in relationship beingit b - courses.helsinki.fi · like the translation ol interpretation of a text. ... The Buss-Perry Aggression ... It has 29 statements, and the subjects are asked

/

3 Theories,lnterpretations,and Concepts

The phrase "social science" sometimes seems like an oxymoron. The sciencesare a model of knowledge because they have demonstrated striking powerover the last thlee centuries, By postulating laws, imagining mechanisms toosmall to view with the unaided eye, and by ever more plecise measurement,the sciences have yielded a mass of detailed empirical knowledge. \Øhen weturn our attention to the social world, however, theories of this sort seemimpossible. A.r'e there Can democracy be mea-sured? There seémp to s that makes them iniþósl.ffiIä-ià -."rur.,{,p..d of a social science ere notobjects at all, thêy ale own ideas, concepts, andperspectives. \Øe can study the stals without asking what they care about;not so with humans. Human subjectivity, many have concluded, constitutesa barriel to sqientific theorizinpEbout the social world. In this chapter, wewill examine tËë questíon of natulalism through rhe lens of social scientificcznceptl lVhele do social scientific concepts get their content? And what isthe relationship beween the theorist's concepts and the ways that the subjectsthink about themselves?

Scientific concepts are different fi'om everyday concepts because they aredeliberately refined. Notions like "atom" or' "gend' were developed to explainobselved phenomena; in the social sciences we have "inflation" and "culture."AII by itself, a concept does not explain anything, just as a single word doesnot say anl'thing. \Øords must be combinecl into sentences, and it is thesentences that are true or false, believable or unbelievable, well or poorly justi-fied. Similarly, concepts change in the light of new evidence only insofar asthey are parts of theories. Think of the dlamatic changes in the concepts of"atoni' and "gene" in the twentieth century as the theories of physics and

Theories, lnterpretations, and Concepts 35

of the social world. The arguments for this position discussed in Section 3.3will tuln on the relationship ti!-ts' concepts ancl theconcepts ofthe people beingit inierpretivism ðoncludethãfthe concepts developed b must be closely relatedto rhose employed by the subjects. This makes social scientific theorizinglike the translation ol interpretation of a text.

The capacity of humans to apply concepts to themselves is the basis forarguments that the entities postulated by the social sciences are very differentfrom those postulated by the natural sciences. Natural science, one mightargue, discovels natulal kinds-the fundamental properties of the naturalworld. Natural kinds might be things like quarks or DNA molecules. Theanalogues in the social world might be institutions like schools or statuseslike being married. One might argue that if a couple does not think ofthemselves as married and if they are not treated as being married by neigh-bors, churches, or courts, then they âr'e not married at all. By contrast, theDNA molecule doesn't care how it gets treated; it doesnt conceptualize itselfThis not only entails that there is a metaphysical difference between thesoffirnd'äatural sciences; it means that the relationship between sociálsðlentific cbncepts and the coriceptualizati,ons of the people studied is âtüvo-way street. \Øhile the social scientists are trying to form concepts ancltlïeories about a group of people, the subjects are observing the social sci-entists and learning from them. Humans take social scientific concepts andmake them their own, Section 3.4will explore this fascinating phenomenonand its implication for the social sciences.

3.1 Aggression,Violence,andVideo GamesDo violent video games cause agglessive behavior? As video game technologyhas made violent intelaction more vivid and pelsonal, concerns about theileffects have risen. A broad research program in social psychology has inves-tigated this question with both experimental and correlational studies. Severaldecades of research have caught the attention of parents and politicians. In2000, the American Academy of Pediatlics declared that, "At this time, welland concepts: Both venerable pedigrees

36 Theories, lnterpretations, and Concepts

over' 1000 studies . . . point overwhelmingly to a causal connection betweenmedia violence and aggressive behaviol in some childlen" (American Academyof Pediatrics 2000). Some investigators have embraced such strong claims,but others have been more circumspect, arguing that violent video games(and other media) at most modulate underlying aggressive tendencies thathave other causes.

Social learning theories are one prominent approach to aggression. Accord-ing to this kind of view, children lealn how to lespond to social situationsby observing and modeling the behavior of others. An individual's knowledgeof how to act in a given kind of situation is encoded in memory. \X/hile thedetails of the encoding vary among social learning theories, the upshot islargely the same. Childlen learn patterns of response to social situations. Aswe matufe, these responses get hardened into personality stfuctures thatresist change. An investigatol with this theoretical standpoint would expecra direct causal link between the environment in which a child develops andaggt'essive behavior. Proponents of different social learning theories havedifferent views about which elements of the environment provide the stron-gest models for emulation. It may be that greater exposure to violent videogalnes, especially first-person shooting games, would effectively train childrento responcl aggressively to social situations. An alternative explanation wouldbe exposure to live aggression, especially family violence, is a much strongerforce in the shaping of behavior.

Another issue in the literature is whethel tendencies toward violent behav-ior have a genetic basis. Studies that compare identical twins reared aparthave supported the notion that there is a genetic predisposition to a violence-prone pelsonality (Eley, Lichtenstein, and Moffitt 2003).On this view, someindividuals are prone to interpret situations as threatening and are morelikely to lespond violently. The environment is the basis for some learnedresponses, but this only enhances or suppresses en innate disposition. Someinvestigators interested in media violence have used this theory to suggestthat the correlation between playing violent video games and aggressivebehavior is explained by a third undellying factor: the innate disposition,Video game violence (and other violent entertainment), these investigatorshave argued, has a negligible effect in the folmation of an aggressivepersonality.

