Eight—
"Observing" Political Culture

Author's Note : This essay was completed just before this book was prepared and is published here for the first time. Because it deals with the second main criticism of "culturalist" theory, I have adjoined it to the chapter outlining a culturalist theory of political change. However, it deals with a major methodological issue and thus also belongs in the earlier section on "Political Science."

Introduction

The political-culture approach to theory and explanation has, in just one generation, passed from undoubted domination of macropolitical studies, to doubt or rejection, to what appears to be an early renaissance—even if in more or less modified forms.^[1]

Doubts about the approach no doubt arose from its too-quick popularity, its rapidly acquired faddishness.^[2] More important, they arose for two much more serious reasons: the apparent inability of culturalists to account for political change without cheating; and the fact that culturalists made fundamental something subjective, hence not immediately observable, which they variously called orientations to action, predispositions, or (dubiously) attitudes. In essence, they made central to the construction of "positive" political theory, akin to theories in the natural sciences, what had long been considered the principal barrier to doing so by the German Geistes - or Kulturwissenschaftler (cultural scientists, as conceived by Dilthey, Windelband, and Rickert) who influenced Max Weber's conception of verstehende Soziologie . This barrier was the meaning, or understanding, of experiences and actions to actors (cognitive, affective, evaluative, as usually

broken down), governed by "rules" that are always culturally defined,^[3] thus probably unique, and acquired by learning within cultures.^[4] One reaction to this emphasis on subjectivity (on highly variable "lifeworlds," as Husseri called them) was a sort of crusading antipositivism: the reduction of social science to the mere "interpretation" of cases. The rejection of social science on the ground of its subjectivity was pushed under more or less fancy labels: hermeneutics, phenomenology, semiotics, ethnomethodology, or "thick description."^[5] In able hands, like Geertz's, studies in this vein produced fascinating information and striking insights, but at extreme costs in reliability and validity (in their technical senses), not to mention parsimony and predictive power. Compared to earlier reactions to the philosophers of cultural science, especially Weber's, the later reactions argued, in effect, for throwing out both baby and bath water.

A second reaction also was extreme. In gist, it amounted to solving the problem by circumventing it. For the most part, this was done by rational-choice theorists, who assumed a universal disposition toward "efficient" action in cost-benefit terms. Culturally learned preferences would, at some point, have to be poured into utility functions, but not necessarily and certainly not initially.^[6] To a lesser extent, and less clearly, the difficulty also was evaded by the various forms of "structuralism," drawn from the diverse sources of Lévi-Strauss, Piaget, and, of course, structural linguistics. We might note, however, that structuralism, rather strangely, has been symbiotically related to the "phenomenological" reaction.^[7]

Now a substantial return to culturalist theory seems to be in process. But this is not due to the fact that culturalists have solved the decisive problems that caused earlier doubts about their approach. The renaissance of culturalist theory appears, in most cases, to be more a reaction against the earlier reactions. The revival of the approach occurred, in part, because "phenomenology," for all intents and purposes, abandoned positive theory, and also because clever formalistic exercises upon rational-choice assumptions are still a far cry from validity (and have been empirically impugned—e.g., in coalition-theory) as well as being based on dubious, or highly time-bound and, of all things, culture-bound assumptions. To justify the renaissance of culturalism, more is needed than the shortcomings of other modes of analysis. The problems that made culturalism doubtful must be plausibly solved, if we are not to go through this cycle of rejection and counterrejection all over again.

In an earlier essay I tried to deal with the issue of political change within the culturalist framework (see chapter 7). Here, I want to confront the second issue: how to deal with "meaning" for purposes of "positive theory."^[8]

The Subjectivity Problem

In the work that started political-culture theory, Almond and Verba, after granting that culture is a word overburdened with meanings in social study, define it succinctly as "psychological orientation toward social objects . . . the political system as internalized in the cognitions, feelings, and evaluations of its population,"^[9] instilled by socialization. Despite the large variety of meanings attached to the concept, before and since, "psychological" (that is, subjective) orientation has always been central in the idea of culture. To be sure, terminologies differ. For instance, Geertz writes about instilled "programs" that govern behavior, as against behavior per se,^[10] and Barnes identifies culture with "shared assumptions about how the world works and should work."^[11] Kaase used terms like Einstellungen (attitudes) and Prädispositionen as central to culture, as against Verhalten (behavior) or Handeln (acting).^[12]

