moved to New York City." Keep a count of how many in each class sign the petition and how many inform you that the UN headquarters is already located in New York City. 3. Do people consider New York or California the better place to live? a. Consulting the StatisticalAbstract of tlze Ulzited States or a similarpublication, check the migrationrates into and out of each state. See if you can find the numbers mov- ing directly fromNew Yorlcto California and vice versa. b. The national p o h g companies-Gallup, Harris, Roper, and so forth-often ask people what they consider the best stateto live in.Look up some recent results in the library or through your localnewspaper. c. Compare suicide rates in the two states. 4. Who are the most popular instructors on your campus, those in the social sciences,the natural sciences,or the humanities? i a. If your school has a provision for student evaluationof instructors, review some recent results and compute the average ratings given the three groups. b. Begin visitingthe introductory courses given in each group of disciplines and mea- sure the attendancerate of each class. c. In December, select a group of faculty in each of the three divisionsand ask them to keep a record of the numbers of holiday greeting cards and presentsthey receive from admiring students. Seewho wins. The point ofthese examples is not necessarilyto suggestrespectable researchprojectsbut to illustrate the many ways variablescan be operationalized. OperationalizationGoesOnandOn Although I've discussed conceptualizationand operationalizationas activitiesthat precede data collectionand analysis-for example,you must design questionnaire itemsbefore you send out a questionnaire-these two processes continue throughout any research project, even if the data Určeno pouze pro studijní účely 140 .Chapter5: Conceptualization,Operationalization,and Measuremt have been collectedin a structured mass survey. As we've seen, in less-structuredmethods such as fieldresearch, the identificationand specification of relevant conceptsis inseparablefrom the ongoing process of observation. As a researcher, always be open to reexamining your concepts and definitions.The ultimate pur- pose of social research is to clarify the nature of so- cial life. The validity and utility of what you learn in this regard doesn't depend on when you first figured out how to look at things any more than it matters whether you got the idea from a learned textbook, a dream, or your brother-in-law. Criteriaof MeasurementQuality This chapter has come some distance.It beganwith the bald assertion that social scientistscan measure anythmg that exists. Then we discovered that most of the things we might want to measure and study don't really exist.Next we learned that it's possible to measure them anyway. Now we concludethe chapter with a discussionof some of the yardsticks against which we judge our relative success or fail- ure in measuring things-even things that don't exist. PrecisionandAccuracy To begin, measurements can be made with varying degrees of precision. As we saw in the discussion of operationalization,precision concernsthe fineness of distinctionsmade between the attributes that composea variable. The descriptionof a woman as "43 years old is more precise than "in her forties." Saying a street-corner gangwas formed in the summer of 1996is more precise than saying "dur- ing the 1990s." As a general rule, precise measurements are su- perior to imprecise ones, as common sense would dictate.There are no conditionsunder which im- precisemeasurements are intrinsically superiorto precise ones. Even so, exact precision is not always necessary or desirable.If knowing that a woman is in her forties satisfiesyour research requirements, then any additionaleffort investedin learning her precise age is wasted. The operationalization of con- cepts, then, must be guidedpartly by an under- standingof the degree of precision required. If your needs are not clear,be more preciserather than less. Don't confuse precision with accuracy,how- ever. Describing someone as "born in New En- gland is less precise than "born in Stowe, Ver- mont"-but suppose the person in question was actuallyborn in Boston. The less-precisedescrip- tion, in this instance, is more accurate, a better reflection of the real world. Precision and accuracyare obviously important qualitiesin research measurement, and they proba- bly need no further explanation.When socialsci- entists construct and evaluate measurements, how- ever, they pay special attention to two technical considerations:reliabilityand validity. Reliability In the abstract, reliabilityis a matter of whether a particulartechnique, appliedrepeatedlyto the same object,yields the same result each time. Let's say you want to know how much I weigh. (No,I don't know why.)As one technique, sayyou ask two differentpeople to estimatemy weight.If the firstperson estimates 150pounds and the other esti- mates 300, we have to concludethe technique of havingpeople estimatemy weightisn't very reliable. Suppose, as an alternative,that you use a bath- room scale as your measurement technique. I step on the scale twice, and you note the result each time. The scale has presumably reported the same weight for me both times, indicating that the scale provides a more reliable technique for measuring a person's weight than does askingpeople to esti- mate it. Reliability, however, does not ensure accuracy any more than precision does. SupposeI've set my bathroom scale to shave five pounds off my weight just to make me feelbetter. Although you would (reliably)report the sameweight for me each time, you would alwaysbe wrong. This new element, called bias,is discussedin Chapter 8. For now, just be warned that reliabilitydoes not ensure accuracy. Let's supposewe're interested in studying morale among factory workersin two different kinds of factories.In one set of factories,workers have specializedjobs, reflecting an extreme division of labor. Each worker contributes a tiny part to the overallprocessperformed on a long assemblyline. In the other set of factories, each worker performs many tasks, and small teams of workers complete the whole process. How should we measure morale? Following one strategy,we could observe the workers in each factory,noticing such things as whether theyjoke with one another, whether they smile and laugh a lot, and so forth. We could ask them how they like their work and even ask them whether they think they would prefer their current arrangement or the other one being studied. By comparingwhat we observed in the differentfactories,we might reach a conclusionabout which assembly processpro- duces the higher morale. Now let's look at some reliabilityproblems in- herent in this method. First, how you and I are feeling when we do