The concept of the "case" is a basic feature of social science research, and yet manyquestionsabouthowacaseshouldbedefined,howcasesshouldbe selected, andwhatthe criteria arefora goodcase orset ofcases are farfrom settled.Are casespreexistingphenomenathatneedonlybeidentifiedbythe researcher before analysis can begin? Or are cases constructed during the course of research, onlyafteranalysishasrevealedwhichfeaturesshouldbeconsid-ereddefiningcharacteristics?Willcasesbeselectedrandomlyfromthetotal pool of available cases? Or will cases be chosen because of their uniquequali-ties?To whatdegreemustcasesbecomparable?Isthelogicofquantitative research, usingalarge numberof cases, fundamentallydifferentfromthat of qualitative research, using only one or a few cases? These questionsandmanyothers areaddressedbythe contributorstothis volume astheyprobethenatureofthecase andthewaysinwhichdifferent understandings of what a case is affectthe conduct and the results of research. The contributors find a good deal of common ground, and yet they also express strikinglydifferentviewsonmanykeypoints.Inhisintroduction,Charles Raginprovidesaframeworkfordistinguishingfourfundamentallydifferent approaches to case-based research. These approaches are organized around two dichotomiesinhowcasesareconceived:whethertheyareconsideredtobe empiricalunitsortheoreticalconstructsandwhethertheyareunderstoodas examplesofgeneralphenomenaorasspecificphenomena.Eachapproach involvesproceduralandanalyticalguidelinesthatwillaffectthecourseof research and the conclusions the research draws. As Ragin argues and the other contributors demonstrate, the workof any givenresearcheroftenis character-izedbysome hybridofthese basic approaches, anditis importanttounder-stand that most research involves multiple definitions and uses of cases, both as specific empirical phenomena and as general theoretical categories. What is a case? WHAT IS A CASE? Exploringthefoundationsofsocialinquiry Edited by Charles C. Ragin, NorthwesternUniversity and Howard S. Becker, University of Washington mm CAMBRIDGE UNIVERSITYPRESS CAMBRIDGEUNIVERSITYPRESS Cambridge, New York,Melbourne,Madrid,CapeTown,Singapore, SaoPaulo,Delhi,Dubai,Tokyo,MexicoCity CambridgeUniversityPress 32 Avenueof the Americas, NewYork, NY10013-2473, USA www.cambridge.orgInformationon thistitle: www.cambridge.org/9780521420501CambridgeUniversityPress1992 Thispublicationisincopyright.Subjecttostatutoryexception andtotheprovisionsof relevantcollectivelicensingagreements, noreproductionof anypartmaytakeplacewithoutthewritten permissionofCambridgeUniversityPress. Firstpublished1992 12th printing2009 A catalogrecord forthis publicationis availablefromtheBritishLibrary ISBN978-0-521-42050-1Hardback ISBN978-0-521-42188-1Paperback CambridgeUniversityPresshasnoresponsibilityforthepersistenceor accuracyof URLsforexternalor third-partyInternetWebsitesreferredtoin thispublication,anddoesnotguaranteethatanycontentonsuchWebsitesis, orwillremain,accurateorappropriate.Informationregardingprices,travel timetables,andotherfactualinformationgivenin thisworkarecorrectat thetimeoffirstprintingbutCambridgeUniversityPressdoesnotguarantee theaccuracyofsuchinformationthereafter. Contents ContributorsIntroduction: Cases of "What is a case?"Charles C. Ragin ICritiques of conventional practices1Cases of cases . . .of casesJenniferPiatt 2What do cases do? Some notes on activity in sociological analysisAndrew Abbott 3Cases are for identity, for explanation, or for controlHarrison C. White 4Small N's and big conclusions: an examination of the reasoning in comparative studies based on a small number of casesStanley Lieberson IIAnalyses of research experiences5Making the theoretical caseJohn Walton 6Small N's and community case studiesDouglas Harper v VII119215383105119121139viCONTENTS 7Case studies: history or sociology?Michel Wieviorka 8Theory elaboration: the heuristics of case analysisDiane Vaughan IIIReflectionson "What is a case?"9Cases, causes, conjunctures,stories, andimageryHoward S. Becker 10"Casing" and the process of social inquiryCharles C. Ragin NotesIndex 235227217205203173159Contributors Andrew Abbottis Professorof Sociology at the Universityof Chicago. His research spans occupationsandthe professionsand basic issues in social science methodology. His bookThe System of Professions wonthe Distinguished Contribution to Scholarship Award of the American Soci-ological Association in 1991. HowardS. Becker is Professorof Sociology at the Universityof Wash-ington.HispublicationsincludeOutsiders,Boys inWhite,Art Worlds, Doing ThingsTogether, andWriting for SocialScientists.He has taughtat Northwestern University, the Visual Studies Workshop, andthe Museu Nacional (Rio de Janeiro). Douglas Harperis Professorof Sociology andChair of theDepartment of Sociology at the University of South Florida. His works include Good Company andWorking Knowledge: Skill and Community in a Small Shop. Stanley Lieberson is Abbott Lawrence Lowell Professorof Sociology at HarvardUniversityandPastPresidentoftheAmericanSociological Association. He is the author of numerous articles and books, including Making It Count: The Improvement of Social Research and Theory, A Piece of the Pie:Blacks and White ImmigrantsSince1880,and From Many Strands: Racial and Ethnic Groups in Contemporary America. JenniferPiatt is Professorof Sociology at the Universityof Sussex. She hasbeeneditorofthejournalSociologyandPresidentoftheBritish SociologicalAssociation.Herresearchinterestsareinthehistoryof sociologicalresearchmethodsandinanalyticmethodology,especially the logic of case studies. vii viiiCONTRIBUTORS CharlesC.RaginisProfessorofSociologyandPoliticalScienceat NorthwesternUniversity.Hiswritingsonmethodologyspanmacro-sociologyandcomparativepolitics;hissubstantiveworkaddresses ethnicity and political economy. His book The Comparative Method: Mov-ing BeyondQualitative and Quantitative Strategies wonthe SteinRokkan Prize for comparative research. John Walton is Professorof Sociology and Anthropologyat the Univer-sity of California, Davis. He has published various books and articles in the fields of Third World development, historical sociology, revolution, andcollectiveaction,includingtheaward-winningReluctantRebels: ComparativeStudiesof Revolutionand Underdevelopment.Hischapterin thisvolumeisareflectionontheresearchthatledtohis1992book Western Times and WaterWars: State, Culture, and Rebellion in California. Diane Vaughanis AssociateProfessorofSociologyatBostonCollege. Her research areas include the sociology of organizations, deviance and social control, transitions, and qualitative methods. She is the author of Controlling Unlawful Organizational Behavior: Social Control and Corporate Misconduct andUncoupling: Turning Points in Intimate Relationships. Cur-rently she is writing a historical ethnographyexplaining the Challenger launchdecisionandabookthatdevelopstheideaspresentedinthis volume. Michel Wieviorka is Directeur d'etudesat the Ecole des HautesEtudes en Sciences Sociales in Paris and Deputy Director of the Centre d'Anal-yse et d'InterventionSociologiques. He is the authorofseveral books, includingTheWorkersMovement (withAlainTouraineandFrancois Dubet), Societeset terrorisme(Englishtranslationforthcomingfromthe University of Chicago Press), and L'espace du racisme. Harrison White is the authorof the forthcomingIdentity and Control: A StructuralTheoryof SocialAction (PrincetonUniversityPress,1992), as wellasbooksonvacancychains,ontheriseoftheFrenchImpres-sionists, andon role structures, plus articles modeling social networks, industrial markets, andsocial mobility. Currently he chairs the Depart-ment of Sociology at ColumbiaUniversity. Introduction: Cases of"What is a case?"1 CHARLESC.RAGIN The precept of case analysis Social science methodologyis anchoredby a numberof basicprecepts that are rarely questioned by practitioners. One precept that is central to the logic of analysis is the idea of having cases. Social scientists use terms like"N ofcases,""case study,"and"sampleofcases"withrelatively little consideration of the possible theories and metatheories embedded in these terms or in the methods that use cases andmake conventional forms of analysis possible. For example, we may describe aninvestiga-tionasacase studybecauseitinvolvesethnographicresearchinone setting, without ever considering what constitutes a case study or whether therearemethodologicallydecisivedifferencesbetweencasestudies andother kinds ofstudies. Anotherexample: In quantitativeresearch, we use the terms "cases" and"units of analysis" interchangeably with-outconsideringtheproblemsthatmightcomefromconflatingdata categoriesandtheoreticalcategories(Ragin1987:7-9). Oneresearcher may use families as data-collectionsites in a survey study; anothermay writeabookcalledWhat Is Family?(GubriumandHolstein1990). A third example: A study that uses interviews of employees to construct a picture of the informal organization of a firm looks superficially like one thatusesinterviewsofemployeesto addressvariationin jobsatisfac-tion.Bothstudiesuseinterviewsofemployeesastheprimarydata source, butthefirstis aboutthefirmasawhole, whilethe secondis aboutemployees'subjectivestates.Inshort,theterm"case"andthe various terms linked to the idea of case analysis are not well definedin socialscience,despitetheirwidespreadusageandtheircentralityto social scientific discourse. Implicitinmostsocial scientificnotionsofcase analysisis theidea that the objects of investigation are similar enough and separate enough topermittreatingthemascomparableinstancesofthesamegeneral phenomenon.Ataminimum,mostsocialscientistsbelievethattheir 1 2WHATISACASE? methods are powerfulenough to overwhelm the uniqueness inherent in objects andevents in the social world. The idea of comparablecases is implicatedinthe boundarybetweendominantformsofsocialscience andothertypesofdiscourseaboutsociallife(e.g., journalismand,in manyquarters,history).Theaudiencesforsocialscienceexpectthe resultsofsocialscientificinvestigationtobebasedonsystematicap-praisalofempiricalevidence.Useofevidencethatisrepetitiousand extensive in form, as when it is based on observations of many cases or of variedcases, has provedto be a dependable way for social scientists to substantiate their arguments. Social scientists who conduct case stud-ies argue that their cases are typical or exemplary or extreme or theoret-ically decisive in some other way. Thus, even in case-study research the principle ofrepetitionis oftenimplicatedin statementsconcerningthe relationbetweenthechosencaseandothercases.