Recent revised versions of political-culture theory retain the centrality of the subjective. The latest of these versions is "grid-group" culture theory, first developed in the work of Mary Douglas and in Douglas and Wildavsky on risk,^[13] and fully developed in a forthcoming work by a team headed by Wildavsky.^[14] "Grid-group" cultural theory à la Douglas and Wildavsky essentially does two things. It posits, first, a small number (five) of general clusters of "preferences" for ways of political life, to reduce (make "parsimonious") the vast variety of particular cultural traits and themes; Almond and Verba also did this, differently, in their work of 1963. Second, it proposes a theory of political viability: polities can survive if, and only if, social preferences and social relations are congruent. The preferences (viz., "ways of looking at the world") manifestly are general normative dispositions and thus subjective. It seems plausible to hold that the perception of congruence between preferences and actual relations also will largely be culturally defined, by cognitive dispositions—like those that led French revolutionary crowds to rally to the king, as the (perceived) people's guardian.

Political-culture theories, then, in all forms, focus inquiry on subjective matter. Anything subjective can, at best, only be indirectly observed in the phenomena we experience. This raises questions. Can subjective dispositions be observed at all? (Reichel,^[15] for example, has severe doubts about this; he holds that the decisive problem in political-culture research is "uncertainty": that is, problems of determining just what empirical observations are to be considered as manifestations of political culture.) If dispositions are to be considered observable, then how? And how rigorously can they be observed: how reliably and validly, in the technical methodological senses of these terms?

We may be tempted to dismiss these problems too easily, in the manner of Hartz on the role of "ideas" as data in political research.^[16] Very briefly, Hartz argues that positivist study is study of the "existential" (what exists); ideas exist; ergo, political ideas are as appropriate a subject for positivist study as any other: quod erat demonstrandum . This is simplistic. Hartz happened to be concerned with ideas that are the observable artifacts produced by political philosophers—with what the old-style behaviorist, Bentley,^[17] called "writing-activities." These undoubtedly "exist" and are directly observable, even if often hard to decipher correctly. But the matter is not so simple when one deals with silent and implicit orientations, dispositions, perceptions, especially on the macrolevel; they are not to be found in libraries or bookstores. There may be strongly suggestive evidence that orientations to action do "exist"; Hyman's seminal work on political socialization,^[18] among many others, provides heaps of it. But a version of Lord Hewart's dictum about justice applies here: facts, for positive study and theory, must not only exist; they must be seen to exist; they must be observable facts. The issue is not put correctly by Hartz. It is not whether ideas are facts, but whether they are observable and rigorously so.

In regard to cultural dispositions, the old behaviorists had a point not to be ignored. They held that one can observe only objective behavior and objects, and that ideas, attitudes, dispositions are, in Bentley's pungent words, nothing but "mind-stuff," "soul-stuff," or "spooks" imputed to behavior. Note that this was never meant to say that people have no subjective dispositions. It meant that if inquiry focused on the dispositions it must inevitably be infirm (unbestimmt , in Reichel's word), especially; to a considerable extent, arbitrary and intuitive. Not the least difficulty discerned was that subjective explanations would tend to be truistic and, in the philosophical sense, trivial (mere words, saying nothing). The danger of triviality in political-culture studies is in fact serious. It exists chiefly in the tendency of culturalists to state observed patterns of behavior in a cultural language and then to claim that doing so somehow explains the patterns. It has, for instance, been observed that some countries have multiparty systems. Why? A typical political-culture explanation is that the countries have fragmented political cultures. What is the evidence for this? Often, in large part, that they have multiparty systems. In the same vein, some countries experience frequent political violence, supposedly because they have "cultures of violence," for which the chief evidence is that they experience frequent political violence. The fallacy in this sort of "explanation" is evident. Political-culture theories tend to be circular: statements in which an explanandum (something to be explained) is identical with a proposed explanans (an explanation). Otherwise put, political-culture ex-

planations have tended to be translations into a special language, and no more.