These different theories of aggression have been tested both inside andoutside of the labolatory, In the lab, it is easy enough to expose subjects toviolent or non-violent video games, The trick is to find away of identifyingaggression in an expelimental setting. The investigator cannot simply instigatea fistfight, obviously, so some sort of proxy has to be used. Some studiesadminister a questionnaire after game-play, asking subjects to report theirfeelings or thoughts. Another sort of protocol seeks to identifz aggressivethoughts by asking the subjects to complete open-ended stories, The storiesdescribe a fi'ustrating situation, such as being involved in an auto collision,and ask the subjects to list possible acrions, feelings, or dialogue for the

Theories, lnterpretations,and Concepts 37

characters. Another test identifies aggressive action more directly. In thisorotocol subjects are told they are playing a computer game against anåopon.nr. The winner of each trial gets to set the volume and duration ofa'noise blast heard by the loser. They are, in fact, playing only against acomputer. Louder and longer blasts are taken as more aggressive.

In colrelational studies, investigators use surveys to look for a relationshipbetween playing violent video games and various folms of aggression. Again,while it is easy to understand how a questionnaile might ask about whatvideo games a person has played, it is more difficult to see how aggressionmighr be measurecl. Direct measules would look at actual events, such as

ar,.ãst .e.ords. Aggression is a lalger phenomenon than criminal activity, ofcoul'se, so investigators must cast a wider net. Surveys of attitudes are acommon way of indirectly measufing aggression. In contemporary surveymethodology, qu€stions are divided into groups that correlate highly witheach other. The identification of such "factors" is an important (and chal-lenging) palt of developing a useful survey instrument. Each factor is sup-posed to explain or account fol a significant poltion of the variation inãnr*.rr. The idea behind the constluction of factors is that while a phe-nomenon like aggression is manifested in a variery of ways, these differentmanifestations arise fi'om a smaller number of underlying ploperties. Thefactors should identi$r these deepel structures,

The Buss-Perry Aggression Questionnaire (Buss and Perry 1992) is a rypi-cal survey instrument. It has 29 statements, and the subjects are asked torâte them on a scale of I (extlemely uncharacteristic of me) to 7 (extlemelycharacteristic of me). Vhile the subject answers the questions in a random-ized order, the investigators know that questions divide into groups corre-sponding to foul factors: physical agglession, verbal aggression, anger', andhostility. For example, there are five questions that compose the velbalaggression factor' (Buss and Perry 1992, 454):

. I tell my friends openly when I disagree with them.. I often find myself clisagreeing with people.. \7hen people annoy me, I may tell them what I think of them.' I cant help getting into arguments when people disagree with me.' My fi'iends say that I'm somewhet algumentative.

A subject's answers on the scale of l-7 arc combined to create a single verbalaggression score fol that person. A rypical question of correlational researchon video game violence, then, is whether subjects who report higher levelsof exposure to violent video games have higher levels of physical aggression,verbal agglession, ange¡ and hostility as measured by factols of thequestionnaire.

The study of video games and agglession raises a number of importantphilosophical questions, ancl we have ah'eady noted several of them. Is itreally possible to measure things like "agglession'? The methodology of


survey research presumes that indirect measures can capture the relevantpatterns which will help us understând, explain, and predict aggressivebehavior. Do these substitutes ("loud" blasts punishing a computer, answersto the Buss-Perry Aggression Questionnaire) really cepture the'concept ofaggression? And what lr the concept we are trying to capture, and howwould we know that we have it right? Does it matter if the teen gamersthink about aggression differently than the social scientists who study them?'We will discuss these questions in the next two sections. Others arise aswell. The question with which we began is about causes: Do violent videogames cause aggressive behavior? How can a questionnaire be evidence fora causal association? And even if we could discover something like a causallaw linking video games with aggression, could it really explain why Fredpunched his little brother? The problem of whether human action can beexplained causallywill be discussed in Section 5.1, while the more generalissues about the metaphysics and epistemology of causality will be discussedin Chapters 9 and 10.

3,2 DefiningTheoretical Concepts

The EmpiricìstYiew of Conceþts ondTheory Structure

The body of empirical research briefly described above asks whether thereis a causal relationship between playing violent video games and aggression.One of the philosophical questions that arises from this research is: 'Vhatis "aggression" as defined by these studies, and how is it measured or empiri-cally identified? The theories ofaggression discussed above all postulate thataggression arises from an enduring feature of a person: an "aggressive per-sonality." The difference beúveen the social learning theories and the innate/genetic theories is primarily a difference about the causes of this kind ofpersonaliry and the degree to which it is environmentally malleable. Theaggressive personality is a tlteoretical entity or posit. It is not directly observ-able; it is something known only through knowledge of the theory. In thisrespect, theorists in this area ere adhering to a philosophical picture ofscientific theories and concepts that goes back at least to the scientific revo-lution. The empiricist view of theory was formulated in a strong and clearway by mid-twentieth-century philosophers of science, but the central ideascan be found in a broad range of philosophical and scientific writing. It hasbeen commonly expressed by social scientists in their writing about theoriesand theory construction.

On the empiricist view of theory, a theory is a structure of general state-ments that explains some phenomena and permits predictions about them.These statements are often called "laws" or "nomological generalizations."Newton's mechanics was taken as an exemplar of a scientific theory. Givenknowledge of some initial conditions (such as mass, force, position, or veloc-ity), Newton's laws of motion explain current motions of an object and

Theories, lnterpretat¡ons, and Concepts 39

^Dpear in the theory's general statements. In precisely quantified theories,

,hå "or..prt are often associated with the variables of the laws. Newton'ssecond law of motion is that F = ma. It relates force, mass, and acceleration,which are three of the central concepts of Newtonian mechanics. The con-cepts of a theory might be more or less directly observable or measurable.V/è learn in high school physics that mass and weight are different. '\ü?'eight

is what I feel when I pick up a barbell. Mass is a technical concept ofNewtons theory. In the absence of gravity or acceleration, the barbell isweighdess, yet it retains its mass. \Øeight is easily observable, but mass mustbe calculated, via the statements of the theory, from weight in a gravita.tionalfield or some other observables. \What are often called "theoretical concepts"are rhose that are relatively remote from observation. On the empiricist view,the statements of the theory give content to (we might even say "define,"but here we have to be careful not to beg important questions) the theoreti-cal concepts. Thus, in Newtont theory, the theoretical concept of mass isthe quantity that is equal to force divided by acceleration (if .F = rnd., then¡n = Fla), the quantity that is equal to momentum divided by velociry (sincep = mu, m = p/u), and so on.

