Michael A. Faia, College of William & Mary 1993

For Caitlin, Josephine, Gusty, and Pancho

Et si je connais, moi, une fleur unique au monde, qui n'existe nulle part, sauf dans ma planète, et qu'un petit mouton peut anéantir d'un seul coup, comme ça, un matin, sans se rendre compte de ce qu'il fait, ce n'est pas important ça!

--Antoine de Saint-Exupéry

Table of Contents
 

Introduction
 

Chapter 1: WHAT'S WRONG WITH THE SOCIAL SCIENCES?   1
 

(1) The dialectics of disenchantment 2
 

(1.1) Predictability, vulcanology, seismology, and (especially) meteorology 6
 

(1.2) Molecular mysteries and habits of the quark 8
 

(1.3) Rosaldo revisited: How would the catcher in the rye have felt? 12
 

(2) The trouble with feminist theory 13
 

(3) Titles and tribulations 17
 

(4) Solicitous gatekeepers, #1 25
 

(5) Itching and scratching: a Lazarsfeldian digression through an SPSS 27 hologram
 

(6) Transcending the transcendentalists 30
 

(7) What do you presuppose, and when did you presuppose it? The Sisyphus of the social sciences 33
 

(8) The meaning of politics and the politics of meaning 38
 

Chapter 2: MICHEL FOUCAULT, MACHINES WHO THINK, AND THE HUMAN SCIENCES   57
 

(1) "Man the machine--man the impersonal engine" 57
 

(2) Good, bad, or ugly? 67
 

(3) Mitigating circumstances 70
 

(4) In conclusion: Oodles of Boodles 74
 

(5) Why Foucault needed Lindroth, and why Lindroth needed Foucault 75
 

Chapter 3: IN PRAISE OF THE NULL HYPOTHESIS: THE MYTH OF "THE VALUE-FREE MYTH"   83
 

(1) The nature and extent of bias in scientific research 83
 

(2) Quashing the indictment: Can a Comtean rule a country? 85
 

(2.1) Positions 86
 

(2.2) Correctives 87
 

(2.3) Motives: A series of acts contrary ... 88
 

(3) Quashing the indictment: The case against functional analysis 89
 

(3.1) From Spencer to Weber 90
 

(3.2) Pareto 94
 

(3.3) Malinowski and Radcliffe-Brown 95
 

(3.4) Parsons 96
 

(4) Overturned on appeal: Research on political power 101
 

(5) A new hypothesis 102
 

(6) Conclusion: Science as bias control 102
 

Chapter 4: SOCIAL SCIENCE SYNTACTICS AND EMPIRICS: ALLEGED PROBLEMS ...   111
 

(1) Three papers in search of an argument 111
 

(1.1) The causal conundrum 112
 

(1.2) A transcendence of triangulation 120
 

(1.3) The contours of quantiphobia 127
 

(2) Solicitous gatekeepers, #2 131
 

Chapter 5: ... AND HAPPY PROSPECTS   141
 

(1) Introduction 141
 

(2) Sex discrimination in faculty salaries: Toward projections of the middle range 145
 

(3) Allocating faculty salaries with Minitab BASIC, and allowing feedback from future states 154
 

(4) From a FIRM grip on reality to the AIDS epidemic: The future of social simulations 156
 

(5) Conclusions 159
 

Chapter 6: CULTURAL MATERIALISM IN THE SYSTEMIC MODE   171
 

(1) An excursion into "cultural" functionalism 171
 

(1.1) An emergent consensus 172
 

(1.2) Sub-cranial worlds and other pitfalls 174
 

(2) The POET paradigm: A transcendence of hierarchy 178
 

(3) Reciprocal interaction: Organization and technology 179
 

(3.1) Wagons, waterways, and William F. Ogburn 179
 

(3.2) A functionalist overview of an ecosystem 186
 

Chapter 7: SOCIAL SCIENCE SEMANTICS: A SAD STORY   195
 

(1) Introduction 195
 

(2) Contemporary thesauri with taxonomic overtones 200
 

(3) The logic of taxonomies 205
 

(4) Sampling, search, and test strategies for Sociological Abstracts: Hierarchies without perks 209
 

(5) First conclusion 214
 

(6) Codicil 1991 215
 

(7) Frequencies and amplitudes: Toward a taxonomy of time series 216
 

(8) Online searches as a data source: Notes on the future of taxonomy 219
 

(8.1) Data reduction: Obtaining simple counts 221
 

(8.2) Data reduction: The REPORT command 222
 

(9) Final conclusion 223
 

Index

Introduction

The contention of this book is that the contemporary social sciences in America have the qualities of a 1936 Cadillac V-12 sedan making a belated trip to the repair shop: The classic lines, the elegance, the power; the smooth, sonorous forward thrust usually evident, but momentarily interrupted by an assortment of minor failures and afflictions, as yet undiagnosed, brought about by many years of not entirely benign neglect, by a shortage of replacement parts, and by bad luck; the unhappy proprietors vaguely conscious of flashier models whizzing past wildly, their half-crazed occupants oblivious of the nature of real workmanship, real quality, real substance, real inspiration, real art; and finally the hope, the faith, that a few new wires here and there, a better thermostat that no longer permits the cooling system to overheat, a little grease pumped into the strange sounding transmission, a few carefully sewn patches over nicely tucked but tattered upholstery--and once again we take to the highways, once again we have a touring car that gives us, and gives our passengers, free and easy access to exotic places that we do not ordinarily expect to see. As far as I'm concerned there never will be a trade-in: The new models do not impress me, I love the old Caddy, and I'm willing to ante up in order to get her fixed.

By 1936, the year of my vintage Cadillac, market segmentation had not yet been discovered (or invented). Detroit had not yet begun producing cars designed mainly for the female market, let alone cars for white female yuppies earning $70,000 or more per year, or cars for rural retired elderly gentlemen of Latin-American extraction and with a need for high motility in a harsh climate, or cars for wives of young black blue-collar aristocrats with strong Republican proclivities, an overweight mortgage, and two kids in private schools. The 1936 Cadillac was classy, and it was also androgynous: It combined the best of the masculine with the best of the feminine. Achieving such a combination is one of the central intentions of this book, which originally was to have been entitled Androgynous Sociology.

Chapters 1 and 2 argue that, for mysterious reasons, a vociferous anti-science movement has developed in recent years among students of society. While the size, structure, and sources of this movement remain obscure--although I think feminist theory and qualitative methodology have given it impetus--it has attracted the support of several influential scholars within the social science disciplines. The movement may be an attempted adaptation to a contemporary state of affairs in which, according to a recent President of the American Sociological Association, the field finds itself "in the doldrums." These opening chapters anticipate remaining parts of the book by arguing (1) that the social sciences already have achieved successes that rival those of the natural sciences, and that the current round of self-criticism, bordering on self-flagellation, is therefore an inexplicable failure of nerve; (2) that anti-science, arising in part from math phobia, is an indirect and unjustifiable attack on quantitative structuralism; (3) that advocates of anti-science wish to focus primarily on questions that lie outside the realm of science, or that constitute the preliminaries of scientific inquiry, or that constitute only one aspect of scientific inquiry; (4) that advocates of anti-science misconstrue the similarities and differences between scientific and non-scientific thought patterns, and do not appreciate the interactions and complementarities of these thought patterns; (5) that the anti-science movement wrongly asserts an ineluctable tension between science and "interpretationism," the latter exemplified by scholars such as Foucault, with their strong nominalist and logocentric tendencies; and (6) that the embattled advocates of interpretationism and structuralism should declare an immediate and open-ended truce.

Chapter 3, derived from my own brief history of functionalist theory (Faia, 1986: appendix), asserts that in recent years we have made a shibboleth of the belief that social science theories are inescapably invaded, pervaded, and distorted by one's values, ideology, social class background, current class situation, age, sex, race, and so forth; in brief, by one's "biases." Sometimes the process is alleged to operate in reverse: One's theoretical commitments, for instance, are said to influence one's ideology. Because shibboleths tend to terminate thought, hardly anybody bothers to define and distinguish among the central terms of the discourse, such as theory, values, ideology, class background, gender. Nor is there strong concern about testing the shibboleth as if it were just another hypothesis, which it is.

I have found an intriguing way of illustrating my argument, presented mainly in Chapter 3, about bias. This illustration uses the idea of triangulation, an idea that is to be of central importance throughout this book. When seismologists try to find the epicenter of an earthquake, they use triangulation. A given seismographic observation station, analyzing the various wave series set in motion through the earth's crust by an earthquake, ascertains the distance from the station to the epicenter; the direction cannot be ascertained. A circle is then drawn around the station, with a radius equal to this measured distance, and it is assumed that the circle, at some point, must touch the epicenter. If a second station follows the same procedure, the two circles usually will intersect at two points, one of which is the epicenter. A third station, swinging a similar arc, then decides between these two intersections. If the process does not work, then presumably somebody's measurements are off; or, something may be wrong with theories about earthquake foci, epicenters, seismic wave behavior, etc.¹

Now, go to your blackboard. Draw a diagram of this elaborate search process. You will see that three seismographic stations do indeed enable us to find an epicenter. Then, erase one of the stations, along with the circle drawn around it. Using information from the other two stations, and retaining your knowledge of the location of the epicenter, try to determine the position of the third seismographic station. Voilà! It is harder to find than Graham Greene's third man; in fact, one cannot find it. My (null) hypothesis, then, derived from geology, is that the perspectives adopted as we interpret the real world, as we try to locate what we consider the major social epicenters, will not necessarily tell our readers anything about our own location, our own presuppositions, even if readers have an epicenter and two snitches.

There are excellent social scientists who have repeated the truism about large and inescapable biases so many times that they now accept it as gospel, virtually without reservation, and my fear is that the habit tends to induce give-up-itis, or at least to contribute to the failure of nerve mentioned above. Why should we insist that our science has any special merit, when it is as fallible as any other system of thought? In Alan Sica's (1988:257) sad claim--the sadness is mitigated by the fact that Sica is probably wrong--social scientists do not know much more about human fertility than do the more militant members of pro-choice or pro-life movements.