[Foranextended discussionofthisissue,seeFeagin,Orum,andSjoberg(1991)and especially Sjoberg et al. (1991).] Nomatterwhichcaseorunitinvestigatorsuseintheirempirical analyses,theytypicallyinvokeadditionalunitsinthepresentationof their research. An analysis of individual-levelsurveydata froma sam-ple ofadultsinthe UnitedStates, forexample, providesafoundation forstatementsaboutindividualsandaboutthe UnitedStates (in addi-tion to the rangeof unitsin betweenthese two poles -e.g.,communi-ties, cities, socialclasses). A studyofthistypecan beseenbothasan extensive analysis of many cases (the sample of individuals)andas an intensive case study of the United States. Further, the United States may be seen as a case in several differentsenses. For example, it may be seen as a memberofa largersetofbroadlydefinedobjects(e.g., advanced societies), or as an instance of an importanttheoreticalconceptor pro-cess(e.g.,partialimplementationofmeritocraticprinciples),orasan intrinsically interesting historical or cultural entity in its own right (e.g., a country that is changing in historically specific or decisive ways). This example shows as well that while it is tempting to see the case study as atypeofqualitativeanalysis,andperhapseventoequatethetwo, virtually every social scientific study is a case study or can be conceived as a case study, often from a variety of viewpoints. At a minimum, every studyisacasestudybecauseitisananalysisofsocialphenomena specific to time and place. When presenting their results, investigators manipulate both empiri-calcasesandtheoreticalcases, andthesedifferentcasesmayvaryby level, as whentheyare nestedor hierarchicallyarrayed, andtheymay varyinspecificity.Agivenbodyofempiricalevidencealsocanbe constructedinto a varietyof differentsubstantivecases (a case of mis-Introduction: Cases of "Whatis a case?"3 management, or a case of overwhelming external pressures, or a case of inertia,etc.). To thequestion"Whatisacase?"mostsocialscientists wouldhavetogivemultipleanswers.Acasemaybetheoreticalor empiricalorboth; itmaybearelativelyboundedobjectora process; anditmaybegenericanduniversalorspecificinsomeway.Asking "Whatis a case?" questions manydifferentaspects ofempiricalsocial science. Conceptions of "cases" and social science discourse The various usages andmeanings of the term"case" havefar-reaching implicationsforthe conductof social science; this factalone is enough to justifyquestioning its status. The issue also deserves carefulconsid-eration because differentconceptionsoftheterm"case" are centralto theenduringgulfbetweenquantitativeandqualitativesocial science. The term"case"isoneofmanybasicmethodologicalconstructsthat have become distorted or corrupted over time. The typical pattern is for a key methodologicaltermto gain multipleandsometimes contradic-tory meanings. Consider, for example, the term "cause." A fundamental rule of quantitative social science is that"correlationis notcausation." Yet social scientists routinelymakestatementsthatone variable causes variationinanother,whentheevidenceisbasedentirelyoncorrela-tionalpatterns.Theterm"cause"hasbeenpermittedmultiple,non-overlappingmeanings.Itisonlywhencriticschallengearesearcher whousescorrelationalevidencethat"cause"inthesenseofdocu-mentedempiricalconnectionsisaddressed(seeBecker'sdiscussion, Chapter 9). The term "control" offers another example. Originally, "con-trol" referredalmost exclusively to experimental designs, where causal factorsare directlymanipulatedbytheinvestigator(hence theappro-priatenessoftheterm"control"). Todayweusetheterm"control"to refer as well to arithmetic adjustmentsbasedon assumptions of causal linearity andadditivity in analyses of naturally occurring data. [A con-ceptuallyorientedoverviewofthese practicesis offeredbyLieberson (1985).] Thus, we say that we control statistically for the effectof paren-talsocialclasswhenweassesstheeffectsofpublic-versus-private-schoolattendanceonstudentperformance,wheninfactwehaveno real controlatalloverwhoattendswhichschool.Thetermhasbeen broadened in meaning to refer to very differentresearch activities. The same holds true fortheterm"case." The view thatquantitative researcherslookatmanycases,whilequalitativeresearcherslookat only one or a small number of cases, can be maintainedonly by allow-ing considerable slippage in what is meant by "case." The ethnographer 4WHATISACASE? who interviews the employees of a firm in order to uncover its informal organization has at least as much empirical data as the researcherwho uses these same interviews to construct a data set appropriate for quan-titativeassessmentofvariationamongemployeesinjobsatisfaction. Both have data on employees andon the firm,and bothproducefind-ings specific in time and place to that single firm. Further, both research-ers make sense of their findings by connecting them to studies of other firms.Yettheethnographerissaidtohavebutonecaseandtobe conductingacasestudy,whilethequantitativeresearcherisseenas having many cases. In this light, muchofwhatis consideredlarge-N research alsomust be seen as case-studyresearch, andthe tendencyto conflate qualitative study and case study should be resisted. To apply the same term to vastly differentmethodologicalconstructs serves only to increase the percep-tion that the differentkinds of social science are irreconcilable andthat their practitioners speak mutually unintelligible languages. We need to strive for greater clarity in what we mean by "case" and differentiateits variousmeanings. Thisemendmentofcurrentpracticeswillsimplify the task of linking qualitative and quantitative research and bring greater richness and unity to the conduct of social science (Ragin 1991). Considerthisbookafirststepinconfrontingthisimportanttask. Collectively,thecontributorshavequestionedtheterm"case"froma variety of viewpoints; their contributions to this volume can be seen as the groundwork for future efforts toward refining the various meanings of"case."Asthecontributionsshow,itisdifficulttoask"Whatisa case?" withoutaddressingother bases of social scientificmethodology. Askingthisquestioninitiatesalong-overdueconversationaboutthe foundationsof social science andthe meanings ofthe terms we use to describe what we do. Conversations about "What is a case?" The conversation about the term "case" presented in this volume had its originsinotherconversations.Theissueofcasescameupoftenina logic-of-analysisworkshop that HowardBecker andI conducted in the winterandspringof1988. This particularworkshophadits origins in still otherexchanges, basedon oursharedreactionto theunexamined status of the case in social science methodology. Thispeculiarstatusofthe"case"wascleartomeinmywork The Comparative Method(Ragin1987). InthatworkI showedhowconven-tionalvariable-orientedcomparativework(e.g.,quantitativecross-national research), as comparedwith case-orientedcomparativework, Introduction: Cases of "Whatis a case?"5 disembodies and obscures cases. In most variable-orientedwork, inves-tigators begin by defining the problem in a way that allows examination of many cases (conceived as empirical units or observations); then they specifytherelevantvariables,matchedtotheoreticalconcepts;and finallythey collect informationon these variables, usually one variable at a time -notone case at atime. Fromthatpointon, the language of variables and the relations among them dominate the research process. The resulting understandingof these relations is shapedbyexamining patternsofcovariationinthedataset, observedandaveragedacross manycases,notbystudyinghowdifferentfeaturesorcausesfitto-gether in individual cases. Thealternative,case-orientedapproachplacescases, notvariables, centerstage.Butwhatisacase?Comparativesocialsciencehasa ready-made, conventionalized answer to this question: Boundaries around places andtime periodsdefinecases (e.g., Italy afterWorldWar II). In comparativeandhistoricalsocialscience,thereisalongtraditionof studyingindividualcountriesorsetsoftheoreticallyorempirically related countries conceivedas comparable cases. The conventionalized nature of the answerin macrosocial inquiry made it simple tocontrast variable-orientedandcase-orientedapproaches.It couldjustaseasily be argued, however, that not countries but rather parallel andcontrast-ing event sequences are cases (see Abbott's contribution, Chapter2), or thatgenericmacrosocialprocesses,orhistoricaloutcomes,ormacro-level narrativesare cases. "Whatis a case?" is problematic evenwhere researchersareconfrontedateveryturnbybig,enduring,formally constituted macrosocial units such as countries. Theproblemof"Whatisacase?"isevenmorecrucialwhenthe contrast between variable-oriented and case-oriented approaches is trans-ferredtootherresearchdomains,becauseinmostresearchareasthe answers are less conventionalized. Is a social class a case or a variable? (See Piatt'sdiscussioninChapter1.)Thisisnotatrivialquestionfor scholars interestedin social movements andthe futureof inequality. Is an analysis of United States census data a study of many cases (individ-uals)oronecase(theUnitedStates)?AsIpushedmyideasabout case-orientedresearchinto new substantiveareas, I foundthatI hada lot ofnewquestionsaboutcases. The logic-of-analysisworkshoppro-vided a good setting for exploring these questions. HowardBeckerbroughtmanyquestionsaboutcasestothework-shop, too. His concerns overlap with mine, but also differqualitatively (see Chapter 9). In the workshop, and later in the symposium where the essays in this volume were first presented, he persistently pulled the rug outfromunderanypossible consensusabout"Whatis a case?" From 6WHATISACASE? his perspective,to beginresearchwithaconfidentnotionof"Whatis a case?"(or, moreprecisely, whatthis -theresearchsubject-is acaseof) iscounterproductive.Strongpreconceptionsarelikelytohampercon-ceptualdevelopment.Researchersprobablywillnotknowwhattheir cases are untiltheresearch,includingthetaskof writinguptheresults, is virtuallycompleted.Whatit is a case of will coalescegradually,some-timescatalytically,andthefinalrealizationofthecase'snaturemaybe themostimportantpartofthe interactionbetweenideasandevidence. Inshort,Beckerwantedtomakeresearcherscontinuallyaskthe question"Whatisthisacaseof?"Thelesssurethatresearchersareof theiranswers,thebettertheirresearchmaybe.Fromthisperspective, nodefinitiveanswerto the question"Whatis a case?" canorshouldbe given,especiallynotattheoutset,becauseitdepends.Thequestion shouldbeaskedagainandagain,andresearchersshouldtreatany answertothequestionastentativeandspecifictotheevidenceand issuesathand.Workingthroughtherelationofideastoevidencean-swersthe question"Whatis this a caseof?" Thus, while I hopedthe workshopwouldbring the start of ananswer tomyquestionsaboutcases, Becker,myco-conspirator,hopedtokeep thequestiononthefloor,unanswered.Thesecontrastingorientations made for a lively workshop, withsome participants sharing myconcern foranswers,howevertentativeandincomplete,andotherssharing Becker'sconcernforkeepingthequestionalive.Still,weallleftthe workshopwithastrongsenseofunfinishedbusiness.Thissensethat therewasmuchmorethatcouldbeminedfromthetopicwasour primarymotivationfororganizinga symposiumonthetopic. We decidedto keep the conversationaboutcases going by posingthe questionto aselectgroupofeightsocialscientists inthefallof1989, to befollowedbyasymposiumonthetopicinthefollowingspring.Our primaryselectionprincipleswerediversityandoriginality.Convincing eightdiversesocialscientiststocometoNorthwesternUniversityand offertheirthoughtson"Whatisacase?"waseasierthanwehad anticipated.Potentialparticipantswereeagertotakeacrackatour question.Asalurewesuggestedthefollowingtopicstoourpartici-pantsas possibleissuesfordiscussion: 1.Alternative definitionsof cases, of case study, and of case analysis. 