Returning to the studies of aggression, "aggressive personality" is a theo-retical concept. Having an aggressive personality is not directly observablein the way that playing a video game or punching someone in the nose isobservable. That some people have an aggressive personality is a theoreticalposit. The different theories of aggression agree that there is such a thing,that an aggressive personality causes aggressive behavio¡ and that aggressivebehavior is relatively observable. They differ in their specific accounts ofinternal structure and etiology of "aggressive personality." These theoreticaldisagreements result in differences in what 'ãggressive personality'' means inthe different theories. For a social learning theor¡ an 'ãggressive personality''is a set of memories, behavioral scripts, and learned affective responses tothe environment. These are created by the persont experiences, and theycause a actions as threatening and to reect to thosethreats. is an inner state. The central question ofbehavio relatiye proportion of the contributions of

. Hence, for a behaviolal-genetic theory of aggrcssion,is a phenotyp¿ in the serne sense as songs are part of

d species. That is, it is a patteln of behavior createdjointly by the individual's genes and the environmenr.

Research on the question of whether violent video games cause aggression,and the underlying research on aggression and pelsonality, can be fit into


the empiricist view of theoly. Both social learning and behavioral-genetictheories are systems of generalizations about how aspects of personality aredeveloped, maintained, and expressed. Through this system of law-likegeneralizations, they give content to concepts like "aggressive personality."To test the theory, the investigators derive hypotheses that satisf' two clitelia:(1) The hypothesis should be logically entailed by the theory; and (2) itshould include concepts that are amenable to direct or indirect observation.The logical relationship between hypothesis and theory makes the test pos-sible. If observation shows that the hypothesis is false, then as a matter oflogic alone, some proposition of the theoly must be false. The secondrequilement means that the hypothesis can be observationally determinedto be true or false. Hele, however, there is a substantial challenge. 'Aggres-sive personaliry'is not the solt of thing that can be dilectly observed, Howcan social scientists identify ol develop leliable observation concepts?

Reolisrn, lnstrumentolìsm, ond the Problem of ConstructVolidity

Survey lesponses seem to be a simple and clear species of observable behavior',and as a lesult, they are a common method of theory testing. The vaguenessand ambiguity of everyday life is neatly calved into distinct components.Each question has an answer chosen fi'om a cliscrete set of options. Theuniformity of the answers pelmits direct comparison among responses ofdifferent subjects. Moreover, since the enswers can be counted, investigatolscan calculate plobability, corlelation, risk, odds ratios, and a host of otherlelationships. Patterns of survey response among the palticipants are thereforea popular observational basis for testing hypotheses. However, it would bepointless to folmulate hypotheses that directly clescribed parterns of surveyresponse. After all, we do not learn much fiom the obseryation that 43 per-cent of the respondents answered "yes" to question 7. The discussion of theBuss-Perly Agglession Questionnaire, above, shows how these pattelns aleturned into concepts that are more closely lelated to the theories that socialscientists want to test. The concepts are defined by responses to the questionsand called "factors," "latent variables," or "constructs," and the hypothesestested by a survey are expressed in terms of these constructs. These scientificpractices raise several philosophical questions. \Øhy should we think thatthese pattelns of survey response represent anything? \Øhy should we thinkthat a higher score on the physical aggression factor ofthe Buss-Perry Aggres-sion Questionnaire means that the person is physically agglessive?

In the methodological literature, these philosophical questions are often calledproblems of "construct validity." There a-r'e two broad philosophical positionsfor us to evaluate. A realbt position about constluct validity holds that a validconstruct is one that measures what it purports to measure. In other words,real features ofthe world correspond to theoretical concepts or constructs, andvalid surveys (or other tests) can measure them. \W4rile this position may seemlike common sense, it is not the most common philosophical view. The

Theories, lnterpretations, and Concepts 4l

The canonical expression of constluct validiry is Cronbach and Meehl's essay,

"Consrruct Validiry in Psychological Tests" (1955).They alticulated a positionon theoretical concepts in the social sciences that is consistent with the empiricistview of theory. Again, on this view, a concept has specific defining attributesbecause of its role in the theory. The question of construct validiry then, is toshow rhat the specific attributes of the concept are found in the observationalevidence. Considel the construct of ierbal agglessiori' in the Buss-PelryAggres-sion Questionnaite. Verbal aggression is treated as a theoretical posit, like'hggressive personaliry" but'verbal agglessiod' is closer to the surface of observ-able behavior'. As a theoretical posit, verbal aggression is part of a nework ofconcepts (including physical aggression, anger, and hostility) related by gener-alizations. Sometimes these are rough or pre-theoretical generalizations; some-times they may be substantiated by ealliel research. This is a theory (Cronbachand Meehl call it a "nomological networli'), and it makes preclictions abouthow the different factols should corlelate. Roughly, if the factors do not cor'-lelate in the way that theory pledicts, then the theory must be modified, whichmeans that the content (definition) of the concepts is modified.