The subjectivity problem that afflicts political-culture studies thus involves two issues: (1) Either political culture simply is a pattern of political behavior in a society or subsociety—in which case, why talk about political culture in the first place? Or (2) political culture is something else: the dispositions that underlie the activities and generally pattern them. In the latter case, back to square one: how can one observe the general dispositions with reasonable assurance of validity?

Political Culture and Survey Research

The standard solution of the subjectivity problem in political-culture research has been to use a special technique that supposedly taps orientations directly: survey research into political attitudes. The leading macrolevel example is by Almond and Verba,^[19] whose study of democratic cultures was largely based on an interview schedule of 131 items, many subdivided into numerous subitems: the schedule occupies twenty-three large pages of an appendix to The Civic Culture . Putnam's excellent book on the "beliefs" of politicians in Britain and Italy is another case in point, using both a large number of closed questions and the coding of replies to openended ones.^[20] A plethora of other examples could be given. The most intriguing, perhaps, is a sort of canned, standardized questionnaire proposed by Berg-Schlosser for determining the political cultures of any and all societies (which, if it made sense, would, of course, be very useful for comparative study).^[21]

The idea that underlies the survey-research solution of the subjectivity problem is familiar. One uses artfully devised questionnaires and interview schedules to obtain data considered to reveal unmistakably and directly people's general political orientations or more specific attitudes. Such research is considered particularly appropriate if one deals with large aggregates of respondents, so that psychological testing or even more arduous techniques for getting at perceptual maps are not feasible.

Much is said for the usefulness of survey research in numerous studies.^[22] As a solution to the subjectivity problem in political-culture research, however, the technique poses substantial difficulties, two of which, I think, are insuperable. I deal with most of them only briefly here because the matter is much discussed in the cookbook literature on social science methodology.

First, there is a practical difficulty: survey research is extremely costly. The resources required for it are unlikely to be available to most researchers. One should not dismiss this difficulty because it is only "practical," because the social sciences are, and are likely to remain, the poor relations of research funding. Sometimes it prevents research altogether; often it

forces researchers to make do with what can be done, not what they ought to do.

A second difficulty, practical but less readily soluble, involves access to respondents. Not all polities are open to survey research (especially by Americans), some by law (as in the case of Burma), more often because of obstacles by officials unable to distinguish between research and intelligence work, inquiry and snooping. In some cases, also, potential respondents are too fearful and suspicious to be good subjects for surveys.

A third difficulty, about which much has been written, is intrinsic in survey-research technique: there are nearly always doubts about the trustworthiness of responses to even the best designed survey instruments. Converse has discussed one especially serious reason for this,^[23] the expression of "nonattitudes": attitudes about matters regarding which subjects lack the knowledge to have attitudes in the first place. There are many others. Respondents, if suspicious, may give intentionally misleading responses. Or they may give answers they consider expected of them (a special kind of "nonattitude"); this is especially prevalent among school children who confuse questionnaires with tests. Or responses may vary with contextual factors; for example, there is the well-known phenomenon of deceptively high levels of political knowledge just before general elections. The wording of questions may contaminate responses, as may the presence or demeanor of researchers. And so on. Survey researchers have been ingenious at guarding against the difficulty posed by misleading responses but, of course, imperfectly so. The basic problem is not that distortions may occur, but that it is hard (indeed impossible) to know whether, and to what extent, they have occurred—unless one already knows what surveys are conducted to find out.

The absolutely insuperable difficulty posed by the survey-research solution is logical (and epistemological); that is what makes it insuperable. Responses to survey instruments still are "behavior" that only supposedly represents orientations. They undoubtedly provide more data than mere "raw observation" and conceivably more revealing data. But there is no getting around the fact that they are, in Bentleyan terms, "speakingactivities" or "writing-activities," not the subjective thing itself.