2.Thecontrastbetweenobservationalandexplanatoryunitsandthe implications of this distinction for research findings as representations of social life. 3.The hierarchicalnature of units andthe implicationsof thisstructure for case analysis and the goal of generalization. 4.The place of theoretical andpurposive sampling in social science and the relation between these sampling strategies and case study. Introduction:Cases of "Whatis a case?"7 5.Therelationshipbetweenanalysisofcasesandanalysisofresearch literatures. 6.The differentuses of case studies in social science. 7.The boundary between social science and other forms of discourse and the place of the concept of the case in supporting this boundary. 8.What is a good case study? Ourlureproducedacollectionofcontrastinganswerstothequestion "What is a case?" and an avalanche of new questions as well. Duringthe two-dayconference,whichinvolvednotonlytheinvitedscholarsbut alsomanyoftheoriginalworkshopparticipantsandalivelygroupof studentsandfacultyfromNorthwesternandotheruniversitiesinthe Chicago area, it was difficultto separate ourquestions aboutcasesfrom awidearrayofissuesconcerningthefoundationofmodernsocial sciencemethodologies. Thepresentcollectionofessaysincludesalleightpreparedspecific-allyforthesymposium.However,theessayshavebeenmodifiedin responsetothediscussionatthesymposiumandinresponsetoeach other as well. Starting pointsfor answering"What is a case?" Beforediscussingthedifferentresponses,itisimportanttonotethat noneofthesymposiumparticipantsofferedwhatmightbeconsidered conventionalanswerstothequestion.Forexample,noonepushed methodologicalindividualismortheideathatsociallifecanbeunder-stoodonlyfromthe perspectiveof individualactors. Nordidany ofthe participantsattempttodefendtextbooktreatmentsofcases -theidea thattherearepopulationsofcases(observations)"outthere"waiting forsocialscientificanalysis.(However,acceptanceofthispositionis impliedinLieberson'sessay;seeChapter4.)Nordidanyoneendorse the idea that the definitionof a set or populationof cases was thepurely practicaltask,specifictoeachresearchendeavor,ofdefiningtheuni-verse ofpossiblyrelevantorcomparableobservations. Whensampling cameupfordiscussionatthesymposium,theprimaryfocuswason theoreticalorpurposivesampling,notrandomsamplingfromapopu-lation. Correspondingly,there waslittle discussionofissues inestimat-ingpopulationparameters.Theconceptof"thecase"is logicallyprior bothtotheconceptof"thepopulation"andtheconceptof"thevari-able." In a contextwherethe conceptof"the case" is madeproblematic, these other concepts appearimpenetrable. Whiletheanswersto"Whatis a case?"werediverse, theydisplayed commonthemes. Participantsagreedthatthe preceptof case analysisis 8WHATisACASE? fundamentaltothe conductofsocialscience andthatit has aspecial, unexamined status. They agreed that individual social scientists answer thequestion"Whatis acase?" inremarkablydifferentwaysandthat answers to this question affect the conduct and results of research. And all agreed that cases may be multiple in a given piece of research: What the case is maychange bothin the hands ofthe researcher(duringthe courseoftheresearchandwhentheresultsare presented)andinthe hands of the researcher's audiences. This general agreementon the importance and indeterminatenature of the term"case" shouldnot be taken as evidence thatstrikingdiffer-ences in emphases do notexist amongthe eightresponses. In fact,the differencesaredramatic.Atthemostgenerallevel,thecontributions differin whether the question stimulateda critique of currentpractices or a reflectionon researchexperiences. Fourof the eightcontributions focusoncritiquesofcommonpractices,andfourfocusontheirown experiences. Crosscuttingthis descriptive dichotomy, however, are more fundamentaldifferenceswhichreflectdifferentstartingpointsforan-sweringthequestion.Tounderstandthesedifferentstartingpoints, consider two key dichotomiesin how cases are conceived: (1) whether they are seen as involving empirical units or theoretical constructs and (2) whether these, in turn, are understoodas general orspecific. The first dichotomy (whether the question of cases involves empirical unitsortheoreticalcategories)is commonindiscussionsofsocial sci-ence methodologyandoverlapswiththe philosophicaldistinction be-tweenrealismandnominalism.Realistsbelievethattherearecases (moreorlessempiricallyverifiableassuch)"outthere."Nominalists thinkcasesaretheoreticalconstructsthatexistprimarilytoservethe interests of investigators. A realistsees cases as eithergiven or empiri-cally discoverable. A nominalist sees cases as the consequences of theo-ries or of conventions. The second dichotomy concerns the generality of case categories. Are casedesignationsspecific(e.g., the"authoritarianpersonality"orthe "anti-neocolonialrevolution")anddevelopedin the course of research (e.g.,throughin-depthinterviewsorhistoricalresearch)orarethey general(e.g., individuals, families, cities, firms)andrelativelyexternal to the conduct of research? In many areas of research, generic units are conventionallytreatedascases, andcase categoriesareneitherfound nor derivedin the course ofresearch.They existpriorto researchand are collectivelyrecognizedasvalidunitsbyatleastasubsetofsocial scientists. Specific case categories, by contrast, emerge or are delineated in the course of the research itself. Whatthe researchsubjectis a"case of"maynotbeknownuntilaftermostoftheempiricalpartofthe Introduction: Cases of "Whatis a easel"9 Table I.i. Conceptual map for answers to "What is a case?" TT.,,.Case conceptions Understanding of casesSpecificGeneral As empirical units1.Cases are found2.Cases are objects (Harper)(Vaughan) As theoretical3.Cases are made4.Cases are constructs(Wieviorka)conventions (Piatt) projectis completed.To a limitedextent,this seconddichotomyover-laps with the qualitative-quantitative divide in social science. The cases of quantitative research tendto exist as conventionalized, generic cate-gories independent of any particular research effort. The cases of quali-tative research tend to coalesce as specific categories in the course of the research. "What is this -the researchsubject -a case of?" is aquestion that is best asked in qualitative social science. The cross-tabulationof these twodichotomies(Table Li) yieldsfour possiblestartingpointsforansweringthequestion"Whatis a case?" Considerthe nature of"cases" fromthe perspectiveof each cell ofthe cross-tabulation: Cell 1: Cases are found.In the first quadrant, researchers see cases as empirically real and bounded, but specific. They must be identifiedand established as cases in the course of the research process. A researcher may believe that "world systems" (networks of interacting and interdependent human societies) are fundamentallyimportant empirical units for under-standing the history of human social organization and therefore may seek to determine the empirical boundaries of various historical world systems (verifiable, e.g., through evidence of trade in bulk goods between peoples ofdifferingcultures).Researcherswhoapproachcasesinthiswaysee assessmentofthe empirical boundingof cases as anintegral partofthe research process. Among the eight contributions, the clearest advocate of this view of cases is Douglas Harper (Chapter 6). Harper makes the empir-ical unit "community" problematic and attempts to delineate communities inductively,throughindividuals.Communitiesare boundedindifferent ways depending on their nature, and the boundary of a single community may be fluid and ever-changing. Cell 2: Cases are objects.In the second quadrant, researchers also view cases as empiricallyreal andbounded, butfeel no needtoverify i oWHATisACASE? theirexistenceor establishtheirempiricalboundariesin the course of theresearchprocess, becausecasesaregeneralandconventionalized. These researchers usually base their case designations on existingdefi-nitionspresentinresearchliteratures.Aresearcherinterestedinex-plainingcontemporaryinternationalinequalityforexample,would accept nation-states (as conventionally defined) as appropriate cases for hisorheranalysis.Oftencoupledwiththisviewisaninstrumental attitudetowardcases -theyexistto bemanipulatedbyinvestigators. DianeVaughan'scontribution(Chapter8)offersthebestexampleof this approach. Her empirical cases are conventional units such as orga-nizationsandfamilies. She arguesthatby exploringgenericprocesses (e.g., misconduct) across differenttypes of generic empirical units (e.g., familiesandformalorganizations),itispossibletodevelopbetter theories. Cell 3: Cases are made. Researchers in this quadrantsee cases as specifictheoreticalconstructswhichcoalesceinthecourseofthere-search.Neitherempiricalnorgiven,theyaregraduallyimposedon empirical evidenceas they take shape in the course ofthe research. A cell-3 investigatorinterestedin tyranny, forexample, wouldstudymany possible instances of tyranny. This investigationmight leadto aniden-tificationof an importantsubset of instances with many commonchar-acteristics, which might be conceived, in turn, as cases of the same thing (e.g., ascasesof"patrimonialpraetorianism"orascasesof"modern tyranny"). Interaction between ideas and evidence results in a progres-sive refinementof the case conceivedas a theoreticalconstruct. Atthe start of the research, it maynot be at all clear thata case can or will be discerned. Constructing cases does not entail determining their empiri-cal limits, asincelli,butratherpinpointingandthendemonstrating theirtheoreticalsignificance.MichelWieviorka'scontribution(Chap-ter 7) offersa good example of this understanding of cases (as does John Walton's; see Chapter5). Wieviorka'sessay shows howtheinteraction of ideas andevidence in his researchon terrorismmade it possiblefor him to identifyits sociologically decisive features. Cell 4: Cases are conventions.Finally, in the fourthquadrant, re-searchersseecasesasgeneraltheoreticalconstructs,butnevertheless viewtheseconstructionsastheproductsofcollectivescholarlywork andinteractionandthereforeasexternaltoanyparticularresearch effort.A researcher, for example, might conduct research on"industrial societies,"recognizingthattheassignmentofempiricalcasestothis theoreticalcategoryisproblematicandthatthetheoreticalcategory Introduction: Cases of "Whatis a case?"11 itselfexistsprimarilybecauseofcollectivescholarlyinterest.Inthis view,casesaregeneraltheoreticalconstructsthatstructurewaysof seeing social life and doing social science. They are the collective prod-ucts of the social scientific community and thus shape and constrain the practiceofsocialscience.ThisviewofcasesisthebasisforJennifer Platf s contribution(Chapter1). Amongotherthings, sheshowshow the cases ofa givenstudymayshiftovertime as intellectualfashions change insocialscience andpastworkis selectivelyreconstructedby the social scientificcommunity. This fourfolddivision of case conceptions is not absolute. A researcher couldbothuseconventionalizedempiricalunits,acceptingthemas empirically valid (cell 2), and try to generate new theoretical categories or case constructs(cell 3) inthecourse ofhis orherresearch.Frustra-tions with conventional case definitions and practices (cell 4) could lead researchersto intensifytheirempiricaleffortsandtodefinecasesand theirboundariesinamoreinductivemanner(cell1).Infact,most