Scientists are people who have special skills in semantics, syntactics, and empirics (Hepp, 1956). In Chapter 4 we argue that although social scientists are relatively weak in the area of semantics--the discipline of definition and measurement--we are surprisingly sophisticated in the realms of syntactics and empirics. We have considerable skill in the use of deductive models of all sorts from Aristotelian syllogisms to the stable population model to modern statistics to kinship networks to the propositional calculi of learning theory. And in the realm of empirics, social scientists know a great deal about how to observe the world, about how to register social phenomena such as fertility, mortality, migration, marriage, divorce, crime, contagious disease, abortion, and economic activity, about how to conduct social surveys of all varieties from censuses to political "exit polls," and about how to make plausible estimates, say, of the North American native population of 1492 (Thornton, 1987).

Among our most highly developed deductive models are those which enable us to make causal inferences, and this chapter explores the issues raised by modern scholars who suggest that the pursuit of causal explanation in the social sciences ought to be played down, if not abandoned.

In Chapter 5, we discuss "reverse" regression models and "forward" regression models--the latter often called simulations--as ways of testing causal ideas. For instance, ideas about ways in which the sex preselection of children might influence the sex ratio, ideas about monetary rewards and their allocation in bureaucratic settings such as the modern university, or the idea that it is possible to design a highly efficient, "optimal" traffic-control system for a central city. We then develop the argument that one of the major advantages of the American federated system, and perhaps of the new commonwealth that has replaced the Soviet Union, is that these entities may provide many jurisdictions willing to carry out their own unique experiments. Modern nations and commonwealths should be conducting systematic social science experiments in large numbers, and we should be introducing the results of these experiments into our best social theories while simultaneously seeking opportunities to conduct "social engineering" (Henshel, 1976). We are not yet very good at social engineering, but we are much better than those who currently have the responsibility for it (Dye, 1990).

Chapter 6 argues that the best social science research of the future will make use of biosociological and cultural materialist perspectives within a functionalist/systemic theoretical framework, with appropriate attention to experimentation, to simulations that tie together the past, present, and future, and to interpretationism. This chapter tries to clinch a central argument: The most promising works of the social sciences excel primarily because they triangulate. They do so by combining a satisfactory conceptual scheme, appropriate empirics, and appropriate syntactics; by making sure that the syntactics of a given hypothesis are capable of predicting a fairly wide range of testable consequences; by using an assortment of instruments and methods² to test a given hypothesis and its consequences; by invoking existing knowledge from several related scientific fields; by presenting findings in a way that exemplifies the skillful use of several different symbol systems--verbal systems, quantitative systems, those that are both verbal and quantitative (e.g., programming languages), and those that are neither (e.g., graphics, motion pictures).

The major purpose of chapter 7 is to evaluate the semantics of the social sciences by ascertaining whether the Lazarsfeld generalization-specification model, also known as the "elaboration model," has been routinely and effectively used by social science researchers. If the Lazarsfeld model were widely exploited, then highly abstract assertions--e.g., "factors of type X tend to have a causal impact on variables of type Y"--would be tested for their applicability to a broad range of specific instances of both X and Y.³ For example, the literature discusses a number of situations in which deterrence allegedly increases conformity; generalizations have been derived from these specifications, and the generalizations seem to produce new specifications.

Using the Sociological Abstracts thesaurus of sociological concepts, in combination with online searches of the social science literature, this chapter assesses the extent to which researchers use systematic taxonomies and exploit them through application of generalization-specification strategies. Briefly, the chapter raises questions such as the following: First, do social science researchers and teachers, in a popular metaphor, tend "to mix apples and oranges"--and perhaps an occasional persimmon? Secondly, when they do so, do they arrive at generalizations applicable to broad categories? Thirdly, do they then test new hypotheses about narrowly delimited instances of these broad categories?

Chapter 7 concludes that, despite SA's listing of "broader" and "narrower" terms in relation to a given thesaurus entry, there is little evidence of the use of highly structured taxonomies in the social science literature.

When I speak of the structure of a taxonomy, I definitely intend the implication that the concepts of a scientific discipline, as Durkheim suggests, should have a sort of social organization. Durkheim would surely agree that a single concept has a lot in common with a single personality: It is an abstraction that cannot be clearly understood until it is related to the structures that surround it. The chapter concludes further that in creating workable taxonomies we would do well to begin with the limited classification schemes already available in fields such as "ecological demography" (Namboodiri, 1988), and that it would be wise for us to develop, early on, a taxonomy of time-series processes; the chapter then outlines preliminary steps toward both these objectives. The final section suggests that social science data of the future, gathered primarily by people who are not themselves social scientists, will come to us largely by way of telephone lines, optical cable, and satellite transmissions, and that we need to try to influence the organization of these masses of data.
 

NOTES (Introduction)
 

(1) I understand that it is also possible to find a "shadow zone," 180 degrees away from an epicenter, where seismic waves suddenly disappear.

Spencer (1910:83) gives a similar example: While standing by a lakeside in the moonlight, one sees "... flashes from the sides of separate wavelets ...," but one's "... position in relation to certain wavelets brings into view their reflections of the moon's light, while it keeps out of view the like reflections from all other wavelets."
 

(2) For instance, one would alternate between personal interviews and mailed questionnaires; among surveys and laboratory experiments and library investigations; between exploratory methods and approaches that are highly structured in their use of multivariate causal models, feedback, time-series data, etc. For a discussion of the relationship between theoretical and methodological aspects of research, see Faia (1986:128-41).
 

(3) When we extend the scope of a generalization, we are in essence applying Mill's method of agreement. In a paper written some years ago (Faia, 1968:117-8) I argued that while we make substantial use of Mill's method of concomitant variation (correlation) and his method of difference, we neglect the method of agreement. This claim is close to one of the central theses of this book: That while we have high skill in the realms of empirics and syntactics, we remain weak on the forms of conceptualization that support the method of agreement.
 

REFERENCES (Introduction)
 

Dye, Thomas R. 1990. Who's Running America? Englewood Cliffs: Prentice Hall.
 

Faia, Michael A. 1968. "On Alienation, Structural Strain, and Deviancy: A

Reply." Social Problems 16:117-20.
 

Faia, Michael A. 1986. Dynamic Functionalism: Strategy and Tactics. Cambridge University Press.
 

Henshel, Richard L. 1976. On the Future of Social Prediction. Indianapolis: Bobbs-Merrill.
 

Hepp, Maylon H. 1956. Thinking Things Through: An Introduction to Logic. New York: Scribner's.
 

Namboodiri, Krishnan. 1988. "Ecological demography: its place in sociology." American Sociological Review 53:619-33.
 

Spencer, Herbert. 1910. The Study of Sociology. New York: D. Appleton and Co.
 

Thornton, Russell. 1987. American Indian Holocaust and Survival: A Population History Since 1492. Norman: University of Oklahoma Press.
 

Wildavsky, Aaron. 1988. Searching for Safety. New Brunswick: Transaction.
 

Chapter 1

WHAT'S WRONG WITH THE SOCIAL SCIENCES?
 
 
 
 
 
 
 
 
 
 
 
 

The current controversy about science versus the humanities and the arts is ... quite absurd. The assumption seems to be that the advancement of science and especially of social science must be at the expense of the other intellectual, artistic, and religious pursuits of man. This is a preposterous assumption. It implies that human knowledge is a finite quantity so that any increase in one department must be at the expense of some other. Actually, the advancement of science can only liberate, stimulate, and advance also the arts.

--George Lundberg (1947:14-15)
 
 

As a field for cultivation, literature starts with two advantages: it deals with common experience, couched usually in common language, and it has a long tradition of being looked at and talked about. The corollary is that when the man of science finds it difficult or impossible to read Dickens, the deficiency is in him or in his training; but when the humanist finds it difficult or impossible to talk about science, the deficiency is in him, in his training, and also in the present state of the subject.

--Jacques Barzun (1964:26)
 
 

When scientists deal with each other, they should never think about tactics. Or, at worst, they should think about tactics only in the way a happily married couple does.

--Anonymous
 
 

Pain is a perception, not a reality.

--Shirley MacLaine, in Time magazine,

Pearl Harbor Day, 1987
 

Introduction
 

Lundberg's lively little volume arrived at the bookstalls nearly half a century ago. It was standard reading for us undergraduates of the early postwar years, and given what we knew about famine, disease, war, depression, genocide, and political repression, the book made big promises. I don't know whether the social sciences have delivered on these promises, but I do know that we have established a credible basis for continuing the effort, more or less unimpeded and with an appropriate Lundbergian optimism. I have many colleagues, however, who do not share this assessment, and they have launched an attack--sometimes subtly and sometimes not so subtly--against the established positivist premises of the social sciences.¹ In my opinion these colleagues are basically wrong, perhaps disastrously wrong, and if they succeed in destroying the social sciences they will find themselves enjoying the spoils of a Carthaginian victory.

Perhaps, however, the forlorn victors will be able to deny the reality of their disappointment: The world according to MacLaine is a free-lance filibuster constructed on a simplistic denial of Barrington Moore's (1973) universal forms of human misery. Or, perhaps I'm simply wrong. I am told that I exaggerate, and that the anti-science critique is not all that deadly. But on the other hand, it is possible that the victors will eventually stumble upon a fundamental truth: Sub-cranial pain falls mainly on MacLaine,² while the realities of pain fall upon the rest of us like inexorable Malthusian checks. In any case, and with due respect for Jacques Barzun, I would be satisfied if the anti-science philosophers were to learn the language of science in the same way that most of us learn the language of Dickens: We use it--with at least a modicum of skill.
 