Cronbach and Meehl take an instrumentalist approach to construct valid-ity. The goal is to find a theory that fits the observation, and the only realquestion is whether the data points corlelate or cluster in ways that fit theconcepts. The lure of instrumentalism lies in the way it resolves the epistemicpredicament of the scientist. The predicament arises when the theory pos-tulates the existence of something and we have no way of directly obselvingol measuring it. The only epistemological resources are more observations:mole tests, mole sulveys, more expeliments, and so on. The best we can dois to compare these observations with each other, using them to rule outtheories that do not conform to the data. In the end, a good theoly is onethat accounts for all of the observatiolrs. \Øe can ask no more, yet a realistseems to want more. The lealist wants to ask whether the theoretical con-structs realþ corcespond to the hidden properties or structures. The instru-mentalist replies that since there is nothing but observations that couldanswer that question, the realist demand cannot be satisÊed.

In response, realists have argued that there is something incoherent aboutthe instrumentalist approach to concept validity. The idea of measurementt'equires two conmitments, and instrumentalism violates both. Filst, measuringthe properties of an entity requires commitment to its existence. For example,it is impossible to design an insrrument that would cletect the presence of


Santa Claus in the chimney. The problem is not that we have no theory ofSanta Claus (we do!), but that we think Santa Claus does not exist (sorrykids), Therefore, it is incohelent to sa¡ "This instrument measures aggrrssion"and at the same time deny that aggression exists, Instrumentalism thus seemsto be an inconsistent position. Second, any meesurement instrument must becausally related to the things measured, such that changes in the object causechanges in this instrumenr. This is true of mechanical measuring devices, likethelmometers, as well as all kinds of surueys and tests, Denying such a causalconnection rendels insignificant any claim that a difference in tesr score (etc.)means that there is some diffelence in the things measured, and it would makemeasurement irlelevant to both theory and practice. To adopt an instrumen-talist attitude toward theoretical concepts, the realists argue, is to deny thevery thing which makes measurement possible.

A practical consequence of this debate is that some social scientific researchseems to be superficial precisely because the construcrs are so thin and under-developed. Starting with a large enough pool of quesrions, it will always bepossible to find co¡r'elations among them and to derive "factors" from thosecorrelations. Of course, survey developers never start with random questions;they start with questions that they guess will capture the phenomena in whichthey are interested. But if these initial concepts are wealdy developed, ambigu-ous, or vague, the surueys cannot yield significant results. The instrumentalistunderstands this line of criticism as showing that good resealch requires goodtheories. \Without well-developed and carefully articulated theolies to suppolrthe constructs, the suryeys will be super'ficial. The realist will lespond that thecriticism goes deeper. Some social scientific research tries to measure entitiesthat we have, as yet, no good reason to believe exist over and above patternsof correlated survey responses. The result is studies that are methodologicallysound, but theoretically vacuous,

The debate between realism and anti-realism is a conflict between rwobload views about the project of science. A realist takes the goal of scienceto be the discovely of truths, including mechanisms and entities that arenot directly observable. The anti-realist denies that the goal of science couldbe the discovely of truths or theory-independent facts, Anti-realism takesmany forms, depending on the grounds for rejecting lealism. Instrumenral-ism is a form of anti-realism which arises within an empilicist fi'amewolkfor theory construction. \Øe have seen in this section how the diffelencebetween realism and instrumentalism has importanr consequences for theway social scientists understand their research.

3.3 lnterpretivismYou may already have had the thought that the research on aggression doesnot capture the way the subjects think about, reacr to, or feel aggression.For example, context seems to matter. If I were to say, "That is an incoher-ent statement" to my daughtel over breakfast, it woulcl be taken as aggressive:


Indeed, some have algued that the whole empiricist appalatus of law-likeseneta)ization and hypotheses testing is blind to the subjective charactel ofíocial realiry. Therefore, some have concluded, the social sciences need con-cepts and forms of theory that are diffelent fi'om those of the natural sci-ences. Following common philosophical usage, this form of anti-natulalismwill be called "intelpletivism."

ldeol TyPes ond Verstehen

Max Weber, the great social theorist of the late nineteenth and eally twentiethcentury, articulated an infuential vision of social inquiry. Central to hismethodological approach was the notion of an "ideal rype" (\Øeber 1949119041,85-110). Concepts ofthe social sciences, \feber argued, need tocapture the meanings and motivations that are significant fol the subjects.To do so, the social scientist should first describe the actions to be explainedin terms of their typical motivations, The concept is then formed by abstract-ing characteristics of the action fi'om a valiety of observations. The uniquecharactel of social inquiry is that ideal types must inclucle rypical moriva-tions, attitudes, and meanings. Ideal types therefore relate observable behav-iors by iclentifying the motivations that stand behind them. Tre concept of"aggression," Webel would say, must be developed as an ideal type concepr.It is not sufficient to define it solely in terms of observable behavior, suchas selecting the loudness of a sound blast. The concept needs to iclentify therypical feelings, beliefs, and social meanings that stand behind the behavior.In this respect, the definition of aggressive behaviol in rerms of the "intentto cause herm" (Andelson and Bushman 2002,28) is on the right trackbecause it picks out a typical motivation. However, since the commonunderstanding of agglession also involves othel feelings, motivarions, andresponses, -Veber would say rhat the deÊnition remains incomplete.

How would an investigator determine that hel concepts ale incomplete orotherwise inadequate? \Øe have seen the empilicist answer; Concepts get theirmeaning from their role in a theor¡ and a theory is a system of laws. Themeaning of a concept develops as the theory is modified in response to resr-ing. \Øeber could not acc€pt this account because he rejected laws as a neces-sary basis for social scientific understanding. While the discovely of law-likegeneralizations might be impoltant and useful in some conrexrs, he algueclthat the expression of laws in the social sciences al'eady required well-developedideal type concepts. Ideal rypes are thus independent of and pliol to

/


generalizations. \Øeber emphasizecl that ideal types were idealizations; theywere tools for the identification of central aspects of a phenomenon (\Øeber1949119041,90). Yet, unlike the idealization of a "fi'ictionless surface," idealtypes are to be compared with the evidence and modified in its light, A¡ideal type predicts actions, historical changes, or social developments on thebasis of the palticular motivations it identifies. If these actions do not arisein the expected context, there is reason to modi$r the ideal type.