research involves multiple uses of cases, as specific or general theoreti-cal categories and as specific or general empirical units. These multiple uses occur because researchcombines theoreticalandempiricalanaly-sis, andthe two kinds of analyses neednot use parallelcases or units. The pointof Table Liis notto establishboundaries betweendifferent kinds of research, but to establish a conceptual map for linkingdifferent approaches to the question of cases. The eight answers As noted, independentof starting point, the contributions split equally into two main groups. The firstfourare critiques of conventional prac-tices. The second four are analyses of research experiences. JenniferPiatt(Chapter1) searchesbroadexpansesoftheterrainof social research in her effortto explore the diverse ways sociologists use the term"case." This sweep includes both qualitative andquantitative empirical research from the past and present. She uncoverssurprisingly little consistency. Both empirical and theoretical cases are multiple within mostresearchefforts,andinvestigatorsonlyoccasionallyseemcon-cernedtomatchtheoreticalcasesandempiricalcases.Forexample, muchtheorizingaboutsocialclassesascaseshasoccurredoverthe historyofsociologyandpoliticalscience.Yetmanyrecenteffortsto studyclassesempiricallyusesurveydatafromsamplesofresidents (oftenmales only) andinferclasses andtheircharacteristicsby aggre-gatingthecharacteristicsofindividuals.Thedistancebetweenthese artificial statistical constructions and the theoretical categories are obvi-12WHATisACASE? ouslygreat,especiallywhenviewedfromtheMarxistperspectiveof classes as historical actors. Yet this way of studying classes has become conventionalized(Piatt1984).Anotherconfoundingfactorinsocial scientists' uses of cases is whether investigators see the research setting itselfas a meaningfulhistorical case (e.g., Great Britain in the1970s) or asjustoneamongmanypossibleequivalentsettingsforresearch(a postindustrialsociety).Piattalsonotesthatinmanystudiescrucial arguments dependon evidence about other cases in other studies; thus the evidence used to supporta conclusion may be secondhandor even thirdhand.Researchersmayconstructargumentsfromcontrastsbe-tweentheirowncases andthoseofotherresearchers, evenwhenthis contrast involves using secondary cases in ways that conflictwith their original uses. This feature of social research underscores the communal nature of case use in social science. AndrewAbbott'scritiqueofconventionalpractices(Chapter2)fo-cuses onthe tendencyformostsocialscientiststoconflatethedichot-omy of "population" (or large N) versus "case study" with the dichotomy of"causalanalysis"versus"narrativeaccount."Hearguesthatsocial scientistsshouldconductnarrativeanalysisacrossmanycases.After dissecting several"population/analytic"studies (large-N, variable-ori-entedinvestigations), Abbottconcludesthatthese conventionalforms ofanalysiscannotsystematicallyaddress action,agency, andcomplex eventsequences. Thesestudiesdescribecasesasactingonlyincrude and ad hoc ways. For example, some version of the rational actor model may be invoked to explain an anomalous statistical relationship. Abbott contendsthatinvestigatorsshouldask"Whatdocasesdo?"firstand thatnarratives,ascases,aretheappropriateunits.Inductively,the investigatorconstructs narrativeaccounts andexplanations fromevents, whichinturnarefoundincolligationsofoccurrencesandotherevi-dence. Thus, narratives can be discerned in specific empirical evidence. The end product is not a mere collection of narratives, however. Abbott points to the possibility of building generic narrative steps andgeneric plots from the events and sequences that make up individual narratives [see also Abbott and Hrycak (1990) on patterns and sequences of events]. Harrison White (Chapter 3) bases his analysis of social scientists' use of cases on an examination of "worldly" conventions in their use -how nonscientists use them. He findsthree basic worldlyuses: (1) to estab-lishidentity,(2)toexplainorresolvebyinvokinggeneralprinciples, and(3) to accountforwhyevents unfoldin one wayandnotanother, with the idea that such knowledge can be usedto control situations or tofixtheminsomeway.Thesedifferentworldlyusesofcasesare paralleled in differentkinds of social scientific work. An example of the Introduction: Cases of "What is a case?"13 firsttype is Immanuel Wallerstein's Modem World System (1974), which establishestheworldcapitalistsystemas asingularandfundamental unitforsocialscientificthinking.Anexampleofthesecondtypeis BarringtonMoore'sSocialOriginsof Dictatorship and Democracy(1966), which accounts for a range of political outcomes with a single explana-tory framework.Jane Mansbridge'sWhy We Lostthe ERA (1986)offers anexcellentexampleofthethirdtype. Whiteshowsthatthereisno simple correspondence between these differentgoals in the use of cases andthe kinds ofevidenceor types ofempiricalunitsusedin astudy. Mansbridge'sbook,forexample,aclassic"control"casestudyin White'sframework,includessurveydataonindividualsshowing broad-basedsupport for the equal-rights amendment and the ideas it embodied. White argues forcefullythat scholarly recognition of these worldlyconventionsin the use of cases wouldimprovesocialscien-tificthinking. The relative utility of differenttypes of empirical units forformulat-ingortestingcausalargumentsisStanleyLieberson'skeyconcern (Chapter4).Heistroubledbywhatmightbecalledpseudo-causal analysis (Mill's method of agreement and his indirect method ofdiffer-ence) mechanically applied to small numbers of cases. Researchers work-ing with small N's have argued that Mill's methods permit rudimentary causalanalysis[see,e.g.,theexchangebetweenNichols(1986)and Skocpol (1986)]. Lieberson disagrees and frameshis contribution to this volume as a cautionary tale: Although it may be temptingto do causal analysisofsmallN'susingMill'smethods,theendproductwillbe a seductivelydeterministic,andprobablyfaulty,causalgeneralization. Liebersonattemptsto demonstratethat whenN's are small, the possi-bility of formulating causal statements that are both general and reason-able (e.g., "state breakdownis acause ofsocial revolution")isgreatly diminished.Theproblemsthatcropupresemblethosethatoccurin quantitative analysis when researchers attempt to maximize the propor-tion of explained variation(Lieberson1985). While this cautionarytale bypassesdiscussionofthewaysinwhichMill'smethodshavebeen superseded(e.g., Ragin1987:85-123), it is significantbecause much of thediscussionof"Whatis acase?"focusesonsmallN's.Lieberson's implicitmessageisthatinvestigatorswhowanttomakevalidcausal generalizations should use generic empirical units and conduct large-N, variable-oriented investigations. Lieberson illustrates his arguments not through analyses of examples of small-N research, but by showing what might happen if these methods were applied mechanically to artificially constructed,generic empiricalunits -automobileaccidentscontrasted with nonaccidents. 14WHATisACASE? JohnWalton(Chapter5)arguesthatcasesaremadebyinvoking theory. The process of justifyinga case -as a case of somethingimport-ant-involvesshowingthatthecasebelongstoaspecificfamilyof phenomena. This family, in turn, is important because of its relevance to general social scientific thought ("theory"). This theoreticallygrounded characterofcasesexplainsbothwhytheyarecentraltotheadvance-ment of theory -why case studies appear prominently in the history of social thought -andwhy cases can be made andremade as newtheo-ries are appliedto knowncases (a pointalso made by JenniferPiatt in Chapter1).Whatcasesare"casesof"maychangeasourtheories change.Todemonstratehisarguments,Waltondescribeshow"the case" shiftedin his own study(Walton1991) of"California'slittle civil war," an episode of conflict between the residents of Owens Valley and the City ofLos Angeles over rights to the valley's water. The case had obvioushistoricalsignificance,butitssociologicalsignificancewasat firstelusive.Movingbackandforthbetweentheoreticalideasand historical evidence, Walton eventually founda suitable theoretical con-text for framinghis case sociologically. DouglasHarper'scontribution(Chapter6) offersa clear example of howcases areconceivedinmuchofqualitativesociology. Heplumbs the boundaries of communities through intensive studyof individuals. In effect,the individualprovidesa windowonthecommunity.Inthe processoffindingcommunitiesthroughindividuals, empiricallimits areestablished.Forexample,throughethnographicinvestigationof "Willie," a ruralhandyman, Harper(1987) unearthsa complexweb of formalandinformalexchanges andinterdependences.Willie'sskill in makingrepairs andcreatingusefulobjectsoutofrefuseis a feature of thiscommunity,notsimplyanaspectofWillie(i.e., Willie'shuman-capitalendowment).Harper'sanswerto"Whatis acase?"argues, in essence, thatcasescan be foundinductively,piecedtogetherfromthe livedexperiencesofindividuals.Whencollaboratingwithresearchers who see communities as givens (i.e., definedby formal political bound-aries or census tracts), Harper chafes and struggles because he feels that animportantpartoftheresearch -findinganddelimitingcases -has been assumed away. Michel Wieviorka opens his discussion(Chapter 7) by examining the factorsthatmake somethinga "goodcase," focusingfirston the pecu-liar status of the good case in medicine. A medical case is "good"when itisbothrareanddiagnosable;itembracesbothanindividual,the patient, and an important or new category in the professionalliterature. Wieviorkathenmovestothecontrastbetweengoodcasesinhistory andsociology.Whatmakesacasegoodinhistoryoftendiffersfrom Introduction: Cases of "Whatis a case?"15 what makes it good in sociology, even though the two approaches may be applied to the same historical facts. A historical case is good because ofitssignificanceforsubsequentevents;asociologicalcaseisgood because of its theoretical significance. These two don'talways go hand inhand,however;thesociologicallydecisiveaspectsofthecasemay not be relevant to questions about why the case followedone historical pathandnotanother.To illustratehisarguments, Wieviorkapresents hisownworkonterroristgroups, wherehisprimaryconcernwasto identifytheirtheoreticallydecisivecharacter(i.e., afeaturethatwas general to these groups andthat clearly differentiatedthem fromother groups)(see Wieviorka1988). This search ledto the developmentof a new theoretical case or construct, which in turn allowed him todifferen-tiate among terrorist groups. As Wieviorka explains, some groups which originally had been classifiedas terroristin journalistic accounts could now be seen to differfromterroristgroupsmarkedbythe presence of theoretically decisive features. Thus, the end result of empirical research for Wieviorka is a new or refinedtheoretical case. Diane Vaughan's contribution (Chapter 8) focuses on general empirical unitsas cases -families,organizations, nation-states, andso on. But she seesinthisdiversityofempiricalunitsagreatpossibilityforstudying specific processes in vastly differenttypes of settings. She opens by noting that the typical academic career requires social scientists to know more and more aboutless andless andthatthis specializationalso oftenentails a focus restricted to a single empirical unit (e.g., the family). This restriction impedes theorydevelopmentandelaboration because manyof the phe-nomena that interest social scientists and their audiences appear in many differenttypes of empirical units, at various levels of complexity and size. Patterns observed studying a phenomenon in small units (e.g., misconduct in families) can lead to theoretical and analytic insights in the study of the same phenomenon in larger units (e.g., organizations). This creative sym-biosis is further strengthened because different kinds of evidence are avail-able in differenttypes of units. For example, it is difficultto do an in-depth interview of a formal organization, but possible to do so for the members of a family [as in Vaughan's study (1986) of couple breakups using princi-ples from organizational theory]. Formal organizations, by contrast, leave manywrittenrecordsoftheirday-to-dayoperations; families,relatively few. Thus, Vaughan's answer to "What is a case?" celebrates the diversity of generic empirical units and the many opportunities for social scientists this affords. 