(1) The dialectics of disenchantment
 
 

... during the interwar years, [there was] a widespread questioning of the founding program of the American historical profession: the scientific and detached search for impartial, objective historical truth. ... the interwar criticism of the objectivist posture was a moment in a philosophical debate that went back to Aristotle and Protagoras.

--Peter Novick (1988:250)
 

There is, then, nothing new under the sun. However:

Karen Winkler's early aperçu on antipositivist opinion (26 June 1985) should not surprise anybody when it claims (with George Lundberg) that the social sciences have ties to the humanities. The natural sciences have similar ties, and these derive largely from the fact that all the sciences have their origins in the humane disciplines. This historical reality negates the deplorable tendency among Winkler's interviewees to suppose an irreconcilable conflict between the thought patterns of the humanities and those of the sciences. As Arthur Koestler (1964) argued in The Act of Creation many years ago, the thoughtways of the sage, the artist, and the humorist, although not identical, have large areas of overlap. The disastrous confrontational atmosphere created by Winkler and her interviewees is degraded further by the absurd claims that "the discipline of economics is impervious" to non-scientific thought and that "the mainstream of sociology still almost unconsciously accepts the natural-science model." If we tend to do things unconsciously we will surely find it difficult to "interpret" what we do--a point to which I shall return. For now, suffice it to say that if there is an "arbitrary boundary" between the social sciences and the humanities, then surely Winkler's brand of journalism--and the anti-science school is big on journalism--helps to produce and perpetuate it. But intellectual boundaries, I believe, have no objective existence; they are akin to the boundaries that create "periodization" in history. In fact, the "scientific" and "artistic" hemispheres of the human brain are joined together objectively by a structure called the corpus callosum, the ideational equivalent of which we desperately need.

"Economic man," that quintessential human machine created centuries ago by the earliest economists, has been suspect for as long as he has existed, but Winkler and her interviewees decided that it might be fun to trot him out once again. Nobody who has studied sex or violence or non-answers to Donald McCloskey's question "If you're so smart, why aren't you rich?" has failed to notice that the "means-end, cost-benefit model is far from covering all aspects of human activity and experience." Social scientists who study capital punishment, or who experiment with ways of punishing men who beat women, have known for many decades that deterrence--a simple principle based on the cost-benefit model--cannot be relied upon as a means of social control. And when students of teenage fertility discover adolescents who believe that the withdrawal method of birth control cannot fail, or that they are too young to get pregnant, or that they can only get pregnant just before or just after menstruation, or that they can avoid pregnancy by standing during sexual intercourse and perhaps (both?) facing north, we are in the presence not of rationality but of poetry. Ironically, Economic Man has been converting himself lately into a handy straw man, despite the fact that very few contemporary social scientists have ever tried to prop him up. Like Joan Rivers, we only prop up a dead man when we're desperate.

In American intellectual history it is not unprecedented for economists to make large departures from the means-ends, cost-benefit model. The famous "Phillips school," for instance, believed that the staunchest supporters of slavery in the United States continued to favor the institution despite its irrationality from a cost-effectiveness standpoint. To Phillips and his followers slavery persisted largely because its advocates believed it to be a morally superior form of socio-ecological organization (Fogel and Engerman, 1974:II,176). The Phillips school no longer prevails, but one does wish that its willingness to treat the means-ends model as a mere hypothesis, rather than as a normative goal toward which society ought to strive, were impressed on several unreflective economists currently in the employ of the American embassy in Mexico City, who seem to believe that it is divinely preordained that the central valley of Mexico be transformed into a replica of the San Joaquin valley of California. Don't bet on it.

Remember, too, that one woman's rationality is another woman's madness, even when the women in question happen to be social scientists. Discussing the San Joaquin model as applied to Mexico, De Janvry and Vandeman (1987:107) conclude that
 
 

For Third World countries under the imperative to rapidly increase food production and reduce rural poverty and slow down urban migration, the U.S. model is of little relevance. Nevertheless, increasing food production through large-scale farming, international transfers of technology, and a dominant agribusiness is clearly a tempting way to satisfy such pressing needs. ... The Third World increasingly is adopting this approach. Yet because they lack the labor absorption capacity and off-farm income sources that characterized the U.S. development, these countries are finding the social costs of the U.S. model simply unbearable. ... If rural development policy is dominated, as it is in the United States, by large landholders' lobbies, its purpose will be agricultural development couched in the rhetoric of rural development.

De Janvry and Vandeman (1987:98-104) also point out that agribusinesses do not emerge solely through the operation of market forces. They are highly subsidized. Beyond that, there is a possibility that the "unbearable" costs of agricultural technology may fall more heavily on women than on men (Harriss and Watson, 1987:100-101); among native Americans of the far north, the unbearable costs of the highly efficient snowmobile impose themselves without regard to gender (Georges, 1993).

In any case, McCloskey is wrong. Social scientists--excepting those assigned to the U.S. embassy in Mexico City--are clearly capable of evaluating the cost-benefit model scientifically. It is ironic that a book devoted largely to a critique of positivism, empirical investigation, quantitative methods, the rationality model, and their alleged dominance in the social sciences (Sica, 1988:20, 40, 78-9, 94-5, 118, 136-41, 259), ends by concluding that both Weber and Pareto gave large play to irrational elements in human behavior. It is untenable to argue, as Sica does (1988:262), that "...human rationality is a relatively rare event, ..." and it is equally untenable to argue that social scientists have had little aptitude for the irrational.

And even though McCloskey may be justified in accusing many economists of relying too heavily on formal tests of statistical significance (1985: Chapter 9), one must point out that a major controversy over statistical significance occurred in sociology a generation ago (Morrison and Henkel, 1970) and was more or less resolved in favor of the notion that statistical significance is merely one means to the attainment of "substantive" significance (Selvin, 1957; Walster and Cleary, 1970). One of my students (Flippo, 1987) recently completed a master's thesis in which she was interested primarily in whether high educational attainment tends to delay sexual initiation for a longer time among nonwhite women than among white women. Her idea was that ambitious minority women develop a higher aptitude for deferred gratification than do ambitious majority women. I don't recall whether the observed difference in sexual initiation--age at first intercourse--between the two groups was statistically significant, but I do know that it was not substantively significant.

What bothers me about McCloskey is that he does not acknowledge these processes of intellectual evolution, and he therefore does not see the adaptive character of social science knowledge.

Historians, whether "scientific" or not, always have had strong ties to the humanities, and many universities still cannot decide whether to classify the history department within the humanities or the social sciences. It is not surprising that traditional historians often express their misgivings about cliometrics and other exotic forms of scientific history. But cliometrics is here to stay, probably because historians perforce must deal more and more with historical epochs in which human beings had long since established themselves as a "data-generating animal." The media have arrived, and we cannot ignore the message. Historians should welcome the opportunity to triangulate among speculative history, text analysis, the intensive case-study method, and scientific, quantitative, data-oriented history. Anybody who believes, with Randall Collins (1984:331), that historians are better off when they do not permit "inroads from the statistical side," should take a close look at Fogel and Engerman (1974).

Winkler quotes Alan Sica to the effect that various historical circumstances have contributed to the latest outburst of antipositivism. Sica's speculation about the impact of NSF funding policies would have far greater validity, however, if it were tested by cliometricians with a talent for deriving and assessing its empirical consequences. Sica implies that Congress must have expressed great faith in the scientific value of the social sciences back in the golden days of the 'fifties and 'sixties, that social scientists who had the benefit of Congressional largesse must have shared a similar perspective, and that the recent reduction of funding has somehow undermined the scientific faith among these erstwhile advocates and exemplars of a pure social science. I suspect, on the contrary, that the attitude of Congressmen and of "scientific" social scientists has changed very little, and that the challenge to positivism has occurred primarily because the antipositivists in academe have discovered that less money for their colleagues means less power.³ The Reagan/Bush administrations did not create a revolution in our minds; they created a revolution in our budgets. On the other hand, perhaps Sica is right. Perhaps diminishing public support, involving explicit disparagement of sociology by the president of the United States, has created a crisis of confidence in which the discipline finds itself "in the doldrums," a prospect taken seriously by a recent president of the American Sociological Association--this is not a healthy form of presidential accord. If we actually have allowed Reaganite anti-intellectualism to undermine our confidence in ourselves as scientists, then I say that our behavior is approximately as rational as that of the adolescent lovers mentioned earlier.
 

(1.1) Predictability, vulcanology, seismology, and (especially) meteorology
 
 

The cases presented here ... give evidence of the fact that social science research can be effective. ... an extreme view of the ineffectiveness of social science research, it would now seem, is untenable.

--Bernard Barber (1987:154)
 

The antipositivist's claim that social scientists cannot make predictions is nonsense, and anybody who plans to abandon scientific thoughtways because of this alleged defect is conducting a most inauspicious withdrawal--I cannot seem to escape adolescent techniques. Social scientists have become so good at predicting electoral results that we allegedly keep people away from polling places. Using an extraordinarily powerful theory known as the stable population model, we can predict (among other things) the future burden of social security; if the Social Security Administration did not have the benefit of our predictions (or "projections," as demographers prefer to say) the results could well be disastrous. We can predict, again within a small margin of error, the number of limited wars that will be fought by, say, the end of the current millennium. Social scientists predicted, correctly, that the peace settlements following World War I would create instability. During World War II we predicted that racial desegregation would have a generally benign impact on American society. Perhaps our greatest success in the realm of social engineering was the desegregation of the American military, a process strongly urged by social scientists despite the widespread belief, both within and without the military, that it would weaken America's warmaking capability. A few years ago a number of us predicted the recent resurgence of capital punishment in the United States, and we also had a few things to say about the prospects of reducing violence through this means. I do not claim that we can predict everything with complete accuracy, but I do claim that our ability to make predictions is at least as impressive as that of the vulcanologists working a few years ago in the vicinity of Mount St. Helen's, and more recently in Colombia (Marlowe, 1993). Decision theory, learning theory, the stable population model: These scientific devices, and their ability to anticipate the future, are comparable to much of what is found in the mighty arsenals of the natural sciences.