There is an ineluctable "epistemic gap" (as Northrop called it)^[24] between objective behavior and subjective dispositions. Certainly, survey research cannot bridge it; logically no observational technique can.

These arguments may seem glum, but they at least allow us to state the decisive problems in observing political culture. There are two. First, how can culturalists get more reliable, more trustworthy data than by survey research? Second, how can the apparently unbridgeable "epistemic gap" between behavior and orientations be narrowed, even if not eliminated, so that its existence makes little difference to theory?

A Proposed Solution of the Subjectivity Problem

Obtaining Data

If any single technique for getting data about a subject is likely to be flawed, it seems plain ordinary sense always to use a variety of such techniques. The results can then supplement one another; up to a point, they might cancel out errors that arise from each technique used singly; they provide checks on one another; and if they produce overlapping findings, one may be especially confident that something correct has been found. Campbell has called this procedure the use of multitrait-multimethod matrices.^[25] The method is analogous to geological "triangulation": measuring heights not otherwise measurable, or distances not readily traversable, by the use of a series of imperfect measures (of triangles from points on a baseline). That method is, in fact, sometimes used in preference to direct measuring, even where that is possible, because of its exactness.

In trying to get at culture, a considerable variety of different measures is, at least potentially, available for the similar purpose of observing the apparently unobservable. One of them, for all its imperfection, is survey research. There are many others. A large number of textbooks have been written about them. Without being comprehensive, I will discuss a few.

One method designed to get at the inherently unobservable is content analysis, a once widely used and quite sophisticated technique that seems now, unaccountably, to have lapsed into disfavor and even unfamiliarity in many cases. Its purpose was the quantitative analysis of communications, using explicity defined categories to capture their content; the use of the categories was binding in analysis to prevent the arbitrary selection of materials that might merely strike inquirers as intuitively "interesting"; and the measurement of the categories was carried out by appropriate, rigorous procedures. Newspaper editorials and speeches were the sort of materials studied, in all cases to find the "meanings" (mostly values) they conveyed. For instance, Lasswell, Leites, and their associates devised a "symbol analysis" that was used during the Second World War by several branches of the U.S. government to analyze the contents of newspapers, noting the frequency of the symbols they used and whether positive or negative meaning was attached to them.^[26] White similarly studied the prewar speeches of Hitler and Roosevelt, classifying every value statement (5,326 of them) under predefined categories.^[27] A good overview of the content-analysis technique, although it emphasizes attitudes in international relations, is provided by North et al.^[28]

Content analysis, of course, has many flaws of its own, including problems regarding the predefined variables used in carrying it out, problems in measuring them accurately, and problems of the "significance" of find-

ings because it is only too easy to get hooked on counting as such. Not least, it poses sampling problems; for instance, it seems heavily biased toward elite communications, such as speeches by conspicuous political leaders. However, its use, as stated, is recommended here only as a flawed way of deciphering meanings, among other such ways. And content analysis has undoubted advantages over survey research in getting at cultural meanings. One is that it is designed to get at implicit meanings of which subjects themselves might not be aware. Of course, details and nuances that idiographers usually catch get lost in the mechanical use of preselected variables and methods of measurement, but all techniques of generalizing, even about a particular society, pay that price—and the generalizers know it and consider it worthwhile.

A method that political science culturalists seem somehow to have missed, yet that seems tailor-made for deciphering meanings, is Osgood's "semantic differential" technique; its nature and rationale is discussed comprehensively in Osgood, Suci, and Tannenbaum's The Measurement of Meaning ,^[29] and the technique is applied, almost entirely by Osgood and his collaborators themselves, in a long-sustained series of works that covers some three decades.^[30] The technique is a type of projective technique. It is used to elicit responses to stimuli—in this case, cards that present words which are genuine opposites, like honest and dishonest, not short and tall—in regard to some object (a picture, for example) that may potentially reveal meanings with a minimum of deception by respondents (as in the case of nonattitudes) and contamination by the presence of observers.