The concern with which this section began was that a naturalistic perspec-tive on the social sciences, particularly the empiricist view of theory andassociated methods, failed to capture the subjective character of the socialworld. \Webert ideal types go some way toward addressing this concern, butone might argue that they clo not go far enough. Ideal types treat the sub-jectivity of the agents as a construction of the social scientist. It is the socialscientist, after all, who chooses to emphasize one aspect or another of thesocial environment, This replacement of the subject's view with the scientist'stheol¡ some have felt, is plecisely the problem with survey research andsimilal methods. \Øebert idea might be pushed farther by recognizing thatthe subjects have their own ideal rypes. That is, as participants in a socialgroup, we have oul own understanding of the possibilities for action. \Øeknow the expectations, regulations, roles, and institutions that make up oursocial reality. Alfrecl Schutz called this "common-sense rhinking," and heargued that uerstehez is "the particulal experiential folm in which common-sense thinking takes cognizance of the social cultulal wolld" (Schutz 1954,264). A full-blooded uerstehen approach to the social sciences would thuscapture the subjects' own unclerstanding of their world.

Schutz suggested that the social sciences require two levels of ideal rypes.The first is the level of common-sense thinking, and it contains the "theories"used by a group ofpeople to understand each other. The seconcl is the socialscientiÊc models of motivations, feelings, and meanings. The social scientist'stheoly construction proceeds with the second-level ideal rypes. Since theaim of social science, on the uerstelteru view, is to understand the subjects'perspective, the second level must reflect the first level in some way. Articu-lating this lelationship between first- and secondlevel idea rypes is a difficultphilosophical problem, Schutz ploposed that the second-level ideal typeshad to be "consistent" with the common-sense thinking. As he phrased this'þostulate of adequacy" it lequiled that:

Each term in a scientific model of human action must be constructedin such a way that . . . the rypical construct would be understandablefol the actor himself as well as his fellow-men in terms of common-senseinterpretation of evelyday life.

(Schutz 1953,34)

\íhile this arguably does more than lVeber's idealjective dimension of social concepts, it too has its

rypes to capture the sub-challenges. The postulate


able to understand the constt'ucts, not that they agree with them or usethem in their daily lives. On this reading, the requirement is too weak,because we lose any sense ofhow the second-level ideal types correspond tothe trst level,

Hermeneutics ond Meoning

Vhat is the lole of language in common-sense thinking about social realiry?Another way of replesenting Schutz's problem is that there ale two languagesinvolved in the social sciences: the subjects' own ancl the social scientists'.Even if the scientist comes from the same social milieu and speaks the samelanguage as his or hel subjects, the social scientist is developing a technicalterminology (ideal types) with which to describe and explain social events.\fhat is the relationship between the subjects' language and the specializedlanguage of the scientist?

Empiricist approaches ú'eat the ploblem of lelating the scientist's technicallanguage to the subject's language as dìe problem of relating theory toobservation. \(/e saw the outline of this view in Section 3.1. At the outsetof social Lesealch, we do not know the subjects' motivations, feelings, mean-ings, and so on. Inquir¡ empiricists argue, must thelefore begin with aneutlal description of behaviols. From this pelspecrive, one can see theattraction of survey and testing methods, since they need only a very thindescription of the action, e.g, that the pelson selected "Vely like me" inresponse to question 14,

Interpretivists have been very suspicious of the idea that such an impov-erished observation language could possibly support interesting social scien-tific research. Language, many have pointecl out, is deeply integrated intosocial life. Events of social life get their identiry-even their very exisrence-because of the way in which language ancl action are mutually embedded.Charles Tâylor's essay, "Interpretation and rhe Sciences of Man' (1971), usesthis point about the lelationship between language and action to algueagainst any naturalistic approach to the social sciences.

Consider two playels at e game of chess. Suppose one player moves herknight fivo squares up and one over, and we want to explain why. A neutlalobservation language would let us describe the state of the board and theregularities of motion of the pieces. But simply saying that the horsey-shapedpieces generally move two up and one ovel will not suffice. \XZhile true, it

;


actions get their identity fi'om the rules. \Øe do not know what "rhe move-ment of a knight" means until we have learned the rules of chess. The ¡ulestell us what counrs as a legal move of a knight. (\x/e will lerurn to this ideaand develop it further in Section 8.3.)

Tâylor argues that a neutral language restricted to the motions of bodies,answers on a survey, or even descriptions of the beliefs and attitudes of the

Thick Description and ¡rs Chorrenges

One of the clea'est articulations of the idea that social scientific conceptsare translations is found in CliffoLd Geertz's essa¡ "Thick Descriptiôn:Toward an Interpretive Theo'y of culture" (Gee'tz r973b). Geertz is an


,{ maximally thin descliption would be something like the behavioristobservation language criticized by Tâylor. To describe a person as simplywalking would be a thin description. Thin descliptions have minimal rela-tionships to othel descriptions. That is, to say of a Person that she is walkingis to say almost nothing about the person's goals and motivations. A thickerdescription of the same event would be that the person is walking to BowdenHøll, or that she was walhing to class. These ale thick descriptions becausethey show how the action is related to othel aspects of the person's life andsocial environment. "Hurrying to class" suggests a possible set of motivesand specific relationships to prior and subsequent activities. It invokes theidea of getting to class on time, the possible alternative modes of tt'anspolta-tion, and the reasons why a person woulcl prefer walking to driving or takingthe bus. Thick descriptions ofactions or social events have, in virtue oftheirmeaning, specific lelationships to other actions, events, motivations, possi-bilities for response, outcomes, stt'ategies, and so on.