A few social science success stories have been so impressive and are now so completely taken for granted that they seem to be imbued in the very nature of things and we no longer receive due credit for them. Take, for instance, the matter of human mortality. Available data on the death rates of, say, 150 or more years ago suggest that these rates were highly unstable and unpredictable: They could readily vary from 30 per thousand annually to perhaps 200 per thousand annually during famines, plagues, and wars, and these variations were often not anticipated. Nowadays we can say with nearly complete certainty that the death rate of the United States throughout the 'nineties will be very close to 9.5 or 10 per thousand annually, as it has been for several decades. One of the reasons why I cherish this particular example is that it is arguable that, during the nineteenth century, large reductions in mortality as well as the stabilization of death rates were brought about not primarily by advances in the medical sciences but rather by improvements in transportation, communication, environmental sanitation, and public education (McKeown, 1988); that is, by inventions that took place in the realm of human ecology, broadly defined, and that were dependent on social diffusion. In brief, these changes were the work of people who could well be called applied social scientists.

For those contemporaries who have the courage to practice applied sociology (Whyte, 1986), the ability to predict and explain social phenomena becomes crucial: Enough mistakes, and one's job disappears.

As Seymour M. Lipset correctly points out (1979:5-7), we have had some notorious failures of prediction--for instance, our failure to anticipate the "stagflation" of the 'seventies, or our failure to predict changes in human fertility. In 1992 he added to his list the highly debatable allegation (Lenski, Lenski, and Nolan, 1991:261) that social scientists failed to anticipate the disintegration of various communist regimes. But for every demonstrable failure there are several impressive successes, and we can even provide instances in which prototypical social scientists such as the authors of The Federalist Papers have designed and correctly predicted the future behavior of elaborate social systems (Bellah et al., 1985:254-55; Diggins, 1990). If social scientists were more concerned about winning for themselves the prestige of the natural sciences (or, better yet, the prestige of the Founding Fathers), instead of slinking off to the English department for this afternoon's pomo brownbag, they would probably have many more opportunities, eventually, for designing and testing social systems or large elements of them. A few of our colleagues in West Germany (Rehberg, 1985; Stehr, 1986; Tenbruck, 1984) seem to believe that our past successes have produced the sort of anti-sociology backlash that gives some of us a sense of being in the doldrums, but I predict that we will rise above this petty affliction with minimal effort.⁴ I believe that what Tenbruck is saying about the social sciences is largely the sort of thing that economists might say about TV sets (Cook and Campbell, 1979:313): If manufacturers sold lots of TV sets last year, they will not do so well this year because they have satisfied the high levels of demand that existed last year.

This year's low returns, in other words, may be an indication of last year's successes--or the last century's successes, to return to the mortality example cited above. And even regarding prediction, social scientists had as many successes last year as did vulcanologists, seismologists, or meteorologists--few of whom find themselves in the doldrums.
 

(1.2) Molecular mysteries and habits of the quark
 

Another inappropriate but highly predictable twist taken by the antipositivist argument (Alexander, 1988:80) is that a science of social behavior is impossible because men and women behave more subtly than do molecules. Molecules, says one of Winkler's interviewees, do not lie. While I know little about the veracity of assorted chemical compounds (except for tea leaves and sheep entrails), I do know that anybody who studies molecules is impressed by the fact that, as one scientist phrased it, nature is more subtle by many times than the puny ability of the human mind to comprehend it. And this generalization applies to physics just as clearly as it applies to the biological and social sciences, despite Dawkins' untenable claim (1986:2) that physics studies "simple things." Any good physicist--say, a specialist in quantum dynamics--knows that her field is infinitely complex (von Baeyer, 1992), and if you try to tell such a colleague that the infinite complexities of your life are greater than the infinite complexities of hers, she is likely to conclude that you are mentally deranged.

As for biology, there are viruses that seem to have read every textbook ever published on molecular biology and a small set of manuscripts not yet published, submitted, or conceived. When Dawkins writes about his own field, which is biology, he speaks compellingly: Biology, especially at the level of molecular genetics, evolutionary theory, and their interrelationships, is a science of complexity. One cannot find anything in Habits of the Heart that makes its subject matter more complex than that of The Blind Watchmaker.⁵

To those who point out that there is nothing in the universe more complex than the human heart and the human mind, one can only reply that the complexity of human heads, hearts, hands, and health is to a large but indeterminate degree a matter of molecular biology.

In short, the world is very impressive; human thought, as I shall suggest in Chapter 2, is not much more impressive than computer thought. Nature presents infinite complexity in all its dimensions. The subtlety of nature tends to flamboozle all scientists, not just social scientists, and we do not find here any unique disqualification of the social sciences--even the great Michel Foucault, as we shall see, was mistaken on this question.

Winkler recounts Renato Rosaldo's experiences among the Ilongots, apparently trying to show that Rosaldo's emotional make-up contributes to the quality of his scholarly work. Her presentation in this case undercuts the anti-science stance, because one cannot see the alleged relationship between Rosaldo's sad experience of losing his wife and his excellent hypothesis that exchange theory may help us to explain (and "postdict," if not predict) head-hunting among the Ilongots. This is a typical social science hypothesis and it is testable regardless of whether one is capable, in Rosaldo's words, of getting "close to the emotional force involved." The emotions of the subjects are likely to be irrelevant to the explanatory tasks of the scientist. As Marvin Harris argues (1979:333-36), it does not matter what was going through the minds of the victims of Aztec sacrifice as these unfortunates were being dragged up the temple steps to the slaughter. (Harris, to be sure, overstates the case: Would that we could have interviewed some of the potential victims!) I believe that we could readily assemble a panel of anthropologists who could outline a procedure that would enable us to decide whether Rosaldo's hypothesis fits reality. In selecting this panel, we would be wise not to pay any special attention to the emotional make-up of its members. Selection by expertise rather than emotion or other irrelevancies is the hallmark of science and one of the major reasons why it progresses--at least in the sense that it can make decisions about hypotheses. And this, of course, is a good definition of scientific progress.

Rosaldo reminds me of George Bush: He has strong feelings, but once he has called this to our attention a few times it tends to lose its force, and its relevance.

It is puzzling to me that Rosaldo somehow has come to be associated with the antipositivist critique, because his writings on the topic insist that social scientists continue pursuing scientific knowledge according to the standard positivist maxims. He tells us (1986), for instance, that Le Roy Ladurie's studies of fourteenth-century French village life may have been impaired by the fact that the major source of data was the "inquisition register" of Bishop Fournier; many of the respondents may have slanted their answers out of fear for their lives and/or salvation. Fournier, it appears, was no Torquemada--but he wasn't Charles Booth or George Gallup either. Rosaldo makes similar claims about the classic studies of the Nuer by Evans-Pritchard, who worked under the auspices not of the heavenly powers but rather the earthly powers of the majestic Government of the Anglo-Egyptian Sudan (1986:88).

The cardinal rule of positivism, of course, is that we must strive for objectivity. Rosaldo is merely insisting that we abide by this maxim.

In any case, it may well be that an anthropologist whose personal experiences and emotions are comparable to those of Rosaldo is more likely to arrive at a particular hypothesis, and it is for this reason that we should not act as if the emotionality of science were somehow incompatible with insight and objectivity. The scientist who discovered a volcano on one of Jupiter's moons a few years ago speaks as if her emotional constitution--at least her persistence in poring over data--had a lot to do with the discovery, and I believe her argument is at least as compelling as that of Rosaldo. In Watson's The Double Helix (1968), one discerns the strong emotions involved in the competitive search for DNA. Guillemin and Schally, who received the Nobel prize for their work on the ways in which the brain controls the endocrine system, were swept along (or perhaps impeded) by intense emotions (Meites, 1977). Again, Koestler (1964:101-267) shows that emotional experiences, dreams, accidents, coincidences, serendipity, and the like have always been at the center of scientific procedure and progress, and I believe that social scientists know this as well as anybody. In Hammond's view (1964:3), the best sociologists have a "feel" for the tactics and strategy of science, and Becker claims (1987:25-26) that neither natural scientists nor social scientists work with total impersonality and objectivity. Again, Sica's (1988) insistence that social scientists place too much emphasis on human rationality is undercut by these sorts of considerations.

On the other hand, when we evaluate the contribution of emotion to the sciences, we must remember that powerful emotions underlie the astronomy of Ptolemy, the biology of Lysenko, and the psychogenetics of Arthur Jensen. Science is the process of getting experts to agree about things regardless of their strongest emotions. The emotions that lead us to pursue the beauties of Jupiter's moons do not dictate a strong emphasis on emic explanation, and they do not dictate an abandonment of the central thoughtways of science. Incidentally, it is contradictory to argue, as do some of Winkler's interviewees, that a social science is impossible because the scientist is part of his subject and that the social sciences suffer because so many of us remain emotionally detached from the subject. It is interesting that Barzun (1964:87) considers detachment to be the most civilized aspect of science, an extraordinary accomplishment for the "restless rampaging beast" that spends most of its time trying to whip nature into shape.

Finally, Winkler's article wrongly asserts an ineluctable tension between science and interpretation, implying that scholars who opt for interpretation must necessarily declare war on science. On the contrary, it is entirely possible to study interpretation, the meanings of symbols and of human experience generally, from a scientific standpoint. There is, for instance, a highly developed positivist psychology of human perception and cognition. When scholars begin speaking about interpretive styles of investigation, however, we often encounter a distressing tendency that is dangerously exacerbated by the Winklerian world-view: The problem is that interpretations--either those of the observer or the observed--are taken as the ultimate reality. But social scientists who contribute to the debunking function, who (with Collins, 1984) take the null hypothesis seriously, have reason to believe that most interpretations, most of the time, are wrong. It is arguable, in fact, that this assertion is the central premise of all the sciences, applying therefore even to the finest etic interpretations. And if we cannot muck through with the best, what, pray tell, do we make of the rest?