Again, I refer to the semantic differential technique only as part of the arsenal of techniques available to get at meanings. However, that method seems to me to be especially notable because it has in fact produced enlightening findings; for instance, to cite just one, Israelis react more intensely to words denoting activity than to those that evoke strength, whereas Arab cultures seem to react intensely to cues that involve strength but not those that denote activity.^[31] On the face of it, this finding illuminates much experience. (After all, Nasser seemed genuinely surprised that his threats to drive the Israelis into the sea were taken at face value in 1967; he was, after all, only flexing muscles.)

Greenstein and Tarrow have used what they call a semiprojective technique that seems particularly suitable for studying children's dispositions because children are likely to confuse questionnaires and the like with tests, but their technique obviously has wider applicability.^[32] In gist, subjects are told the beginning of a story and then asked to continue it; their responses are then subjected to more or less systematic analysis. To illustrate, one story, designed to get at dispositions toward political authority began with this cue:"The [head of state] has to attend an important meeting. He [she] is late, so he [she] drives to the meeting over the speed

limit. A policeman stops him [her]." Here is part of how a twelve-year-old French girl continued, after considering that the president, then de Gaulle, might be issued a speeding ticket:

Girl : But they must let it pass, because it is the president. . . . He tells them he is late, that he has to go to a meeting. Perhaps the gendarmes let him go, telling him to go on less rapidly, because he could have an accident. Let's say that he pays attention to their orders, if he is afraid, or perhaps he does not pay attention to them saying to himself: "It is not right that a gendarme should give me orders." And maybe he will tell his chauffeur to speed up , or at any rate not to slow down.

Interviewer : What do you think? Would he say go slower or go faster?

Girl : Perhaps in order to be safe he slows down a little, for he says to himself: "If there is an accident, I will not be there any more to direct the country [la patrie ]."

Compare how a British working-class girl of the same age continued the story:

Girl : Well, really the police wouldn't really stop the queen driving the car because somebody would be escorting the car to the meeting and they would stop the rest of the cars so she wouldn't really need to be going fast. People have got time to wait for the queen. She wouldn't really be late.

Interviewer : Let's suppose for the purpose of this question that the queen was driving by herself in a car and she wanted to go out and get away from other people and the police stopped her for going too fast.

Girl : Well, the police might have stopped her, but as soon as they saw it was the queen they wouldn't take her in for speeding or anything like that.

Interviewer : What do you suppose the policeman would say?

Girl : I think he would be astounded, because really the queen, she doesn't drive the car on her own. And she knows Britain's speed limits, so she wouldn't have gone really fast.

Or take the way an American worker's son elaborated the scenario (not, to me, very reassuringly):

Boy : The policeman was giving him a ticket and everything. And once he saw it was the president he told him he could go and that. And the people that were standing around thought it was unfair and everything. So they started running after the car. They were pulling the bumper and that. They were wrecking it, and they forced the policeman to give him a ticket. He was late for his meeting.

Interviewer : Well, why did they think it was unfair that he didn't get a ticket?

Boy : Well, because the president is just a person and everything.

Interviewer : Well, what did the president think about all of this?

Boy : He thought it was all O.K. He thought it was fair that he should get one.

Interviewer : What does the policeman think when he sees that he has stopped the president's car?

Boy : He thinks he'd better let him go or else he'll get mad and that and start telling the chief and he'd fire him or something.

These responses surely tell one a lot, even on just an intuitive basis, and Greenstein and Tarrow do subject them to intensive analysis; but they also use rigorous quantitative techniques on the aggregates of responses.

Not to overlook the obvious, there also is "raw observation" (though we may feel better if we give it a label that sounds more dignified, like "participant observation") to provide clues to cultural meanings. A strange notion seems to be abroad in the social sciences: that such observation does not provide "data." When I used it in my own work, I was accused by a survey researcher of being merely "anecdotal"—not "scientific."^[33] Natural scientists, it seems to me, would regard this notion with disdain. Granted that they use vastly complex mechanisms and techniques of observation. They do not use them merely to observe, but to observe more and better: a necessity at a stage of development we hardly have reached. Moreover, raw observation can be carried out thoughtfully, even with considerable rigor (see the literature on participant observation). And it has several advantages over special techniques of observation designed to decipher the subjective. It is not artificial and is therefore unlikely to be either mistaken by subject or contaminated by observers; and, unlike survey research, one can do it almost anywhere.