The conceptual relationships expressed by thick description are alreadyembedded in the language, symbolic system, and actions of the subjects.They thus correspond to Schutzt common-sense thinking, However, Geet'tzdoes not follow Schutz in regarding common-sense thinking as the subject'stheolies about their own society. Following \Øittgenstein, Geertz treats themeaning of a word as its conditions fol use in a communiry, This notionof meaning can be generalized to symbols of all kinds, and to the meaning-fulness of actions as well, Because meaningfulness depends on patterns foundacross the whole community, thick description captures what the membersof the communiry have in common-theil culture. Culture is, in Geertz'sphrase, an "acted document" (Geeltz I973b, 10). Schutzì common-sensethinking becomes nothing more ol less than the totality of actions, utter-ances, and social events that make up the cultule. The goal of interpretivesocial science is to thickly describe the culture, and theleby express in theinterpreter's language the lelationships that make the subjects' social worldmeaningful.

Interpletivism has a number of consequences that have tloubled bothphilosophers and social scientists. Thick descliptions are generalizations abouta group, and some have found generalizations about "the" cultut'e problem-atic. There is a strong tendency in interpretive social science to understandthe beließ, values, meanings, symbols, norms, and actions of a gloup as asingle coherent system. Against this, one might point out that social groups


are typically riven by conflict and contradiction, Different people do notfind the same meaning in a social event or symbol; they clo not undelstandthe demands of a rule or norm in the same way. Moreover, differences ininterpretation can be closely tied to social relationships of powel and domi-nation, By plesenting a single narrative of "the" culture, the interpretationnot only misrepresents the social reality, it takes up a position within thepower structures of the sociery. One group's common sense is highlightedas the true account, while other', dissenting voices are eclipsed. Thick descrip-tion, some conclude, is inevitably a political act. In the terms of Section 2.3,interpletations ale not value-neutral. One criticism of ealliel anthlopologistshas been that they were not sufficiently sensitive to the unintended politicalconsequences of their work. As a result of these criticisms, contemporeryinterpreters are often concerned to represent conflict and difference ofvoicein the interpretive description of a group. It has also lesulted in a demandfor reflexiuity on the part of the interpretel a requilement that will beexamined in Section 4. L

The problem of generalization goes deeper still, insofal as it questionswhether cultures have clear identities at all. Social groups not only disputemeanings, they borrow meanings and practices fi'om each other. Any thickdescription will depend on the interpreter's choices about which meanings,values, symbols, and so on are rypical of the cultule, In an important sense,then, "the culture" is nothing more than the artificial creation of the anthro-pologist, or so it has been alguecl. This clispute raises some very difficultphilosophical questions: A.le cultures bounded entities with rypical features?And if cultures or social groups are always artificial, what consequences doesthis have fol the social sciences?

3.4 Realism and Social Concepts

Sociol Constructions

A defender of interpretivism might argue that the objections mentioned atthe end of the last section arise from not taking the commitments of inter-pretivism seriously enough. To be concerned that thick description mightmisrepresent the culture is to treat an interpretation as if it were a theoreticalrepresentation of an independent object. On the contrery, by treating cultureol society as a text to be read, intelpretation collapses the clifference betweensubject and object, As Geertz put the point:

In sholt, anthropological writings are themselves intelpletations, andsecond and thircl order ones to boot, . . . Th.y are, thus, fictions; fic-tions, in the sense that they aLe "something made," "something fash-ie¡sd"-¡þs original meaning of f.cto-,not that they are false, unfactual,or merely "as if" thought experiments.

(Geeltz 1973b,15)


ln Secrion 3.2, we noted that a realist position about science held that there2re enrities which cot'respond to scientific concepts, and they exist independentlyof our theories. Interpretivists often defencl a form of anti-realism known as

sociøl constructiuism, Social constructivists argue that social factors play anineliminable role in the constitution of scientific concepts and theolies. 'Whatis taken as a "scientiÊc fact," they algue, is the result of a process of negotiationthat is thotoughly imbued with political, moral, and economic interests.

One way to sharpen the issues is to invoke a distinction between naturalal kinds are denoted by terms like "gold" or "riger."of these terms share a cluster of propelties. The

I co-occurrence; the properties that constitute goldatural kinds are supposecl to be distinct from those

objects that ale grouped together because we find it useful or convenienr."German beer'" would be an example of such a nominal kind. \X/hether rhisterm is true of a pafticular bottle of beer does not depend on any naturalfeatures of the beer-a beer made across the border in the Czech Republicmight be indistinguishable. \Øhile Czech and German beer connoisseursmay dispute the diffelences, the distinction depends entirely on humaninrerests. In an obvious sense, nominal kinds are socially constructed,

In the social sciences, the analogue of natural kind terms are social events(the Bolshevik Revolution), institutions (Apple Corporation), roles andsraruses (married, police officer), as well as race, gencler, and other identities.We might call these social þind¡. The question, rhen, is whether social kindsare like natural kinds, ol whethel social kinds are nominal, If social kindsare like natural kinds, then a realist artitude toward the social sciences wouldseem to be warranted, Convelsel¡ e constructivist stance would be supportedif social kinds are nominal.