16WHATisACASE? Lookingahead Oneofourauthorscommentedatthesymposiumthatthequestion "Whatis a case?" waslike a Rorschachtest andcapableofproducinga varietyofresponsesfromsocialscientists,evenlike-mindedones.Itis truethatthe questionproducesdiverseresponses. It is trueas wellthat thequestionisaprism.AsBeckershowsinhisreflectionsonthe contributions(Chapter9),itispossibletoseethepracticesofsocial science in new ways throughthis prism. The issue of cases touches basic questionsinhowwe,associalscientists,produceresultsandseemto know whatwe know. The essays in this collection emanatefromtheuse of this prismin diverserealmsofsocial scientificpractice. In many respects"What is a case?" is a conversationthat for us hasno real beginningorend.But wealso feelthatin somerespects it hasbeen amissingconversationinthesocialsciences, becausealltoooftenthe "case"is abasic, taken-for-grantedfeatureofsocialscienceresearch.It isimportanttoexaminetaken-for-grantedfeaturesbecausetheylimit our understandingandvision bothof social life andof social science. In thissense,"Whatisacase?"isonequestionamongmanyothers(e.g., "Whatisapopulation?"or"Whatisavariable?")waitingforserious attention.We hopethatourprojecthasgiventhisneededconversation newlifeandthatthiscollectionwillstimulateneweffortsbothto answer"Whatis acase?" andto askother basic questionsabouttaken-for-grantedelementsofsocial scienceresearch. References Abbott,Andrew,andAlexanderHrycak(1990)."MeasuringResemblancein Sequence Data." American Journal of Sociology 96:144-85. Feagin, Joe R., Anthony M. Orum, and Gideon Sjoberg (1991). A Case for the Case Study. Chapel Hill: University of North Carolina Press. Gubrium,JaberR,andJamesA.Holstein(1990).What Is Family?Mountain View, CA: MayfieldPublishing Co. Harper, Douglas (1987). Working Knowledge: Skill and Community in a Small Shop. Chicago: University of Chicago Press. Lieberson, Stanley (1985). Making It Count: The Improvement of Social Research and Theory. Berkeley: University of CaliforniaPress. Mansbridge, Jane (1986). Why We Lost the ERA. Chicago: University of Chicago Press. Moore, Barrington, Jr. (1966). SocialOriginsof Dictatorshipand Democracy:Lord and Peasant in the Making of the ModernWorld. Boston: Beacon Press. Nichols, Elizabeth (1986). "Skocpol and Revolution: Comparative Analysis Ver-sus Historical Conjuncture/'Comparative Social Research 9:163-86. Introduction: Cases of "Whatis a case?"17 Piatt, Jennifer(1984). "The Affluent Worker Revisited/'pp. 179-98 in C. Bell and H. Roberts (eds.), Social Researching: Politics, Problems, Practice. London: Routledge & Kegan Paul. Ragin, Charles C. (1987). The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. Berkeley: University of CaliforniaPress. (1991)."Introduction:theProblemofBalancingDiscourseonCasesand Variables in Comparative Social Research/' pp. 1-8 in Charles C. Ragin (ed.), Issues and Alternatives in Comparative Social Research.Leiden: E. J. Brill. Sjoberg, Gideon, Norma Williams, Ted R. Vaughan, and Andree Sjoberg (1991). "The Case Approach in Social Research: Basic Methodological Issues," pp.27-79mJoeR- Feagin,AnthonyM. Orum,andGideonSjoberg (eds.),ACaseforthe CaseStudy.ChapelHill:UniversityofNorth Carolina Press. Skocpol, Theda(1986). "AnalyzingConfigurationsinHistory:A Rejoinderto Nichols." Comparative Social Research 9:187-94. Vaughan, Diane (1986). Uncoupling: Turning Pointsin Intimate Relationships. Ox-ford: Oxford University Press. Wallerstein,Immanuel(1974). The Modern WorldSystem:CapitalistAgriculture and the Origins of the EuropeanWorld Economy in the SixteenthCentury. New York: Academic Press. Walton, John (1991). Western Times and Water Wars: State, Culture, and Rebellion in California. Berkeley: University of CaliforniaPress. Wieviorka, Michel (1988). Societes et terrorisme. Paris: Fayard. Parti Critiques of conventionalpractices 1 Cases of cases . . .of cases JENNIFERPLATT This chapter's broad concern is with the ways cases are used in practice to build arguments, and how this relates to conventionalmethodologi-calimperatives.Thewayscasesarechosen,analyzed,amalgamated, generalized,andpresentedareallpartoftheiruseinargument.Itis assumed that an argument is designedto reach a conclusion which the reader (and the writer) will find convincing. It is thus always relevant to consider the intendedaudience, and the use of cases may be treatedas partofawork'srhetoric.1Thisessayfollowstheimpliedthemesor questionsthroughaseriesofexamples,chosentoprovidediversity along relevant dimensions. The more specific issues raised emerge from close considerationof what is done in the books analyzed.2 In the light of what is found, the ends are pulled together into some general ideas. We lookfirstatsome workswhichareinrelativelyobvioussenses case studies, whether or not their authors describedthem as such. The Jack-Roller In The Jack-Roller (1930) Shaw gives extremely intensive data on just one individual, a juvenile delinquent; it presents his lifehistory writtenby himself, as well as a varietyof material abouthim fromothersources. The individual it studies is not well known or historically significant; it is clearthathe is quite like a lot ofotheryoung menat the sametime andplace. Thisimpliessensesinwhichhecanbetakenasacaseof something,andShawclearlyintendshimasacase ofayoungdelin-quent. It is interesting, however, that he is also a case in a quitedifferent sense: acasefortreatment,aproblem -andoneonwhom,inconse-quence, a lot of material has already been compiled by caseworkers and othersemployedbyvarioussocial agencies. (The instancethusunites two featuresof the interwar"case-study method," one definitionaland theothermerelyempiricallyverycommon: richqualitativedata,and 21 22WHATisACASE? the use by sociologists of data collectedby social workers.) Such agen-cies are by necessity, however large the numbers they process, forcedto concern themselves with particular individuals, since they are responsi-ble for theirtreatment. Shaw'sbookis alsonotableina quitedifferentwayforits useofa singlecase.Thecaseof"Stanley"wasonlyoneof200similarcase studieswhichShawhad,andhealsohadaprogramofquantitative research on delinquency. Shaw declares that the purpose ofpublishing this study was to "illustrate the value of the 'own story' in the study and treatmentofthedelinquentchild"(1930:1). Thusthecaseisoneofa treatment,orofamethodofelicitingdataneededfortreatmentpur-poses, as muchas of a person. As Becker points out in his introduction to the 1966 paperbackedition, in its original time andplace it waspart of a larger program which gave it a context of other sorts of data, and so needs to be understood as such. Some of the relevant data were in other studiesoftheChicagoarea,carriedoutbyotherresearchers.Shaw himself, justifyingthe use oflife-historydata, saysthatthey"afforda basisfortheformulationofhypotheseswithreferencetothecausal factorsinvolved.. . . Thevalidityofthesehypothesesmayinturnbe tested by the comparative studyof other detailedcase histories andby formalmethodsofstatisticalanalysis"(1930:19). Insofaras thatuseis the one envisaged, the functionsperformedby the case cannot be seen withinthecoversofShaw'sbook,butonlybylookingatthewider program. Inanothersense,too,thebookwaspartofawiderprogram,a programof social reform.Bennett(1981) has arguedconvincinglythat lifehistoriesof juveniledelinquentshave beenproducedundersocial circumstances where reformersneeded them to address particular con-stituencies. Life historiesmakedelinquentsvisible to middle-classpo-tentialvolunteersandphilanthropists,underminetranscendentalor physical theories of delinquency, and persuade readers who are already interested but not professionalsin the field. Shaw's life histories served torecruitsupportersfortheChicagoAreaProjectinwhichhewasa central figure. The Family Encounters the Depression One of the most sophisticated instances of deliberate use of"case-study method"is The Family Encountersthe Depression(Angell1936). Angell's dataweredocumentssolicitedfromstudentsathisuniversitywhose families hadsuffereda loss of at least25% of their income as a result of the Depression. Loss of income in the Depression was chosen as a case Cases of cases. . .of cases23 ofanexternalfactorimpinginguponsocialgroupsofagivenkind whose adjustmentto it could be studied. Although each documentwas writtenbyonemember,the"case"'wasthefamilydescribed,which was characterizedas a unit. Angell explicitly statedthat"our series of cases is not in any sense a representative sample, even of the families of college students" (1936:264), andthatthis does not matter, because the aim is analytic rather than enumerative induction. This "seeks to isolate distinctivetypes, eachofwhichhasitscharacteristicmodeofadjust-ment, so that whenone findsa familyof a certain sort, one canpredict howitwilladjust"(1936:7).However,thenumberofcasesservesa qualitativefunction,ifnota quantitativeone: "A goodmanyfamilies have to be studiedin order to be sure to findat least one example each of the importanttypes. If possible, it is desirable to havecorroborative examples of each one in addition" (1936:7). This is a very interesting strategy, but it has some obviousdifficulties. If there are "distinctive types," a single case differentfromthe others so far foundwill sufficeto exemplifya type; it is not clear, though, why it should be taken for granted that all families of any one type willadjust in the same way, unless the mode of adjustmentis built into itsdefini-tion. The referenceto"corroborativeexamples" is inthis contextpuz-zling, since it is notclearwhattheywouldbe for;this soundslike an inconsistenttraceofaquitedifferentsetofintellectualassumptions. Secondly, theneedto studya numberoffamiliesinevitablyraisesthe question of how one knows when one has studiedenough; that line of questioning seems to lead inexorablyto the suggestionthat one would be wise to maximize diversity in the cases used, and that drawing all of them fromstudents at one universityis not, primafacie, very likelyto achieve that. The descriptionso fargives