Mortimer Adler (1985: Chapter 1) maintains that the separation of consciousness from its objects, the separation of "individualist" and "holistic" perspectives (James, 1984), is a serious philosophical error, and Marvin Harris has argued in several works (1979; Harris and Ross, 1987) that if we rely heavily on our own untested interpretations of social behavior or (worse yet) on the interpretations provided by "knowledgeable" interviewees, we will inevitably go astray. Harris (1979:246-53) shows that if we accept emic explanations of the significance of sacred cows, we are assured of learning next to nothing about the social and ecological dynamics of this custom. Similarly, if we rely on a cross-section of Gallup Poll interviewees for insight into the structure and functioning of American society, we will learn almost nothing about the social science principles that informed the authors of The Federalist Papers, about whether or not America has a ruling class, about the alleged indispensability of "waste" and "waste-makers" to the American economy, or about the extraordinary features of Southern slavery discerned by Fogel and Engerman (1974). Just ask your relatives, friends, and neighbors about the origins and functions of the incest taboo, comparing their replies against what anthropologists have theorized about this topic, and you will see my point. Or, in a more contemporary example, find a dean of faculty, ask him whether sex discrimination in faculty salaries exists at your institution, and compare his reply against the sort of reality discernable by an appropriate regression equation. I do not accuse the dean of malfeasance: As I implied above, if social scientists can do things without conscious awareness or intent, so can everybody else. The dean may simply not know what is going on.

It may or may not be true that "if everybody is somebody, nobody is anybody." I am strongly convinced, however, that if every interpretation is taken to be real, then nothing is real.⁶ And if everyman's interpretation is capable of informing me, and I do my best to assimilate everyman's interpretation, then my own interpretation must be unassailable and I have achieved the superiority of the town gossip. Immediately, then, we are at the fringes of the politics of solipsism, and the consensus-creating practices of science become a mere imposition, an intrusion into the do-it-yourself reality of each of us. Obnoxious little books like Harry Browne's How I Found Freedom in an Unfree World (1973) become our scriptures, and our great commandment, standing alone as an icon of cultural illiteracy, is "Know thyself--and nothing more." Tune in on Browne's discussion of the merits of tax evasion, and compare his view of taxpayers with that suggested below in Section (6).

Some years ago I believed, with Theodore Roszak (1969), that anti-science is an inexorable feature of the "counterculture." Nowadays it occurs to me that anti-science may be the philosophical foundation of the "me generation," and I do my best to suppress the fear that the me generation is merely the latest manifestation of the counterculture.

Science, on the contrary, is integrative, regarding intellectual property as communal in its creation and ownership. It insists upon cooperation. It insists that similarities among things--and among people--are more important than differences. It insists that ego-tripping give way to humility and to an attitude of reverence toward nature. It insists upon its own intermingling with the many non-scientific thoughtways that strengthen it and are strengthened by it. It is ironical that if we aging scientists ever try to bring back the idealism that lived so briefly in our students of the 'sixties, we shall probably have to do so in the name of science. The most promising young radicals I ever knew referred to themselves as the HEADS--they were Human Ecology And Demography Students, they didn't smoke all that much, and they were the wave of the future. And although I do not believe that mathematical competency automatically makes one scientific, I can only hope that Barzun (1964:30) is correct in the claim that "the statistical outlook prevails and establishes an equality that dwarfs the individual." Back in the 'sixties I tried to impress this argument on the Students for a Democratic Society (SDS), but at the time it appeared that they were too individualistic to accept it, not yet having discovered the likes of Althusser and Poulantzas (James, 1984:79-145). As we shall see in the next chapter, however, an intellectual discipline such as statistics is probably not loaded with the ideological biases that are often fashionably claimed nowadays for such disciplines.
 

(1.3) Rosaldo revisited: How would the catcher in the rye have felt?
 
 

I believe that there are fruitful possibilities ... for sociological theorists to shift their reflexive analytical focus from metatheoretical foundational concerns to practical-moral ones.

--Steven Seidman (1991:142), citing Renato Rosaldo (1989).
 

I'm very happy for Rosaldo.

I'm delighted that it is possible for an American anthropologist to live to full adulthood, to become a prominent scholar, to marry, have children, travel across the world many times--and never encounter the "rage in grief" until the awful day when his wife, a beloved friend and colleague, falls over a precipice to her death. I was born in 1938, and I was filled with grief and rage by the time I was three years old. By the spring of 1942--those sad, unforgettable days--I knew that an incomprehensible catastrophe had occurred at a place with a beautiful and mysterious name, Pearl Harbor; I knew that my grandfather, who early one winter morning had given me the exotic, surrealistic experience of walking across Santa Monica Boulevard for the first time, was dead; I knew that my parents were breaking up, that my mother was losing her father and her husband at the same time. And I had my counterpart of Rosaldo's headhunting: I knew that the world is flat, that bullets travel infinitely in a straight line parallel to the flat earth, that machine guns never run out of bullets, and that therefore all I had to do, to be safe, was to acquire such a weapon, press the trigger, and spin round and round. I also knew that I had acquired a "twenty-five dollar United States war bomb" from a very motherly woman named Kate Smith, and that if my machine gun did not work--no problem: My very small pilot would throw open the canopy of his very small airplane, raise my very small twenty-five dollar United States war bomb, and drop it directly on General Tojo.

Which brings me to my current theoretical problem, the AIDS epidemic. I want to know how to model this epidemic: Is some sort of mathematical function appropriate (Brookmeyer, 1991), should I try to develop a simulation (Whicker and Sigelman, 1991: Chapter 6), or should I try to find more data, say, on sexual conduct comparable to the data provided by the National Survey of Family Growth (NSFG) or the General Social Survey (GSS)? Should I use all these approaches at once, and triangulate among them? Am I missing the legs of a larger triangle? In any case, my own endless experience of the rage in grief and (why not?) the grief in rage--and here I must allude to Holden Caulfield, the dialectical opposite of Toe Joe the Ogre--will further my understanding of the AIDS epidemic to precisely the same degree that Rosaldo's "use of personal experience" has furthered my comprehension of headhunting.

One final thought: In the age of AIDS, those who believe in giving full play to their passions should contact the contemporary counterpart of Kate Smith, and get some investments in place.
 

(2) The trouble with feminist theory
 

When Susan Bordo (1987:261) argues that Western intellectual history over the last several centuries has produced a "Cartesian masculinization of thought" and that this masculinization results from "the seventeenth-century flight from the feminine," she implies that there is something macho about the sciences, including the social sciences; presumably, there is something inherently feminine about an identifiable set of preceding centuries. Other advocates and exemplars of "feminist theory" develop the machismo argument more explicitly. Marlene Mackie (1988:1-2), for instance, claims that "... the linchpin of sexism in sociology is its methodology"--referring to those aspects of methodology that purport to be scientific. She laments "... the use of masculine research style ..." and refers (1988:4) to a "machismo element" in sociological research that leads to "exclusion of women through choice of research topics." She tells us (1988:11) that a content analysis of the research literature has revealed that male-authored articles are more likely than female-authored articles to be "masculine oriented"--specifically, 90 per cent versus 71 per cent. (Interestingly, this comparison suggests that fewer than a third of the female authors in sociology write something other than masculine-oriented work; it also leads us to question Mackie's "exclusion of women" remark.) Mackie (1988:12) then arrives at a recurrent claim of feminist theory: That although we often view gender as an important characteristic of individuals, we fail to recognize its significance as a "principle of organization"--a claim to which I shall return. Finally, Margaret Andersen (1987:247) tells us that "... critiques of the scientific method should be a primary concern in feminist revisions of social science courses," and Susan Weisskopf (1978:277) informs us that quantitatively oriented psychologists and sociologists have not succeeded in controlling research biases; instead, they have managed to "... lose touch with the individuals behind the statistics"--another theme that recurs often in works of feminist theory and in the antipositivist critique generally.

In trying to assess whatever support has been adduced for these arguments, I usually find many instances in which scientific method has strengthened the feminist perspective, rather than subverting it. Longino and Doell (1987:165,175), for instance, while focussing on "masculine bias" in the social sciences, point out that
 
 

What the study of contemporary hunting and gathering societies should teach us is that, short of stepping into a time machine, any speculation regarding the behavior and social organization of early humans remains just that. This leaves framework choice subject to influences such as the speculator's preconceived and culturally determined ideas of what human beings are. The distance between evidence and hypothesis ... remains an invitation to further theorizing or, as some would have it, storytelling.

Longino and Doell believe that masculine bias cavalierly fills this gap by telling the story of man the hunter and woman the gatherer. But the central tenet of scientific method is that one does not fill gaps with unsubstantiated guesses; rather, one follows the rules of scientific procedure in order to identify gaps in knowledge and arrive at an appropriate means of closing them. There is no evidence that the "masculine" style of scientific procedure has been remiss in these responsibilities (Lenski, Lenski, and Nolan, 1991:105). In fact, the sorts of scientific attitudes that feminist theory regards as male-oriented are probably associated with more than the usual degree of skepticism about scientific claims, including claims about the earliest sex roles. The many anthropologists and sociologists who laid to rest the various unilinear or parallelistic theories of social evolution were convinced that the earliest human origins, as Sumner once said, are lost in obscurity, and that dogmatic assertions about them therefore cannot be entertained. There is little enthusiasm, today, for Tylor's theory of the incest taboo, for his theory of racial differences, for Freud's theory of the parricidal crisis, or for the Morgan-Engels theory of the origins of the family and private property. Although these scholars told magnificent stories, they could not provide the sort of evidence that would lead us to accept their stories as scientific knowledge. And it is noteworthy, as Josephine Donovan (1985:73-76) points out, that Morgan and Engels did not consistently place the male of the species in the dominant position, in their stories.