Reolism obout Social Kinds

It might seem obvious that all social kincls are nominal. After all, institutions,rituals, political parties, and so on are all cleatecl by humans and depend onhuman action for theil existence. That a group of people form a caste of"untouchables" does not reflect a feature of their genes, psychology, or. physi-ognomy. Rather, it arises fi'om the way rhar they are tleated in a particularsociety. Such treatment may arise during an historical period, last for a rime,and then disappear as the sociery changes. Being an untouchable-nor romention a Republican, a Llnitalian, or an employee of Microsoft-seems nodifferent from being "an object on my desk" or "a plant commonly foundon college campuses," They ale nominal kinds created by us for. particularpurposes. If social kinds are nominal, then there is no "narural" cluster ofpropemies that exists independently of human thought and language. .Veshould therefole be anti-realists about the social sciences.

The foregoing argument for anti-realism (in the form of social consrructiv-ism) gives rise to a puzzle. One difference berween natural and nominal


kinds is that natulal kinds enter into robust, causal generalizations. Knowingthat something is an object on my desk lets you make no pledictions abouiit. \X/ill it burn? \øill it make you sicl< if you eat it? To know that somethingis gold is to know its melting point, specific glaviry capaciry to bond withother metals, and so on. The causal powers of natural kinds are a reason tobe realist about them. the ptzzle is that some social kinds have causal pow-ers. Moreove! there seems to be a difference between those social kinds withcausal powers and those without. To be Black in the United States is to beat a higher risk for diabetes and heart disease than those who are Caucasian.Or again, married people tend to live longer than those who stay single.On the other hand, no similar causal generalizations can be made aboutpeople who shale the postal code 53705, or about those who shop at ama-zon.com. Hence, some social kinds are like natulal kinds and others ale likenominal kinds. Yet, racial categories and social statuses like "married" areexemplary cases of social constructions. Is it possible to lecognize that socialkinds are dre constlucted products of human intelests and action, yet adopta realist stance toward some of them?

One way to get a grip on the above question would be to try to explainwhy some social kinds figure in causal generalizations, while others do not."Gold" seems to do so because of its micro stlucture; social kinds are (pre-sumably) different. Michael Root (2000) argued that real social kinds havethree features:

Local deployment: 'Ihe concept is usecl by people in a community tomake distinctions among themselves.Identi4t: ïrose classited take up the classification and it forms part oftheir selÊconception ol identity,Norms: Thele are llorms that plescribe differential tleatment based onthe classification.

Using these three criteria, we can see why some social classifications wouldÊgure in lobust, causal genelalizations. Consider the difference between thestatus of being married and of having the postal code 53705. Both ale deployedlocally; we use both malital statns and addrcss to distinguish among people.Unlike marriage, however; postal codes are not taken up by individuals as partof their identity. Marriage influences a person's plans, attitudes, and actionsin a way that postal codes do not. Finall¡ norms prescribing differential treat-ment help distinguish between accidental correlations and those that properlyarise fi'om the stãtus. It might be, for example, thar residents with thè SEIOSpostal code ale more likely to contract cancer, since the houses there are builton a conraminated industrial site, This is an accidental correlation beweenpostal code and health. On the other hancl, people who are married enterinto a network of laws, rnolal values, and informal norms. If married peoplelive longer', we woulcl expect that it is because of the normatively sanctionedfolms of mutual caring that are part of the social status of marriage.

heories, lnterpretations, and Concepts 5 I

criteria fit with Schutzt "postulate ofSchutz required that the concepts ofthe subjects, and how this idea was

lee features would guerantee that anya thick description, since they require

he individuals. Real socialkinds muste understanding implicit in a social

o¡eup. Therefole, even if interpretations are fictions in Geeltzt sense, it is

ítill åppropLiate to take a realist stance toward social kinds.Tháre is, then, at least one way in which it is possible to combine lealism

with social constructivism. This conclusion needs to be considerecl carefully,however. Notice how the argument clepends on the use of causal generalizationsas a criterion for treating a social kind as "real." This point might be disputed,and orher criteria fol'i'eality' have been ploposed, Notice also how this argu-menr cloes not assimilate the reality of objects like gold atoms to the reality ofsocial statuses. Those who argue for realism on these grounds are theteforeadmiring that the criteria for counting something as leal in one domain mightbe different tha¡r the cliteda in another. These subtleties should make us waryof quick conclusions in the discussion of realism and anti-realism.

Looping Effects

Social kinds can have real effects on the lives of people. In many parts ofthe world, to be a homosexual opens a person to various forms ofdiscrimination-some subtle, others legally 5¿¡ç¡ie¡sd-that infuence onetprospects ancl possibilities. At the same time, when e person takes up theidentity of being a homosexual, it provides resources for understanding onetacrions, thoughts, feelings, and relationships, and it informs one's plans andgoals, The three criteria foL identifying real social kinds, discussed above,help us understand why social categories have these effects. The content andconsequences of social descriptions like "homosexuel" are not static, Ofcourse, the content of a social câtegory is changed by the way in whichpeople use the category to understand, explain, and justi$r their actions andexperiences. Not long ago, homosexualiry in the United States was treatedas a psychological disease and a criminal status. Today it is treated differentlybecause of the social changes blought about by those who adopted theidentity. 'Ihe gay pride movement succeeded because individuals explicitlydescribed themselves as homosexual and expressed that identity in action.

Social kinds therefole have a dynamic quality that Ian Hacking has called"the looping effect" (1995). This effect is readily appalent in an example likehomosexualiry in the United States. The movement to eliminate the moralcondemnation of homosexualiry and constitute it as a positive, ol at leastmorally neurral, status had the explicit goal of helping people selÊidentify asgay and act on that identification. Hacking has explored more subtle andcontroversial examples, such as child abuse and multiple personality disolder',


sciences. The point is not simply that obselvation requires interaction with,and thereby alters, what is observed; this is true in the natural sciences as

There is a further consequence of the reflexive character of social inquirythat brings us back to the discussion of the value-laden characrer of the

every generalization or form of study that is concerned with an existingsociety consrirures a potential intervention within that sociery: and thisleads through to the tasks and aims of sociology as critical theory.