verylittleindicationofthestyleofthe book,whichwasprobablywrittenasitwaspartlybecauseageneral nonprofessionalreadership was expected.Mostofthemethodological discussion is in an appendix, and there Angell describes how he strug-gledto make sense of the data, trying outdifferentvariablestodefine types until he hadfoundones which seemedto him to dealsatisfacto-rily withevery case. There is no formaloperationalizationofthe vari-ables -andthe nature of the material, even though Angell didgive his students quite detailed instructions, is diverse enough to mean that that wouldhave been very hardto achieve. The main body of the text does not reflect these struggles, or show anything of the process of induction. Mostofitconsistsoflengthydescriptivequotationsfromthefamily accounts, arranged in chapters each of which corresponds to one cell in thetypology.Forpurposesofpresentation,thus,thecasematerialis 24WHATisACASE? arranged as though it were illustrative of the theoretical scheme; it is not clear whetherit is intendedas exampleor ostensivedefinition,or dis-play of data as evidence. Theabsenceofanysystematicmethodofgettingfromindividual casestoconclusionsmeansthattheconvictionofthereaderthatthe conclusion is appropriate must depend either on her own implicit anal-ysis of the cases or on trust in Angell. The latter seems more likely to me, thoughotherreadersmayfinditeasierthanI dotoseeaclearfitto generalideasinthegreatmassofdetailpresented.To theextentthat trust in Angell is crucial, the appeal overtly made is not to authority but to the difficultieshe experienced and his sense of their eventual resolu-tion.However,thefactthatheisanestablishedprofessor,referstoa professionalliterature whichhe has contributedto andwhichappears not to have resolved the intellectual problems at issue, and has a sophis-ticatedmethodologicaldiscussion,presumablyaddstohisauthority andhencetotheweighttobeattachedtohiseventualsatisfaction. Important, too, is the claim that in the end every case has been accounted for; surely a theoretical schema which can achieve that must be valid? I imagineanyonewouldagreethatatheoreticalschemawhichcould achieve that when appliedto a reasonably varied range of cases would be very promising. However, there is an obvious difficultyin addition to the one implied earlier about unambiguous operationalization: in principle, there could be more than one theory which fits all the cases.3 Here it is hard to avoid theissueofqualitativerepresentativenessorrangeoftheparticular cases studied since, prima facie, good fit to a more diverse set of cases is harderto achieve(cf.Polya1968). We maynote thatAngelloriginally intendedhis 50 cases as only an "exploratory" study, to be followedup by a much larger one (1936:271). The larger number of cases was meant toallowfor"verificationandmoredetailedanalysis"(1936:300), and would presumably have met this point; he concludes, however, that "it isdoubtfulwhetheralargersamplewouldyieldenoughadditional knowledge to justifythe effortand expense" (1936:301), though without spelling out the reasons. Therequirementthateverycaseshouldbeaccountedforputsthe maximum emphasis not on the individuality of each, since it is assumed that they will in practice fall into a relatively small number of types, but ontheunimportanceofnumbersofcasestotheoreticalexplanation. Analytical inductionleaves no room forexcuses aboutother variables, or claims which rest only on the proportionof variance accountedfor.4 It is striking that the types Angell eventually arrivedat were definedin termsoftwovariables,eachwithvalueseffectivelyhigh,medium,or Cases of cases. . .of cases25 low; this is a convergence withquantitativestyles whichheraldedthe disappearanceof"case-studymethod"as a recognizedalternative ap-proach. The typological strategy has much in common with that which Lazarsfelddeveloped.Thecasedefinedasapointofintersectionof variables both retains and loses its meaning as a unique historical indi-vidual. Abbott's distinction (Chapter 2) between starting fromvariables andconstructing"cases" frommaineffects,andtakingcasesintheir complexityandthensimplifyingthemdowntokey variables, isvery relevant here. A final feature to note about the book is that Angell chose the substan-tive topic for its methodological interest, to fit the needs of methodolog-ical ideas he hadpreviouslydeveloped.In thatsense it is a case ofthe method. As it happens it started as a case of case-study method and then, because hediscoveredanalyticalinductioninthecourse ofthework, became (also) one of analytical induction. Patterns of Culture Ruth Benedict's Patterns of Culture (1935) is an entirely differentkind of qualitative study. She examines three societies chosen fortheirmarked differencesfromeachother; wholesocietiesandtheirculturesare her cases. Afteran initial chapterin whicha large numberof instances, in thiscaseculturaltraitsratherthanwholesocieties, arementionedin ordertodemonstratetheextremelywiderangeofpossiblehuman diversity,shetreatseachofhermaincasesseparately.Heraimisto show that each culture has an internal consistency based on underlying principles. The cases are notanalyzedin orderto testthishypothesis, norisitpresentedasaconclusionreachedasaresultofexamining them;rather,itistakenfromthestartasapointalreadyestablished, sufficientlyso that no citations or systematic data will need to be given tosupportit.Itisprobablyintendedthatitshouldbeacceptedon authority, the authority(legitimately?)claimed by anexperiencedpro-fessional anthropologist writing a book aimed at nonprofessionals. The bookismeta-anthropology,offeringaninterpretiveaccountofdata collected by others. The approval of the interpretation by these others is invokedto lendfurtherauthoritytotheaccount(1935^), andtothat extenttheweightoftheprofessiongenerallyisputbehindit.The interpretationapprovedis,however,thatofthedataontheseparate cases rather than that of the thesis of the book as a whole. The cases presented have been chosen for a reason additional to their differences from each other. Benedict argues that it is easier to demonstrate the general point by looking at simpler societies; sociallydifferentiated 26WHATisACASE? Western societies are too complex for it to be easy enough in the present stateofknowledgetoidentifytheirunifyingprinciples.Nonetheless, she(1935:39)assertsthat"thisdoesnotmeanthatthefactsandpro-cesses wecandiscoverinthiswayarelimitedintheirapplicationto primitivecivilizations.... Theunderstandingweneedofourown cultural processes can most economically be arrived at by a detour" This confidentassertion is to some extent underminedwhen, toward theendofthebook,Benedict(1935:161)saysthat"ofcourse"notall cultures are equally integrated, andindeedsome may be characterized by lack of integration. If this is accepted, her initial logic of using cases of integratedsimple societies to throwlight on more complexones no longerholds.Now,however,heremphasisseemstohavemoved elsewhere,tothedesirabilityofpassingjudgmentobjectivelyonthe dominanttraitsofourownsocietyandtotheextentoffitbetween personality and culture. The argument on the latter is that personality is highly malleable culturally, but not totally so; cultures thus create devi-anceiftheydonotprovideroleswhichallowforthefullrangeof nonmalleabilityin the personalities of their members. To supportthis conclusion,whatwouldideallybe requiredis cases whichdoanddonotprovideroomfordiversityofpersonalitytype, and this she appears not to have. However, it emerges at this point with specialclaritythatthereisalatentcomparisoncaseintheresearch design, and it is that of contemporary American society. This is seen, on the basisofMiddletown(LyndandLynd1929) andmembers'general knowledge, as not providingsufficientroom fordiversity. The perora-tiondrawsthemoralofthedesirabilityofavoidingextremesandof being tolerant of both individual and culturaldifferences. Inordertoreachthispointshehasmanagedthetask -somewhat awkwardin principle -of both presentingvery alien social patterns as naturalandunderstandableintheircontext,andthereforenottobe criticized fromour cultural perspective, and invoking value judgments aboutcertainpatternsas"psychopathic."(Perhapsitis assumedthat psychiatriccategoriesmaybetakentotranscendculturalrelativity.) Retrospectively,atleast,althoughherearlierlanguagehadalready given us manysignals, we nowsee her cases as chosen at least inpart fortheevaluativeresponsesshehastothem,oranticipatesinher readers: thePuebloIndiansare Good,andtheDobuandKwakiutlin theirdifferentwaysBad.Thebookasawholethususesallegoryin Clifford'ssense(CliffordandMarcus1986).Theparticularwaysin which they are good or bad enable her to make, whether by contrast or byanalogy,thepointsshewishestomakeaboutAmericansociety. It does not seem accidental that the Pueblo case is treated first and in most Cases of cases...of cases27 detail, because to focus attention at once on a case which is both strange andGoodisarhetoricalstrategywhichputsthereaderintheright frame of mind for the total argument. Street Corner Society A verydifferentinstance ofqualitative, case-studyresearchis thatold methodologicalfavoriteStreetCornerSociety(Whyte1943). Whytefo-cusesindetailontwogangs, andmovesoutwardfromthemtotheir local social context. Both in the gangs and in their social context consid-erable detail is given about particular individuals. His initial interest, at that stage very vaguely defined, had been in their area as a slum area; by the end of the book he is reaching conclusions about the area as a whole, thoughnowasacomplexandarticulatedsocialstructure.Hiscases, thus, could be taken to be all of the individuals, the gangs, andCorner-ville as a community. Indeed, one might add the local political machine ornetworktothatlist. Episodes, whetherindividuallike LongJohn's nightmaresorcommunallikethecelebrationofthefeastdaysofthe patron saints, could also be regardedas among the cases studied. The research strategy followedcan hardly be related to a clear initial hypothesis or research question, since there wasn't one, exceptperhaps of a very vague descriptive kind. In this instance there is aparticularly sharpdisjunctionbetweenthewaytheresearchwasapproachedand the way the findingsare eventuallypresented. Cornerville waschosen more or less accidentallyas the areain whichto conducttheresearch. The bookstartswithmaterialaboutthetwogangs, whosestoriesare introduced as showing the range of careers which Cornerville canoffer its young men (but Whyte says its people) and explaining how individ-ualscometofollowdifferentroutes(i943:xviii). Whytesays that"the generalpatternoflifeisimportant,butitcanbeconstructedonly throughobservationoftheindividualswhoseactionsmakeupthat pattern" (i943:xix). He will start with the little guys, move up to the big shots, then look at the structure as a whole. There follows a rich account, both historical and analytical, of what he learned about the gangs. Although one learns a lot about the individuals, and the text is easy to read as just a set of stories, Whyte is consistent in presentinganecdotes as exemplifyingmore general points about the social structure. Some of the general points he makes are backed by more or less explicit quanti-fication,if only of the order