In general, then, findings obtained by a variety of techniques always are worth more than those obtained by any single mode of observation, where all such modes are inadequate. Genuinely rigorous "triangulation," of course, requires one to know precisely the nature and extent of measurement errors in particular imperfect observations, as surveyors do. We know a little about this in regard to some of our methods; for instance, we know that, in community-power research, asking people questions like, Can you tell me who has political power here? will lead to findings that exaggerate considerably the concentration of power in the community. But we know how modes of inquiry distort observations in too few cases, and, where we know, we know too inexactly. However, the principle of the matter is clear: progress is possible; difficulty is not a good reason for not trying; and although it might be foolish to hope that we can attain the precision of geological triangulation, we can approximate it—but only if we try.

Narrowing the Epistemic Gap

The greater difficulty, the crux of the issue of whether political culture is "observable," is how to make negligible the epistemic gap between behavior and orientations (that is, meanings, dispositions, attitudes, mind-

stuff. The solution proposed here is, of course, applicable, mutatis mutandis, in all inquiries that pose the subjectivity problem.

To begin: One must never pretend that subjective orientations have been directly revealed by any technique of observation or by any set of such techniques. There simply is no way around the point that orientations are always imputed to behavior. This, however, is not (yet) a concession to the narrow behaviorists, for the simple reason that imputations may be valid. The issue we have come to can therefore be restated in a form clearly more conducive to solution: How can subjective orientations be validly imputed to behavior? In regard to this, we already have discussed the principal fallacy to be avoided if a plausible answer is to be found. It is the statement of circular explanations: the mere translation of objective observation into a cultural language. Such statements, as I have said, trivialize problems: the explanatory statements simply restate explananda ; they add nothing to them; and they allow us to deceive ourselves into thinking that we have said something when we have said nothing.

The trivialization of theories by translation and circularity is unavoidable if a particular orientation to action is imputed to any particular activity. We simply cannot say, for instance, that a person participates because he is participant. Political orientations must be regarded as dispositions that underlie, thus organize and pattern, sets of perceptions and activities: the larger and more diverse the sets the better for theory. This is inherent in the very conception of "orientations to action" anyhow. They are, as I wrote in an earlier paper, "general dispositions of actors to act in certain ways in sets of situations. [They] pattern actions. . . . Orientations are not 'attitudes': the latter are specific, the former general dispositions. Attitudes themselves derive from and express orientations; though attitudes may, through their patterning, help us find orientations."^[34] (For attitudes here one can, of course, substitute actions or behavior.) If, for example, we say that "X is tolerant," we say that X does not object (or, better, is oblivious) to sets of traits of people or of their behavior widely disapproved in a collectivity. The set might include Jews or Catholics, blacks or Asians, illegal immigrants or guest-workers, homosexuals or eccentrics, self-absorbed or stupid people, and the like; it is unlikely to cover all people or behaviors (say, child abusers: everyone draws a line somewhere), but the statement describes a general pattern of matters directly observable, or else it says nothing.

If orientations are so regarded, then statements about them or sets of them (cultures or subcultures) are statements of "regularity": they are lawlike statements or, more simply, theories. They are never descriptions, not even "thick" ones—though I do not mean to deride the clues to cultural theory that acute observation of a particular activity (say, a Balinese cockfight) may provide.

This, by the way, is rather an old view, dating to the origins of contemporary social science, although recent theorists of culture, and others who stress "meaning" in social inquiry, seem unaware of that fact. Weber, the apostle of verstehende Soziologie , for example, wrote, in his posthumous magnum opus:

Every interpretation aims at self-evidence or immediate plausibility. But an interpretation which makes the meaning of a piece of behavior as selfevidently obvious as you like cannot claim just on that account to be the causally valid interpretation as well. In itself it is nothing more than a particularly plausible hypothesis.^[35]

Weber's point was restated by Morris Ginsburg in 1956:

It appears to be a basic assumption of verstehende Soziologie  . . . that what we know within our minds is somehow more intelligible than what is outwardly observed. But this is to confuse the familiar with the intelligible. There is no inner sense establishing connexions between inner facts by direct intuition. Such connexions are in fact empirical generalizations , of no greater validity than the similar generalizations relating to outward facts.^[36]

This has a crucial implication, which Weber also pointed out. If one is to have confidence in the validity of statements about orientations, then they must be treated like any other hypothetical assertion. At the outset, statements about culture may simply be summaries of observations: verbal summaries or the scientific-looking ones we find in regression tables, frequency curves, histograms, smoothed ogives, and so on. Political culturists usually stop at that point—if they get to it. But theory, to be considered valid, must be tested; and statements about cultures, or anything else, cannot be tested by the data from which they were derived in the first place—even if the nature of a frequency distribution, or our intense immersion in a culture as participant observers, may justify the intuitive surmise that tests will validate or invalidate, so that they might not be worth the cost of being carried out.

If one cannot test theoretical statements by data used to formulate them, it follows that one must posit, by logically deduced prediction, as yet unobserved data that should turn up, given certain "initial (or determining) conditions,"^[37] if our assertions about orientations or cultural sets of them are to be considered valid. (Weber mistakenly thought that the validity of imputed meanings can only be determined by psychological tests—which are still a way of getting directly at the subjective.) Such predictive tests, positing unknowns, can be carried out in two ways.

First, one may predict behavior in situations either that have not yet occurred or, if they have, to which the responses are still uncertain. I will give a real-life example (that happened to turn out well) from a work of

mine that, as mentioned above, was criticized by an eminent authority for using "anecdotal material."^[38]

In my study of Norwegian authority-culture, I argued that Norwegians are consensualists in decision making. More specifically, it appeared that they follow this decision rule: agree widely (unanimously, if possible); if wide agreement is not possible, then appear to be in wide agreement; if that also is not possible, then drop the subject. This is a statement about subjective cognitions—"actions widely agreed upon are likely to be right actions"; about affect—Norwegians "feel" uncomfortable with small majorities; and about evaluation—Norwegians value wide agreement, and processes of forming consensus thereby become their decision processes. Thus, the statement contains all the elements into which Almond and Verba break down orientations.^[39]

The statement is a (low-level) theoretical generalization, based on a large and eclectic variety of mostly raw observations, from the behavior of small neighborhood and club committees to transactions of parliamentary business. Admittedly, the generalization was illustrated by what may accurately be described as "anecdotes." In an essay replying to this point, among others,^[40] I pleaded guilty to anecdotalism but also pointed out that the generalization was based on much else; that empirical generalizations should be based on the best obtainable data; but that despite their basis, what really counted in the end was how such generalizations stood up to appropriate testing by the accurate prediction of unknowns.

As it happened, a "natural" occasion for such prediction had arisen about the time that this debate was going on. In 1965, a general election brought to power what Norwegians called a bourgeois coalition, after thirty years of social-democratic (Labor party) rule. This coalition consisted of a very odd assortment of parties that had in common perhaps only opposition to the Labor party's long dominance. The coalition combined urban, and urbane, business interests with an agrarian party (in a country in which the rural-urban split looms large); nonreligious, if not downright antireligious, parties joined a Christian fundamentalist party; and a small party, chiefly of white-collar intellectual liberals, was added to the odd alignment. The common wisdom in political science is that such coalitions cannot last. They defy the now common rule that coalitions must minimize "preference disagreement," or "policy distance," merely to form.^[41] The common wisdom in Norway, academic and general, was the same. However, the common notions ignored the decision rule I have summarized. That rule led to a quite different expectation, stated in my paper of 1967:^[42] the coalition would endure by simply avoiding divisive issues (like religion or urban and rural interests likely to clash); it would work smoothly by concentrating on the least divisive issues (such as almost universally desired

tax reforms); it would do little or nothing to change the Labor party's welfare-state legislation (Labor still had about 40 percent of the parliamentary seats—hardly a negligible minority); and it would break down if absolutely forced by circumstances to deal with a deeply divisive issue (for example, an issue that involved what Norwegians call "cosmopolitanism": close connections with the continent, divisions about which originated, no doubt, in the long colonial domination of Norway by Denmark). As it happened, all these predictions turned out to be correct. They thus validated—failed to falsify, despite defying both plausible theory and sheer good sense—the cultural generalization. The coalition lasted, although it did little besides tax reform, until the question of joining the EEC came up as an unavoidable issue; it then disintegrated.