the observingwill likely color what we see.We may misinterpret what we see. We may see workers kidding each other but think they're having an argument. We may catch them on an off day. If we were to observe the same group of workers several days in a row, we might arrive at differentevaluations on each day. If several ob- servers evaluatedthe samebehavior, on the other hand, they too might arrive at different conclusions about the workers' morale. Here's another strategy for assessing morale. Suppose we check the company records to see how many grievanceshave been med with the union during some fixed period. Presumably this would be an indicator of morale: the more grievances, the lower the morale. This measurement strategy would appear to be more reliable: Countingup the grievances over and over, we should keep arriving at the same number. If you b d yourself thinking that the number of grievances doesn't necessarilymeasure morale, you're worrying about validity, not reliability. We'll discuss validity in a moment. The point for now is that the last method is more like my bathroom scale-it gives consistentresults. In social research, reliabilityproblems crop up in many forms.Reliabilityis a concern every time a single observer is the source of data, because we have no certain guard against the impact of that Lobserver's subjectivity. We can't tell for sure how Criteria ofMeasurementQuality .141 much of what's reported originated in the situation observed and how much in the observer. Subjectivityis not only a problem with single observers,however. Surveyresearchers have known for a long time that different interviewers, because of their own attitudes and demeanors, get differentanswers from respondents. Or, if we were to conduct a study of newspapers' editorial posi- tions on somepublic issue,we might create a team of coders to take on thejob of readinghundreds of editorialsand classifying them in terms of their po- sition on the issue. Unfortunately, different coders will codethe same editorial differently. Or we might want to classify a fewhundred specificoccupations in terms of some standard codingscheme, say a set of categories created by the Department of Labor or by the CensusBureau. You and I would not place allthose occupationsin the same categories. Each of these examples illustratesproblems of reliability. Similarproblems arise whenever we ask people to give us informationabout themselves. Sometimes we ask questionsthat people don't know the answers to: How many times have you f been to church? Sometimeswe ask people about things they considertotally irrelevant:Are you satisfiedwith China's current relationshipwith Al- bania? In such cases, people will answer differently at differenttimes because they're making up an- swersas they go. Sometimeswe exploreissues so complicated that a person who had a clear opinion in the matter might arrive at a different interpreta- tion of the questionwhen asked a secondtime. Sohow do you create reliable measures? If your research design calls for askingpeople for information,you can be careful to ask only about things the respondents are likelyto know the an- swer to. Ask about things relevant to them, and be clear in what you're asking. Of course, these tech- niques don't solve everypossible reliabilityprob- lem. Fortunately, socialresearchershave developed severaltechniques for cross-checkingthe reliability of the measures they devise. Test-RetestMethod Sometimesit's appropriateto make the same measurement more than once, a technique called the test-retestmethod. If you don't expect the Určeno pouze pro studijní účely 144 .Chapter5: Conceptualization,Operationalization,andMeasuremer Suppose,for example, that you want to study the sourcesand consequences of marital satisfac- tion. As part of your research, you develop a mea- sure of marital satisfaction, and you want to assess itsvalidity. In addition to developingyour measure, you'll have developed certaintheoretical expectations about the way the variablemarital satisfactionre- lates to other variables. For example,you might reasonably concludethat satisfied husbands and wives will be less likely than dissatisfiedones to cheat on their spouses.If your measure relates to marital fidelityin the expectedfashion,that consti- tutes evidence of your measure's construct validity. If satisfiedmarriage partners are as likely to cheat on their spouses as are the dissatisfiedones, how- ever, that would challenge the validity of your measure. Tests of construct validity, then, can offera weight of evidencethat your measure either does or doesn't tap the quality you want it to measure, witliout providing definitiveproof. Although I have suggestedthat tests of constructvalidity are less compellingthan those of criterionvalidity, there is room for disagreement about which ldnd of test a partidar comparison variable (drivingrecord, marital fidelity)representsin a given situation.It is less important to distinguishthe two types of valid- ity tests than to understand the logic of validation that they have in common:LEwe have been suc- cessfulin measuring somevariable, then our meas- ures should relate in some logicalway to other measures. Finally, contentvalidity refers to how much a measure covers the range of meaningsincluded within a concept.For example, a test of mathemati- cal ability cannot be limited to addition alonebut also needs to cover subtraction,multiplication,divi- sion, and soforth. Or,if we are measuring prejudice, do our measurementsreflectalltypes of prejudice, includingprejudice against racial and ethnic groups, religious minorities, women, the elderly, and so on? Figure 5-2 presents a graphicportrayal of the differencebetween validity and reliability.If you think of measurement as analogousto repeatedly shooting at the bull's-eye on a target, you'll see that reliabilitylookslike a "tightpattern," regardless of where the shotshit, becausereliability is a function of consistency.Validity, on the other hand, is a functionof shotsbeing arranged around the bull's- eye.The failureof reliabilityin the figure is ran- domly distributed aroundthe target;the failure of validity is systematicdy off the mark. Notice that neither an unreliable nor an invalidmeasure is Mely to be very useful. WhoDecidesWhat'sValid? Our discussionof validitybegan with a reminder that we depend on agreementsto determine what's real, and we've just seen some of the ways social scientistscan agree amongthemselves that they have made valid measurements.There is yet an- other way of lookingat validity. Social researchers sometimes criticizethem- selves and one another for implicitly assuming they are somewhat superior to those they study.For ex- ample, researchers often seekto uncover motiva- tions that the social actorsthemselves are unaware of. You third