of the understoodclaim that he made a lot of observations andso musthave a goodimpressionofwhatgoes on. These have a differentevidential status from those which rest on partic-ularanecdotes,likethatofLongJohn'snightmaresandtheircure, 28WHATisACASE? where the episode is taken as a sign of the more general factorseen as giving rise to it. To the extent that the point is not also supportedmore quantitatively, the process of inference by which it is reached is not one whichrestsontheinformalsamplinglogicoftheothers.Wemay perhaps take it, too, that Whyte's ability to producea numberof anec-dotes with telling details performs the functionof inspiring trust that he hasindeedreallybeenthereandlearnedalot,andsosupportsthe implicitly quantitative part of the logic. The orderinwhichthe cases whichrepresentthe partsofthe struc-ture are presentedcould be takento be justthe orderin whichWhyte learnedaboutthem.However,insofarasonegeneralmessageofthe book is that Cornervilie is a society in its own right, which should not be seen simply as deviant, to putsmall groupsof friendsmoving intheir own social world first, and the gangsters and politicians who are known to andviewedunfavorablyby the larger society last, is aneffectivestrat-egy. This helps to showthe familiarstereotypicallydeviantas contain-ing strange and not unattractiveconformities. Relatedissuesareraisedbythe"sampling"aspectofhischoice of cases on which to make observations. He addressesthis directly inthe appendix(1943:322-3), describinghowhe came to realizethathewas not writing a community studylike Middletown(Lynd andLynd 1929), whichwaswrittenasaboutpeopleingeneralinthatcommunity.He was dealing with particular individuals and groups, but decided that he could say something significantabout the whole of Cornervilie on that basis "if I saw individuals andgroups in terms of their positions in the social structure. I mustalso assumethat, whateverthe individualand group differenceswere, there were basic similarities to be found. Thus I wouldnot have to studyevery corner gang.. . .A studyof one corner gang was not enough, to be sure, but if an examination of several more showed up the uniformitiesthat I expected to find, then this part of the task became manageable" (1943:323). Hismethodologicalpositionthusrestsclearlyonwhatmightbe described either as a theoretical assumptionor as an empirical general-ization;Whytedoesnotdescribetheintellectualroutebywhichhe arrivedatit.Ineffecthiscases, individualsorgroups,aretreatedas unitsinthesocialstructure,anditis takenthatwithinthatstructure there is sufficientuniformityfor systematic sampling to be unnecessary. The existence of the structure is, however, inferredfromobservation of the individuals and groups. They are taken to represent its operations in aqualitativesense, ratherthanthequantitativeoneimpliedbyusual samplingstrategies.Inaddition,theindividualsareineffectusedas informantsabout units not directly observed, so that Whyte gains more Cases of cases. . .of cases29 generalimpressionsofthecommunity,andcancheckhistentative ideas, indirectly; this again implies an assumptionaboutthesufficient, if not the formal, representativenessof the data available. Whyte could besaidtohaveobservedthestructureasdirectlyasoneevercan observe a structure, and in that sense his observational and his theoret-ical unitsare the same evenif he didnotobserve the whole ofit. The logicbywhichhegetsfromobservedtounobservedcasesisthata structure of the kind which his cases fit and constitute could persist only if the unstudied cases also fit it, and would constrain them to do so. Street Corner Society is especially interesting because, since the publi-cation in its secondeditionof the famousmethodologicalappendix, it has become a case of participant observation; indeed, not just a case but an exemplar. It is arguable that, retrospectively, the prime case studied has become, for many readers, that of Whyte: it is his Bildungsroman (as wellas the buddystoryofBill andDoc). Oneway, andarhetorically very effectiveway, of reaching a conclusion andtaking the readerwith youtothatconclusionistotellthestoryofhowyouarrivedthere yourself. Thisalmostcertainlyentails showingthatyouwereinitially wrong or were surprised by what you discovered. This is a verydiffer-ent strategy from the "scientific" one of concealing human agency in the productionofthefindings,andstartingwithahypothesiswhichhas beenconfirmed. Boys inWhite The final"qualitative"' study to be lookedat is Boys in White (Becker et al.1961),alsoofspecialinterestbecauseofitsmethodologicalself-consciousness. This self-consciousnesswas particularly concernedwith theadequacyoftheevidenceprovidedbyparticipantobservationto supportgeneralconclusionsaboutthegroupstudied,withspecial-though not sole -reference to the question of how much evidence there is. Whengeneralstatementsare madeinthe book,therefore,theyare likely to be supported by figures [e.g., "our field notes yield 87 instances of use ofthe perspective"(Becker et al. 1961:309)]. The datapresented include many such instances; they also include some descriptions of the behavior observedin particular episodes, andlengthy quotationsfrom conversation with the medical students studied. But what are the cases? Thestudywasconductedinonemedicalschool,butitwasnota study of the school as such, either in the sense that the researchers had a specialinterestintheUniversityofKansasMedicalSchoolorinthe sense that they tried to study every aspect of it; their interest was in the effectofmedicalschoolonstudents,andtheyconcentratedonthe 30WHATisACASE? students'experience.TheUniversityofKansasMedicalSchoolwas, thus, usedas a case of a medical school. Some attentionis given to the questionofits representativeness,andsomeremarksare madeabout ways in whichit is likely to differfromothermedicalschools, butthe chapterinwhichthisisdoneconcludes:"We writeintheconviction thatthe way in whichthese youngmen developtheirperspectiveson theirpresentandfutureis,initsessentials,likethatinwhichother medicalstudentsdeveloptheirs"(1961:63).Nojustificationforthis conviction is offeredat that point, but the rest of the book makes it clear thatit rests on rationaltheoreticalgroundsratherthanonideasabout sampling. The majorcauses seen as producing the patterns of behavior observed in Kansas are ones seen as inhering in the nature of what it is to be a medical student faced with a body of potential knowledge larger thanthereis anyrealisticprospectoflearningduringthecourse. The rationalargument,aboutthe naturalnessofcertainresponsesto envi-ronmentalpressure, is also to some extent backedupby theempirical one thatquitedifferentgroups(mostlyindustrialworkers)havebeen foundto reactinanalogouswayswhenfacedwithsimilarsituations. The Universityof Kansas Medical School is, thus, the place wherethe study was done rather than the case studied. Socialunitsintermediateinsizebetweentheindividualandthe wholeschoolarediscussed.Theseunitsaretheyear-groups,seenas developingcollectiveperspectiveswhichfollowfromtheinteraction between students' initial perspectives and their shared experiences, and the fraternities which were seen to structure daily social interaction. It is hardto see these as cases, except perhaps when comparisons are made between differentfraternitiesinterms of theirrecruitmentandits con-sequencesforsocialintegration,andthatis doneonlytoaccountfor differencesbetweenindividualsinthe course ofthe process bywhich the collective perspective emerges. Were individual medical students the cases studied? Obviouslythey wereinthesensethatmostoftheobservationsweremadeofthem. Moreover they are given names, some of the names recur, andsome of theirpersonalsituationsaredescribedindetail.However,thesame individualsare notfollowedthroughas Doc orLong Johnis inStreet CornerSociety;indeed,theauthorsdeliberatelyconcentratedonwhat the students had in common rather than what distinguishedthem (1961: 22). Nor, on the other hand, are numbers of individuals having a given characteristic oftenaddedup. This followsfromthe unusualsampling strategyadopted. The researchersdidnot,asin mostparticipantobservationstudies, attach themselves to one particular group of students and observe them Cases of cases . . .of cases31 as they passedthrough the institution.Instead, they hadshortperiods ofobservationofeachoftheyear-groupsofstudentsinavarietyof situations and at differentpoints over the year. In a real sense, therefore, the sample was one of points of time in the medical-school career more than it was one of individuals.(It thus assumes whatAbbottin Chap-ter 2 calls a stage or career theory of the process of becoming a doctor.) It is not easy to translate that into the language of cases, unless"case" becomesmore orlesssynonymouswith"observation."Insofarasthe mainthingsthatarecountedtoreachconclusionsaretheavailable observationsonaparticulartopic, itmakessenseforitto become so. Some statements are, indeed, made by the authors which supportsuch adescription.Forinstance,theforegoingquotationaboutthe87in-stancesobservedis immediatelyfollowed:"Consideringthatwesaw onlyafewstudentsatatime,itisclearthattheactualnumberof incidents occurring which we were not able to observe must have been much greater"(Becker et al. 1961:309). Thus, althoughthe authors rec-ognize the existence of a variety of social units, none of these theoretical andempiricalentities appearto constitutethe cases whichare usedin their data, although they provide the material and are referredto in the cases. We may still sensibly ask how it is that the cases are used to reach the conclusions. The authorssaythattheystartedtheresearchwithopen mindsaboutdesign,inthesensethattheydidnothavehypotheses, instrumentsofdatacollection, oranalyticproceduresspecifiedinad-vance, andthat their conception of the research problem changedover the course of the work, thoughthey didhave some more general theo-retical and methodological commitments. What in the book is the main researchproblem"becameourcentralfocusonlywhenwewereen-gaged in the final analysis of our materials" (1961:17). The introductory chapterofthe book,however,asisconventional,setstheintellectual scene ina waywhichleadsupto andincorporatesthatcentralfocus, whichisusedasthekeyprincipleoforganizationforalltheother chapters. These chapters followthe students through the chronological sequenceofyearsinmedicalschool,althoughwearetoldthatthe researchers chose to observe the clinical years first although they come laterinthesequence.Thusitisabsolutelyclearthattheresultsare presentedinanorderwhichdoes notrecapitulatethe writers'experi-ence. The "cases" used only became cases of what they are used as cases of retrospectively (cf. Walton, Chapter 5). But within chapters, as well as the bookasawhole, broadinterpretationstendtobepresentedfirst. One of the reasons whythese seem particularlypersuasive is thatthey tendtotreatwhathappensempiricallyascommonsensical,rational, 32WHATisACASE? followinglogicallyfromstructuralfeaturesofthesituation.Forin-stance:"Theenvironment...forcesonstudentscertainchoicesof perspectiveandsuggestsothers"(1961:80)."Theymuststilldecide what is important and worth remembering"(1961:221). "Students' views ofpatients