The case was, in one way, lucky. Something happened in Norway at just the right time to allow a "natural experiment." That probably is only rarely the case. Usually, one may surmise, similar tests do not yet exist; they must be awaited; and they may simply not turn up, anyway not for a long time—quite as in the natural sciences.

Because that is so, a second way of testing whether cultural "meaning" has been validly deciphered may be desired over the first. One may carry out such tests by predicting responses to stimuli controlled by inquirers—not least, by doing survey research into as-yet-unknown responses by logically predicting the findings that should turn up if one's conceptions about a culture are valid. If in fact one has conceptions of culture, however arrived at—multitrait-multimethod "triangulation" is still best—then it should not be difficult to devise questions or other stimuli to test the conceptions. This is the best use, in my view, to which survey research intended to get at orientations can be put, both epistemologically and practically. Models of political culture can be built in all sons of ways, less expensive in funds, time, personnel, and so on, than typical survey researches. Questions can be few if carefully devised for testing (especially if not used in attitudinal fishing expeditions), and samples may be small if thoughtfully selected for a carefully defined purpose. Unfortunately, neither survey research nor other modes of inquiry into culture have ever been so used. Their findings have been treated as definitive results, not as steps toward such results. But survey research, and other techniques, could be so used, at low cost and with potentially high payoffs.

Perhaps one other step that is desirable, even if not absolutely necessary, still is missing. What, in effect, we have, if all the above is done, is simply a tested hypothesis, essentially like any other. To narrow the epistemic gap between meaning and observation, it would still be useful to establish more firmly that a regularity exists because of subjective orientations rather than other factors. One can do this in at least three related ways: by multivariate

analysis intended to establish the relative strengths of direct contextual factors (e.g., SES) as against factors associated with processes of socialization; by comparing the strength of synchronic and diachronic factors in mature subjects (because orientations are supposedly instilled chiefly by experiences early in life and upon my argument are highly resistant to change);^[43] and by observations that might, or might not, establish that adult behavior is relatively stable despite changes in its context: something already done, by summarizing much psychological research, in Herbert Hyman's seminal work on political socialization.

Conclusion

All this is not meant to say that we can directly observe subjective matter by following the recipe here proposed. The epistemic gap between "inner" and "outward" facts, as stated, cannot be more than narrowed. But surely that is of no great consequence for the normal purposes of everyday academic inquiry: constructing, testing, and using theories.

It is especially worth adding that the solution of the subjectivity problem here proposed is exactly similar to how natural scientists treat unobservables, like time, electricity, or magnetism. No one in the "hard" sciences shies away from such concepts simply because they are not directly observable but experienced only through their effects—and in that sense are "spooky." They are, however, treated as theoretical constructs; their nature is imputed to a large number and variety of observations, using various research procedures; and conceptions of their nature have been subjected to careful tests. By now, they seem "real" in the objective sense; one would no doubt be held for an idiot if one were to contest paying one's electricity bills on the ground that one should not be charged for figmental "mindstuff." The construct works in all sorts of ways, and so we, and the utility companies, may safely consider it "real," for our, and their, purposes. Indeed, we must—or else dismiss much good theory.

I make this point to emphasize the assertion that nothing about positivist inquiry compels dealing only with direct observables. That Skinnerian view belongs to primitive and naive positivism; its resemblance to what goes on in the hardest sciences is not even coincidental. Nor does cultural meaning condemn us to hermeneutics or any of its relatives. In "observing" culture one must comply with only one imperative, though it is burdensome: one must be highly cautious, ingenious, rigorous, and, above all, skeptical about one's own ideas. But this is imperative in all researches.