On the question of the early evolution of sex roles, the best studies that I've encountered recently are summarized in Harris and Ross (1987). Harris is strongly committed to "the struggle for a science of culture"; yet, despite this male-oriented proclivity, the Harris-Ross volume provides great insight into the dynamics of excess female mortality (emphasizing female infanticide), the harsh treatment of women generally, the relationship between sex exploitation and class exploitation, and many other such topics. These studies also have the virtue of directly challenging Malthusian theory as it applies historically, and demographic transition theory as it applies to modern populations.

Another illustration of the ways in which science has contributed to feminist theory occurs in Donovan's book. When I came to the page where we learn that "... the witch craze developed in Europe and in New England at the same time as the Newtonian world view was gaining ascendency" (1985:29), I was tempted to write a paraphrase in which I would point out that witch crazes were disappearing at the same time as the Einsteinian world view was gaining ascendency. But before I could do so, I came across the following remarks in a later chapter of the same book (Donovan, 1985:180):
 
 

The Newtonian or scientific world view ... is rooted in the ... masculine psychology ... However, developments in twentieth-century science, such as Heisenberg's Principle ... and Einstein's ... Theories of Relativity ... have challenged the validity of the Newtonian paradigm. The new vision of the universe that is emerging is no longer of an Other that operates in predictable, mechanical fashion, but of a contextual network in which every discrete entity is defined relative to its environment and subject to the positional relativity of the observer.

This sort of concession undermines the anti-science arguments presented by Bordo, Mackie, Andersen, Weisskopf, and by Donovan in the earlier parts of her book. But even as a concession, this passage (and its larger context) leaves a lot to be desired: It does not explain what is masculine about Newton, why relativity is non-masculine, or whether it would be appropriate for us to credit twentieth-century physics with having given major impetus to the feminist movements of this century.

In a few recent works of feminist analysis, one notices an occasional indulgence in what Longino and Doell might call unsubstantiated storytelling. Andersen (1988:17-19), for instance, laments the fact that much of the recent literature on the status-attainment process in the United States involves surveys of men only, but when she castigates Christopher Jencks et al. (1979) on the basis that "... his national sample includes only men ...," we lose sight of the fact that Jencks and his eleven collaborators (including four women) used five national surveys and six special-purpose samples, and that some of these datasets did include information about women. It turns out, granted, that information on women was so limited that, as Jencks et al. say (1979:4-6), "... we reluctantly decided to restrict all our analyses to males ... Nonetheless, this limitation is both serious and regrettable, since sex is one of the most important single factors affecting earnings." This research team, then, had strong misgivings about the neglect of women in status-attainment research, and it is clear that they believed that the surveys available to them missed a lot.

I mention this example primarily because it illustrates the important distinction between theories and "orientations" (Merton, 1957:87-89): If it is true that the status-attainment literature has given little attention to women, this is a criticism of the orientations of researchers in this area, i.e., their ways of selecting problems for analysis, and not of their theoretical practices. Theories and orientations are largely independent of one another. For instance, one could make excellent contributions to conflict theory by developing comparative analyses of international relations, labor-management relations, and race relations, while rarely or never dealing with questions of gender or male-female interaction. And vice versa.

Nor can I agree with Andersen (1988:19) that matters have not improved, that "... in sociological work, gender is seldom considered to be a factor that influences social behavior." In order to substantiate this large thesis she would have to show, among many other things, that approximately 700 gender-related articles and books based on the GSS and published by 1986 (Smith and Fujimoto, 1986: mnemonic index, 50-51), and the many studies published since that time, had little relevance to social behavior. She would have to demonstrate the irrelevance of the U. S. Census, the Current Population Survey, the National Longitudinal Surveys of men, women, and youths, the NSFG, and many other such enterprises in which large quantities of gender-related data have been produced. And then she would have to extend the analysis from survey-based literature to the many other types of literature produced by social scientists. If she were to take these steps, I believe that she would have to abandon her argument, or at least a large part of it.

From time to time, Andersen seems to suggest that the essential problem is that we do not pay enough attention to the sex composition (the sex ratio) of social groups or organizations. But this situation arises from the fact that in order to calculate a sex ratio, one must be able to identify significant ways of grouping or aggregating the data. It is hard to do this with the General Social Survey. The GSS involves a sample of independent households, and there are relatively few ways of grouping data so as to make the sex ratio a meaningful measure. It is entirely possible, however, to use the sex variable from the American Council on Education's massive surveys of college faculty members as a basis for classifying institutions, departments, academic fields, etc., on their sex composition. Many scholars--even male-oriented scholars--have shown no reluctance to take advantage of this sort of opportunity. If it is important, as Mackie suggests in the citation above, that sex be conceptualized as a "principle of organization," this surely must refer to such practices as the use of the sex ratio as opposed to the simple sex classification of individuals. In the Harris-Ross work cited earlier, the sex ratio of children is the most important means of assessing the degree to which infanticide selects against females. If gender as a principle of organization is intended to have additional meanings, they should be stated.

Although feminist theory addresses itself primarily to methodological questions, it is interesting that the one style of theorizing that seems to call forth mandatory malediction is functionalism. When Andersen says (1988:20) that "the focus on norms, roles, and stability emphasizes the status quo ..." and that functionalist theory traps us into using these limiting concepts, one must ask, first, how was Mirra Komarovsky limited by her focus on roles and norms? Second, in what sense are researchers who focus, say, on sex ratios guilty of overemphasizing sex roles, given that these roles are not defined for aggregate data? (If a group or social category can have a sex role, this is a neologism that must be defined with care.) Regarding functionalism, Susan M. Okin (1979:95,240-41) makes a comparable error when she suggests that scholars still see women from a sort of Aristotelian functionalist perspective that insists on a place for everything and everything in its place, including women. The error arises from the fact that sociological functionalism is hardly more teleological than Darwinian functionalism--Darwin and Wallace, in fact, borrowed the central postulates of evolutionary theory from the great functionalists Malthus and Smith. Contemporary functionalism may give human intentionality a larger role than that implied by the Darwinian (or social Darwinian) view, but contemporary functionalism is prepared to take Shulamith Firestone very seriously when she suggests (Okin, 1979:295) that technology, for instance, is capable of liberating women from the "barbaric" character of pregnancy, for starters.

In brief, it is not at all clear what Stacey and Thorne (1985:309) have in mind when they argue that "... positivist knowledge serves the interests not only of dominant social classes ..., but also the interests of men ..." Various demons may rampage through our works, but they are not positivist demons. In the grandest traditions of androgynous sociology we positivists must insist upon both the allegedly masculine practices of abstract, etic, quantitative structuralism, and the presumably feminine practices of giving attention to "... individuals behind the statistics ...," emic exploration, intensive case studies, interpretationism, qualitative methods, and so forth. We also insist that the best sociology is done when these diverse perspectives are combined and made to work in a complementary way.
 

(3) Titles and tribulations
 

Throughout this book I illustrate the harmful consequences of abandoning the high standards of scientific inquiry. In recent months I have found that whenever I try to discuss openly my misgivings about feminist attitudes toward science, I encounter further illustrations. Something is very wrong when a highly productive and capable scholar tells me, in a personal letter, that "I had written a response to you but I didn't send it for publication, for two reasons: 1) I didn't think it my place to respond, 2) I thought [your] editor was the real culprit." There have been several letters and a few phone calls from colleagues who implicitly affirm the prescient words of Eri Atlov: "I do strongly agree with what you say, but I shall be scared to death to say so forthrightly." The sort of colleague who sends these messages reminds me of the "undisclosed source" whose words we read so often even (or perhaps especially) in the better newspapers. But those who are afraid to stand by their own ideas already have surrendered their constitutional rights, no matter how provocative their public expressions; like Emmanuel Goldstein of 1984, these quintessential voices must remain in hiding. And Thomas Jefferson would doubtless remind us that if these anonymous advocates of provocative notions cannot be challenged because they will not reveal themselves, then the marketplace of ideas is already suffering restraint of trade.

As I became more heavily involved in raising the questions of the preceding section, it occurred to me that an appropriate title for this book would be Androgynous Sociology: The Power and the Elegance. After all, the basic thesis of the book is that what we must do is to combine the best of the alleged masculine thoughtways with the best of the alleged feminine (or feminist) thoughtways. But reactions to this idea, from men, were almost always along the following lines: "I would be afraid to self-select a volume with that title for fear of what people might think." And from women, although I usually hear that there is something simplistic about the androgyny theme, these deficiencies are never specified.

And then there is a madman who wrote to a colleague of mine. He asked how "Herr Professor Faia" was doing, said that Stacey and Thorne are merely revealing the obvious when they write that positivist knowledge serves primarily the interests of men, affirmed that science does not attract women because it is "quantitative and logically rigorous," argued that presumably feminine (or feminist) thoughtways have no place in science, excoriated a "geek-branch feminist" for something, involving Mormonism, that I did not understand, and then wound it all up by telling us that he "got out" of sociology because it is "too quantitative and pays too little attention to individuals ..." An unforgivably careless reader, who feels a false affinity with an argument that repudiates his own nefarious analyses, virtually at every turn. Herr Professor Faia assumes that, by now, this gentleman-scholar has met the matriculation requirements of Eastern State, a Virginia institution of some repute.

Bologh (1990), Cancian (1990), and several other colleagues insist that I might modify my misgivings about feminist theory (Faia, 1990) if I were to read a book by Harding (1986). I have done so (not for the first time), and I therefore reconsider as follows:
 

i
 

Harding's indictment of science, its history and its contemporary forms, contains a lengthy bill of charges. She provides an early summation (1986:9): first, the epistemologies, metaphysics, ethics, and politics of mainstream science are androcentric and mutually supportive in their androcentrism; second, modern science serves "regressive social tendencies"; third, scientific applications, technologies, ways of creating meaning, and modes of defining problems and designing experiments are sexist, racist, classist, and culturally coercive. "... I do not wish to be understood," says Harding (1986:10), "as recommending that we throw out the baby with the bathwater;" yet, as we get better acquainted with it, the baby appears to be badly in need of an exorcist.