drawheavily. . .onstudentculture.. . .To the degreethat this is so, it supports our view that medical students' behavior can best be understood by seeing them primarily as students"(1961:314). Each of these statements is made at the beginning of a chapter which then provides evidence to support it. It is, of course, a valuablesocial-scientific achievementto understandandexplainasocial situationsothoroughly that youcan make its outcomes andyourtheories aboutthem look self-evident.One ofthewaysthisis done, however,involves anelement of circularity. Placing a general proposition first, as if it could already be taken as given, makes the data presented next look like confirmationof an idea already to some extent established, and guarantees that the itemsoffered will be interpretedas instances of the proposition rather than as evidence whichmighthaveledinotherdirections, orhavebeengroupedunder other heads. The authors say (1961:22) that they revised their provisional generalizationsif negative cases arose, andsoughtdata relevantto their emerginginterpretations,andtothatextentcannotbeaccusedofthe classic mistake of "testing" hypotheses on the same data from which they were derived. The reader, however, is not offeredthe intermediateinter-pretations and the negative cases -and if she were, one might suggest that the authors were shirking the analysis of their data. Rhetorically,thus, oneis offeredalogicallyambiguousrelationbe-tween evidence on the cases studied and conclusions. When, toward the end of the book, a propositionso general as to be plainlyapplicableto manysituationsnotstudiedis offered["Values operateandinfluence behaviorinsituationsinwhichtheyseemto theactorto berelevant" (1961:431)], wearesurelynotmeanttoseeitasderiveddirectlyand solely from the data of this study. I take it that it is implicitly drawing on other cases, perhaps drawn from general social experience or "common sense" as much as fromsocial-scientificsources. It might well be thought that most of the complexities in the use of cases which we have suggested in the qualitative studies so far examined would not arise in more standard quantitative studies, especially those based on surveys with conventional samples. Let us see if this is so. The People's Choice We startwitha classicsurvey,Lazarsfeld,Berelson,andGaudet's The People'sChoice(1944). This is a panelstudyofthe presidentialelection Cases of cases. . .of cases33 campaign of 1940. It was carried out in Erie County, Ohio. A systematic sampleofthecounty'spopulationwasdrawn,andwithinthisthere were stratifiedrepresentativesubsamples of which one constitutedthe panelandthreethecontrolgroupswithwhichitwascompared.The panel was interviewed seven times over the campaign period, and each ofthe control groupswas interviewedonlyonce afterthe initial occa-sion. The prime functionof the control groups was to allow anyeffects on the panel of repeated interviewing to be identified; however, data on them were sometimes also used to give a larger sample size on some key questions. Threekindsofcasescanimmediatelybeidentifiedinthisresearch design:thecounty,thecampaign,andtheindividual.Lazarsfeldand associates make it clear thatthe particularcountyandcampaignwere notofinterestinthemselves; whattheywereconcernedwithwasthe conditionsinfluencingpeople'sdecisionsto vote as they did.Reasons are given for the choice of Erie County: "it was small enough to permit closesupervisionoftheinterviewers. . .itwasrelativelyfreefrom sectionalpeculiarities. . .ithaddeviatedverylittlefromthenational voting trends. . .it is not unlikely thatErie County was asrepresenta-tive ofthenorthernandwesternsectionsofthecountry5 asanysimi-larly small area could be" (1944:3). These reasons take it for grantedthat one had to choose a county or similar small area, and explain why that particularone couldbe seenas areasonablechoice. But whydidone have to? Two quitedifferentsorts ofanswerscanbe suggested,anditis not clear fromthe bookwhichone is intended. The firstanswerwouldbe thatitwasa basic componentofthe researchdesignto usethesocial context, characterized as it could not be from one individual's question-naire responses, to explainvoting behaviour.The material availableis indeed sometimes used in that way, but no statement of intention to do so is made in the descriptionof the research plan; Glock (1988:34) says, however, that a community base was chosen to ensure thatrespondents hadaccess to the same newspapersandradio stations. Thatgoes well with the initial presumptionthat voters would change their minds a lot duringthecampaign,whichwouldputexplanatoryweightoncam-paignevents. In practice there was notmuch change, so lesstransient factors had to bulk larger in explanation. Thesecondanswerwouldbethattoconfineinterviewingtoone smallgeographicalareais muchcheaper; atthetime ofthestudythe national survey organizations of today, withtheir permanentfieldstaffs, didnotexist,andpracticalsamplingstrategiesofakindthatwould enablethemtobefruitfullyusedhadhardlybeendeveloped.To the 34WHATisACASE? extentthatthefirstansweristherightone,thecountyisacaseofa socialarea,anditsweakclaimstorepresentativenessarenotvery important since the objectis not to predict the distributionof votes but to show how voting decisions are formed. (Note the interestingsimilar-ity to the logic of Angell's lack of concern aboutthe quantitative repre-sentativenessofhiscases.) To theextentthatthe secondansweris the right one, the county is not a "case" in the design at all, but an intellec-tual accident. The individualsinthe samplewerecases ofvoters, andformost of the time they are treated as interchangeable units whose answers can be addedandpercentaged.Wemaydistinguishbetweentwodifferent kindsofaccountgivenoftheaggregatesthusconstituted:oneretains their identity as aggregates, the other treats them as if they were mean-ingfulsocialgroups.Asanexampleofthelatter,taketheheadingof Chart27 (Lazarsfeldet al.1944:78), whichsays: "Those whoreadand listen to campaign materials more than average... end up with a higher degree ofinterestthanthose withexposure belowaverage." Whatthe chart actually shows is that among those withexposure above average 21%showedgreatinterest,whereasonly8%didamongthosewith exposurebelowaverage.Whatthesedataliterallyshowis, thus,that moreofthosewithhigherexposureendupwithgreaterinterest,not that all of them do as the heading appears to say. What is going on in the headingisasortofshorthand,6whichleadstotheimputationofthe characteristicsofsome membersof(whatmay be)anarbitraryaggre-gate created by the research to the aggregate as a whole; it wouldmake substantivesensetodothisonlyiftheaggregatewerearealsocial group,in whichthecharacteristicsofsomememberscouldmeaning-fullyberegardedas somehowcharacterizingothermemberswhodo notthemselvespossessthem.Thisisaformoftheecologicalfallacy, since we are given no reason to believe that in this case it is a real social group. On other occasions the authors seem painstakingly to avoidthis fallacy; I direct attention to it because it is so common, not because of its salienceinThe People'sChoice. The issue is worthfurtherexploration, since one wouldexpectthere to be a close relation between the units implied by theorizing and those usedascasesinthedatacollected.Formuchoftheanalysisin The People's Choice, all variables are equal: a factor measured in the research, treatedas characterizing the individual. Sometimes this has ratherodd consequences,aswhen,inthesectiononcross-pressures,thevoter's own attitudes or past decisions are treated methodologicallyas though they were the same sort of thing as his socioeconomic status (SES) or the political preferencesof his friends(1944:56-69). (This is, in effect,a way Cases of cases. . .of cases35 of working with variables instead of cases while still at leastappearing to retainsomeofthecasenessofcases.)Howtheyareintendedtobe regardedtheoreticallyis not always clear in the discussion.Part ofthe problemis thatsomeofthefactorscouldeasilyoperate inmorethan oneway.SES, forinstance,couldaffectthevotethroughthevoter's unmediatedperceptionofhis objectiveinterests, orthroughthesocial pressure fromassociatesto which it exposes him. Each idea appears at some pointsinthetext,notalwayswithanydirectevidencethatthe interpretationusedisjustified7orisconsistentwiththeargumentat otherpoints. The Catholicchurchis, however,treatedasarealsocial group with a collective history (1944:23). (No issues appear to have been salientinthatelectiontowhichthecontentofCatholicbeliefswas relevant, though in more recent elections that has not always been so.) Broadly,however,bytheendofthebooktheemphasishasshifted decisivelyto the idea thatpersonalinfluencein smallgroupsofdirect associatesiscrucial.Thestudyremainsempirically,though,oneofa sample of individuals, so that group effects can be studied only through theirreportsofexperienceingroups,andofpersonalcharacteristics which are then grouped by the researchers. This indicates a key problem of design, which Lazarsfeldand his associates were to pursue in subse-quentwork(describedin the Prefaceto thethirdeditionofthe book). AmongthisworkisColeman,Katz,andMenzel'sMedicalInnovation (1966), whichhas a sample preciselystructuredto deal withthe prob-lem, as it could be because by then the issue had been theorized and the practicalsurveyexperience hadcumulated.The problemaddressedis that ofthe spreadofamedicalinnovation.The samplechosenwas of doctors in four areas, the target was to interview all the doctors in those areas in relevant specialties, andthe doctors interviewed were asked to nametheirclosemedicalassociates -whoitwashighlylikelywould also have beenincludedin the sample. Thus it was clearthat thedata directlyavailablewereaboutrealsocialrelations(ifnotnecessarily groups)ratherthandemographicaggregates.Thecases,however,re-mainedindividual doctors rather than medical networks, althoughthe networks foundare salient in the interpretations. Anotherissuewhicharisesinrelationtothequestionofwhether groups are real is that of deviant cases andtheir significance.Ifgroups are real it may be presumedthatmethodologicallydeviantindividual casesarelesslikelytooccur,butiftheydooccurtheirsubstantive significanceisprobablynotgreatbecausegroupfactorsarelikelyto overrideindividualones.Ifgroupsarenotreal, however,andthe re-searchisdistinguishingmereaggregates,individualsneedtobeac-countedfor,andperhapsclassifiedintermsofothervariableswhich 36WHATisACASE? place them in more homogeneousaggregates. Deviant cases bydefini-tionmeritspecialattention,althoughtheymaynotalwaysreceiveit. The People's Choice often, but not always, gives them that attention. The categoryofpeople exposedto cross-pressuresis in parta collection of deviantcases of differentkinds, since forat least some of the variables usedtodefinecross-pressureitwasbeingin aminorityin relationto others ratedthe same on it which constitutedthe "cross-pressure." The interestofthebook'streatmentofthisisthatitdoesnotfocuson explaining how people came to be in thatminority, buttreats this as a newvariablewithwhichtoexplainvotingbehavior.Methodological deviantcasesarethusredefinedassubstantivelydeviantintheoreti-cally relevant ways. BeforeleavingThe People'sChoiceweshou