(Giddens 1979,245)

Section 2. genrs er they wantto be or n ntists ial outcomesfol their s are n conuollable.

Theories, lnterpretations,and Concepts 53

3.5 Chapter SummarY

This chapter has sulveyed one of mo-loeical nãturalism. Does human andthËories of the social sciences be L as

the natulalists claim, is thele a single form of scientific knowledge? Many

the empiricist view of theory is that concepts can be identified or measured,and the theories can be tested.

Interpretivists ct'iticize research based on an empiricist view of theory,

is the essence of Geertz's thick description,A second question that has lun through the chapter is whether we should

take a realist stance towald the social sciences. A realist view of the social sci-ences would take its concepts to correspond to real social entities. \7here thenatural sciences discover narural kinds, the social sciences discover social kinds.\Øe saw two forms of anti-tealism in this chapter: instrumentalism and socialconstructivism. The discussion of the looping effect and of the causal conse-quences of social kinds shows that there is room for sophisticated, middle-


This chapter is the opening salvo in a larger battle berween naturalist an4tween an empiricist conceprion of theoryChapter' 5. Some of the most damagingon of theory come not fi.om inter.pretivlto develop causal theories. In Chapters 9

and l0 we will discuss. the critique of the empilicist conception of causalityand law, and we will look ar some recenr alter.natives. Interpretivis- fa..,troubles of its own, and in Chapter 4 we will look at the epistemologicaland methodological issues that it raises, How can we come to þnow whatthe subjects believe or how they conceptualize their activities? Interpretivisrnis not based on a neutral body ofobservarion. \Øhat, then, aLe the èvidentialconstraints on interpretation? \Øhat makes one interpretation better thananother? These are questions for the next chapter.


Further Reading

aggression'.'"îhe classic statement of the empiricist view of theory in the philosophy

of social science is in Book VI of Mill's .S7sra published as Ozthe Logic of the Moral Sclences (1987 lIB72l). cl on the natut'alsciencès, Hempel's little textbook Philosophlt ience (1966) hasclear andnuanced presentations of the standard view and Nagel's monumentalStructure of Sci.ence (1961a) has an extensive discussion of the social sciences.

Mertont Social Tlteory and Soci.al Structure (1957) embeds these philosophicalideas into its methodological discussion. Hage's Ti:chniques and Problems ofTheory Construction in Sociologlt (1972) is a well-known text; Hage updatesand refects on this view in Hage (2007).

Measurement is a lalge topic in the methodology literature, ancl it involvessubstantial technical complexities. Clonbach and Meehlt seminal presenta-tion is surprisingly accessible, and the philosophical ideas are close to thesurface (Cronbach and Meehl 1955). Tlout's Measuring the IntentionalWorld(199S) relates the technicalities of measurement theory to the cleeper philo-sophical questions, and Michell (2007) gives an overview of the issues.

Maltins Wrstehen: The Uses of Under*anding in Social Scienc¿ (2000) isan extensive overview and critique of different conceptions of uerstehen inthe social sciences. \Weber presents his conceptions of uerstehen and idealtypes in several places. "Objectivity in Social Science and Social Policy'(\Øeber 1949 ll904l) is a readable presentation. Interpretation of \Webeltìdeal types can be found in Aronovitch (2011) or Ringer (1997).Classicârguments against the empiricist conception of theories and concepts isfound in Schutz's "Concept and Theoly Formation in the Social Sciences"(1954) and Täylor's "Interpletation and the Sciences of Man" (1971). Foran histolical overview and appraisal of these arguments, see Outhwaite's"Phenomenological and Hermeneutic Approaches" (2007). Naturalisticresponses to this critique include Rudner "On the Objectivity of the SocialSciences" in Rudner (1966), and Føllesdal, "Hermeneutics and theHypothetico-Deductive Method" (1979).

Geertz's essays "Thick Description: Towards an Interpretive Theory ofCulture" and "Deep Play: Notes on the Balinese Cockfight" are importantprogrammatic statements of contemporary interpretivism (both in Geertz1973a), and rVinclis Tbe ldea of a Social Science (1958) was tleated as aphilosophical ally of the position. Critiques relevant to the topics of thischapter are Jarvie, "Understanding and Explanation in the Social Sciences"(in Jarvie 1972), Roth, "Pseudoploblems in Social Science: The Myth ofMeaning Realism" (in Roth 1987), Kincaid, 'A Science of Interpretation?"

Discussion Questions

1. Compare instlumentalist and realist interprerarions of the con-struct "aggression." \Øhich view seems the most plausible to you?\X/hich view can help understand whether violent video gamescause aggression?

2. Both the United Srares census (discussed in Section 2.1) andresearch on aggression require measurement. \What are the differ-ences between measuring 'ãggressive personality'' and measulingthe numbel of people in a country?

3. Schutz's postulate of adequacy holds that social scientific conceptsmust be closely related to those used by the subjects. Can sur.vèysbe designed so rhar they satisfy Schutz's postulate of adequacy?Must they be so designed? \Øhy?

4. To what extent cen (or shoulcl) social science discover features ofa society that are hidden from the members of that sociery? \Øhatwould be an example of such hidden social structures? Vouldyoul example violate Schutzt postulate of adequacy?

5. Does thick description lequire that the gloup under. study beclearly delineated? Can it be applied in modern societies wheregloup identity is fluid, and where people often inhabit multiplecultures, ethnicities, races, or genders?

6. It was atgued in Sections 3.3 and 3.4 thar interpretation cannorbe value-neutral. Does this form of value-ladenness weaken theepistemic status of interpretations?

Documents

in relationship beingit b - courses.helsinki.fi · like the translation ol interpretation of a text. ... The Buss-Perry Aggression ... It has 29 statements, and the subjects are asked