Bologh, Cancian et al. are right: After a careful reading of Harding I have a far better grasp of feminist claims regarding the pervasive character of gender. For instance (Harding, 1986:17,57):
 
 

... as a symbol system, gender difference is the most ancient, most universal, and most powerful origin of many morally valued conceptualizations of everything else in the world around us. Cultures assign a gender to such nonhuman entities as hurricanes and mountains, ships and nations. ... Gender is a fundamental category within which meaning and value are assigned to everything in the world, a way of organizing human social relations ...

And this is only the first installment, the first dimension, of a detailed multidimensional definition that clarifies itself as one continues working through the book.

A second feature of gender traditions, in Harding's analysis (1986:20,123), is that they have led us to view the world with inescapable distortions, such as those arising from a tendency to see reality in binary form, to see things as dichotomies corresponding to the bisexual character of the human species. Scattered throughout the book is an argument on the desirability of breaking up the two-valued logic of machine vs. human, animal vs. human, mind vs. body, reason vs. emotion, objectivity vs. subjectivity, abstract vs. concrete, conceptualizer vs. executor. Harding has a powerful aptitude for the dialectical interlacing of these apparent oppositions and contradictions, as in her claim (1986:71) that there are few differences between research and engineering, or her claim that affirmative action programs are simultaneously reformist and revolutionary (1986:247). She tells us that a major objective of feminist theory is the development of "feminist standpoint epistemologies" (1986:24-9,142-62). We soon realize that these epistemologies, again highly complex and multidimensional, place a strong emphasis on the integration of opposites.⁷ They express hostility toward positivist empiricism (1986:33), and they make a strong commitment to postmodernism with its rejection of the notion of "one true story," an intellectual vice that is alleged to be the essence of science (1986:24-9,194).

One hopes that the feminist standpoint epistemologies, working (at times) within the hoary confines of the American university, will eventually challenge the mindlessness of the EGAD! system--Examinations, Grades, Accumulation (of credits, "contacts," etc.), Degrees.

In the realm of gender symbolism, Harding especially deplores the metaphorical association of woman with nature (1986:114-16) or with the earth; the latter association leads her to the conviction that the Copernican revolution, displacing the female earth with the male sun, was a world-historical degradation of womanhood (1986:85-92). Here, we begin to comprehend her expansive thesis about the dangers of metaphor (1986:113,116,237):
 
 

... feminist historians have focussed on ... rape and torture metaphors in the writings of ... Bacon and others (e.g., Machiavelli) enthusiastic about the new scientific method. ... understanding nature as a woman indifferent to or even welcoming rape was equally fundamental to the interpretations of these new conceptions of nature and inquiry. ... why is it not as illuminating and honest to refer to Newton's laws as 'Newton's rape manual' as it is to call them 'Newton's mechanics'? ... As nature came to seem more like a woman whom it is appropriate to rape and torture ..., did rape and torture come to seem a more natural relation of men to women?

On this argument, more anon.

Harding argues that science selects the wrong problems, defining excessive population growth, for instance (1986:77), as merely a technological issue. With most of the anti-science scholars discussed in this chapter (e.g., Alexander, 1988:80), she makes the predictable claim that the physical sciences are "much less complex" than the social sciences (1986:44), and that the latter, because of their immense complexity, cannot aspire to high scientific standards. "I need hardly even mention the silliness ...," says Harding (1986:46), of assuming that the physical sciences provide a model for social-science explanation. Insofar as she does elaborate on this point, the limitations of the social sciences seem to have something to do with subjects' intentions and learned behaviors; mathematical formulations in the social realm, for instance, are impossible because of these human qualities (1986:52). Finally, within the social sciences there are at least five major sources of androcentrism (1986:85-92,114-16): first, neglect of emotion and irrationality; second, neglect of unofficial, private aspects of life; third--contradicting earlier claims about dichotomous thought--an overemphasis on male-female similarities rather than differences, especially with regard to the meanings of personal experience; fourth, a tendency to ignore sex--"... probably the single most significant variable in history" (1986:89)--as a factor in behavior; fifth, over-reliance on certain methods--e.g., quantification or the practice of having men study women--that may fail to elicit certain kinds of information.
 

ii
 

Harding's book is a most edifying exercise in the thoughtways of reform and revolution. I continue to have major problems, however, with her general hostility toward the social sciences. She believes that the social sciences fail inevitably because their topics are too complex. But her own argument fails primarily because the tasks of defining and measuring the relative complexity of the social and natural sciences are themselves highly complex, and we cannot carry them out. It is ironic that scholars who often do not trust measurement are so cavalier in making a claim that would require an intricate measurement process. Yet, advocates and exemplars of anti-science often make a tropistic rush toward questions that we cannot answer, toward concepts that we can hardly comprehend. In Harding's book, these errors abound.

At the moment, however, I should like to consider one of Harding's claims regarding the role of gender symbolism--the rape-of-nature theme--as a historical factor. Harding quotes from a speech by Richard Feynman on the occasion of his receiving the Nobel prize, in which he says that falling in love with the ideas of physics is "like falling in love with a woman," a woman for whom one's love is not diminished by the discovery that she is less than perfect, a woman whose aging, unfortunately, seems to have created insurmountable problems for Feynman although the implied cardiovascular limitations were doubtless his own. In any case, Harding regards Feynman's speech as a most nefarious act.

Harding's decision to associate Feynman's philosophy with the rape and torture themes of the several preceding pages reflects a failure to apply her own lessons: Feynman's ardor may involve the imagery of unveiling, unclothing, and penetration (1986:117), but he defines an expression of love that, although flawed, is the direct dialectical opposite of rape and torture. If it is plausible to say that Feynman advocates rape and torture, then it is equally plausible to say that devotees of fellation and cunnilingus advocate cannibalism. But a loving practice of fellation and a loving practice of cunnilingus, again, are the precise dialectical opposite of cannibalism.⁸ When these forms of lovemaking are conflated with cannibalism we are in the presence, at best, of sado-masochism; at worst, the madness of General Jack D. Ripper of Doctor Strangelove.

Feynman, in short, is the wrong illustration in a situation where right illustrations may be in short supply. Harding does not provide any convincing support for the rape-torture thesis. It is nothing more than a provocative hypothesis, for which relevant evidence is not yet forthcoming.

Working through the book, I developed further misgivings about many of Harding's arguments. Among other faults, these arguments tend to have one thing in common: They are short on evidence.

Item: Harding makes incredible over-generalizations. For instance (1986:21):
 
 

... science is used in the service of sexist, racist, homophobic, and classist social projects. Oppressive reproductive policies; white men's management of all women's domestic labor; the stigmatization of ... homosexuals; gender discrimination in workplaces--all these have been justified on the basis of sexist research and maintained through technologies, developed out of this research ...

Again it is the lack of evidence that is unsettling, along with the suspicion that if evidence were gathered it would not countervail against the more positive accomplishments of, say, the social sciences. How, for instance, can one fail to give major credit to the social sciences for the suppression of racism in the United States and elsewhere? We've come a long way since the early, racist anthropology of Edward B. Tylor (1898: Chapter III).

Item: Harding claims (1986:35-6) that science does not try to understand itself in a "naturalistic," self-reflexive way. But how can this thesis be sustained in a book that does not even mention Robert K. Merton, or research entities such as the Society for Social Studies of Science? Similarly, how is it possible to criticize contemporary scientists for an alleged over-indulgence in two-valued (Aristotelian) logic while ignoring the massive literature produced by the general semantics movement of the middle decades this century, a movement that surely had a large and inescapable impact on most contemporary scientists?

Item: Harding (1986:48-52) questions "pure mathematics," claiming that there is not sufficient reason to believe that mathematical expressions are value-free. She realizes, however, that the burden of proof is upon her, and her initial argument is that mathematics, alas, has the weakness of seeking internal proofs--"... no conceptual system," she says, "can provide the justificatory grounds for itself" (1986:49). She does not inform us, however, that there are strong traditions of "external," pragmatic testing in mathematics--e.g., Monte Carlo simulations of statistical models, or predicting the positions of planets in celestial mechanics. Regarding the first of these illustrations, it is possible to say that gambling ("double sixes," for example) did for mathematical statistics what warfare did for nuclear physics. In essence, then, Harding asks us to fault mathematics for having both internal and external "proofs."

Given her remarks about internal proofs, it is surprising that Harding (1986:50) makes the Marxist argument that geometry, for instance, is facilitated by systematic agriculture. Here, I see her point: Once we learn about the properties of lines, rectangles, etc., we may begin to value, say, territoriality, a general commitment to private property. But this argument is problematic because it begs the question. There must have been a powerful cultural drift toward systematic agriculture long before geometry appeared (Cohen, 1977), and therefore we must say to Harding precisely what Malinowski used to say to the Freudians: Your theories about the origins of human culture (e.g., the parricidal crisis as invoked among Freudians) assume that human culture is already highly developed prior to the time when your favorite culture-creating mechanisms come into operation. Agriculture formed our minds long before the invention of modern science.

We encounter further elaborations on the alleged poverty of deductive logic: The classic syllogism--"All men are mortal," etc.--goes wrong in Harding's view (1986:51) because of equivocation in the word "men." If one eliminates the equivocation by substituting the words "human being" for men, however, her argument collapses immediately. Harding moves on quickly, making the usual claim (1986:52) that mathematical formulations in the social realm are rendered impossible by irrationality, the presence of intentions, etc. But one who wishes to make this argument must present a convincing critique of the many quantitative studies of irrationality, intention, and the like found throughout the social science literature; Harding does not do so. In essence, she double-binds me: She gives me a fabulous set of metric tools, and then demands that I throw away my English-standard tools. I don't like this concept, and neither does my 'sixties-vintage truck.
 

Item: After becoming deeply involved, early on, in Harding's claims about what the male-dominated sciences have left out, we are given to understand (1986:60) that "... the public justification for women's colleges was that educated women could raise better sons." This is an insult to the Founding Sisters.

Item: Harding raises several questions such as these: First, "was it entirely a coincidence that sexology began to gain status as a science hot on the heels of the nineteenth-century women's movement and women's agitation to enter science?" (1986:80); second, "is it an accident that the novel ... emerges only in the modern world purportedly ruled by scientific rationality?" (1986:232); third--a question already discussed--does the attitude of science toward nature make rape and torture appear to be a more natural relation of men to women?⁹ What all such questions have in common is that we have not arrived at a way of answering them, and perhaps we never will. Postmodernism may encourage us to answer them as we wish, but to a social scientist these are not meaningful questions unless there is some way of defining them more clearly and ascertaining whether they might be approached, say, in ways implied by Harding. Incidentally, professors of medieval Spanish literature will be interested in the presumption that the novel emerged "only in the modern world."

Item: Harding (1986:114-15) finds fault with Copernicus for having replaced geocentrism (earth woman) with heliocentrism (sun man?). But Copernicus knew well that the heliocentric theory violated cultural traditions that tied together religion, the Ptolemaic view, metaphors of earth-as-mother, and other powerful imagery, and that he would therefore encounter serious danger if he were to break up this lovely logic-tight compartment by telling the inexorable "one true story" of the solar system. Yet, because of the compelling nature of the story itself, he, Galileo, and others persisted in acts of immense courage that tell us nothing about their attitude, or the attitude of science, toward women or womanhood. (Nowadays, of course, there are perhaps 1.3 true stories of the solar system.)

Item: Finally, I know of no evidence that Barbara McClintock's "feeling for the organism" (1986:122), along with many additional trappings of civilization, are not shared by other scientists of comparable excellence and stature. Renato Rosaldo, as we saw, has very strong feelings, but it is not at all clear how these feelings contribute to his anthropological research. Toward the end of her book, in fact, Harding (1986:182-85) makes a major concession, presenting a sort of Bridgman-Koestler-Kuhn argument (Koestler, 1964) about the strength of emotional and irrational elements in science. To me, this negates a large part of the preceding thesis, suggesting that many scientists--most of us, one hopes--share McClintock's admirable human qualities, along with the need for making sure that these qualities do not dominate our thinking.
 

iii
 

As we have seen, Harding (1986:50) supports the Marxist argument that plow agriculture facilitates geometry, and vice versa. This I buy; this I am prepared to sell. If you wish to understand the anteriority, constraint, and exteriority of mathematics--and to understand quantitative structuralism in sociology--seek out the appropriate machinery: plows, solar systems, engines, differential gears, airports, computers. Play with this machinery.

Take, for instance, loglinear analysis, a method that nowadays calls forth the vituperation once inspired by the statistic known as chi-square (Collins, 1984; Denzin, 1989:74). I would listen carefully to the various detractors--and to Harding--if they would take the following steps, and then provide a critique of loglinear methods:
 
 

(1) Gather together, and study for a summer or so, the best literature on the nature and uses of loglinear analysis. I strongly recommend Gilbert (1981). More recently, advances over Gilbert's work have shown, for instance, how to develop loglinear models that encompass the logic of multiple regression. This is not an instance of "fads and fashions," as suggested by Denzin, unless one shows that regression logic is no more important than my mothballed Nehru jacket.
 

(2) Using SAS, SPSS, or other appropriate programs, learn how to request and interpret loglinear output.
 

(3) Raise several theoretical questions that appear to be approachable by loglinear analysis, especially questions involving interaction effects. Work with qualitative variables, for it is thereby made clear that the qualitative-quantitative distinction vanishes within the socially constructed loglinear world. Several discussions of loglinear methods have used the Stouffer research on reference group theory. In my own work I have pursued an issue raised by a local politician who claims that she has been victimized many times by virtue of being black, by virtue of being a woman, and especially by virtue of being a black woman. This is an interactive hypothesis for which loglinear methods are well suited.
 

(4) After selecting a theory, don't obtain real data. Start with contrived data. Create many worlds, and make sure that in some of these worlds your theory seems to obtain, while in others it does not. Make sure that you have a hardware-software combination with a quick turnaround: You may have to do many experiments before you understand how the model reflects your various worlds, or what the model will tell you when a given hypothesis has a greater survival prospect than another. Compare loglinear models with simple crosstabulations, and find out what the latter cannot discern. Replicate, say, Gilbert's analyses, modify his data, and make the models behave predictably in supporting alternative hypotheses. In other words, rotate your crops.

After a season or two, reformulate your critique of quantification in the social sciences, and I'll send you, free of charge, a slightly seedy Nehru jacket.

Near the end of her book (1986:250-51) Harding tells us that science is a black box, and that we must beware the GIGO principle: Garbage In, Garbage Out. Where I live, GIGO now receives a novel and truly significant reinterpretation, one that identifies the real danger: Garbage In, Gospel Out.
 

(4) Solicitous gatekeepers, #1
 

This section illustrates what is likely to occur when we become cavalier about the rules of scientific procedure. The remainder of this book provides additional illustrations.
 
 

Date: 4 December 1992
 

To: Colleagues and bulletin-board forums

From: Michael A. Faia, Sociology, William & Mary

Bitnet: MAFAIA @ WMVM1
 

Re: ASA Footnotes and a new concept of censorship
 

Several weeks ago I sent the following letter to the editorial staff of Footnotes, a publication of the American Sociological Association. The letter is self-explanatory:

(5) Itching and scratching: a Lazarsfeldian digression through an SPSS hologram
 

In my experience, for every "quantiphrene" there is at least one advocate of a more or less exclusive use of qualitative methods. We can usually count on the latter to argue that the biggest problem with the former is a tendency to let the methodological tail wag the theoretical corpus--I hesitate to invoke the usual image. The truth is that one cannot see most theoretical problems clearly unless one has the appropriate quantitative models in mind, and one cannot get quantitative models in mind unless one understands the basic logic of theory. There is a beautiful example in Porter's history of statistics (1986:316). Porter says that analysis of variance, a statistical method developed primarily by Sir Ronald Fisher, could be regarded as "... a theory of heredity." In short, we're dealing with reciprocal causation between theory and method. It is for this reason that books like Analyzing Social Settings: A Guide to Qualitative Observation and Analysis, by John and Lyn Lofland, quickly run up against a dead end (1984:100):
 
 

It is sometimes useful and important to count how often a social unit ... occurs and to develop summarizing statistics in such forms as percentages and means. ... Even though they can be useful, such counting procedures are beyond the scope of this manual. There are many guides available ...; thus, to devote attention to them here would be redundant and would only blur our focus on qualitative analysis.

If counts are sometimes "useful and important," one wonders whether it is wise to ignore them on the arbitrary claim that devoting attention to them would be "redundant"--an extraordinary choice of terminology given the quantiphobic tendencies of this little manual. When the Loflands claim that quantitative procedures "... blur our focus on qualitative analysis" the reader is sorely tempted to point out that the authors could have avoided, through a simple statistical demonstration, the mistaken claim (1984:101) that a low correlation between two variables rules out any possibility of causation between them. A good statistician could readily demonstrate that the same conditions that give rise to "spurious relationships" may also give rise to spurious non-relationships.

In my courses on SPSS I use a simple program that shows how spurious non-relationships can be discerned through the use of partial tables, in a manner reminiscent of Lazarsfeld (1958). Notice that, as was true of many of Lazarsfeld's demonstrations, the variables used by this program are simple dichotomies. That is, they are the simplest sorts of qualitative variables, so that devoting attention to them should not be redundant.

I shall reproduce the program below, because I believe that anybody who runs it will realize that (1) the Loflands are wrong, in the place cited above; (2) the easiest and clearest way to demonstrate either of the two types of spuriousness is to put together an appropriate statistical demonstration, as is done by this program; (3) the easiest and clearest way to resolve subtle theoretical issues--e.g., the place of reciprocal causation in functionalist inquiries (Faia, 1986)--is to create statistical models that replicate what the theories say. Finally, I believe that this program should be thought of as a tiny slice of film from a larger hologram that represents the social sciences generally, and that this tiny slice, as is true of holograms, contains essentially the entire picture.
 
 

DATA LIST FIXED FILE=INLINE RECORDS=1

/1

FREQ1 1-2

FREQ2 4-5

X 7-8

Y 10-11

Z 13
 

VARIABLE LABELS

X 'PROBLEM'/

Y 'ADAPT'/

Z 'LAG'
 

VALUE LABELS

X Y -1 'LOW' 1 'HIGH'/

Z 1 'X TO Y' 2 'Y TO X'
 

BEGIN DATA

90 75 1 1 1

10 15 1 1 2

30 15 1 -1 1

30 75 1 -1 2

30 15 -1 1 1

30 75 -1 1 2

10 75 -1 -1 1

90 15 -1 -1 2

END DATA
 

COMPUTE FREQ1=FREQ1+NORMAL(7)

WEIGHT BY FREQ1
 

CROSSTABS TABLES=Y BY X/Y BY X BY Z

OPTIONS 3 4

STATISTICS 1 2 3
 

COMPUTE FREQ2=FREQ2+NORMAL(7)

WEIGHT BY FREQ2
 

CROSSTABS TABLES=Y BY X/Y BY X BY Z

OPTIONS 3 4

STATISTICS 1 2 3
 

FINISH

This program produces six tables, the last three of which are shown in the appendix. The first table, using FREQ1, demonstrates a spurious relationship. This relationship disappears, in the classic Lazarsfeldian manner, when Z is held constant: There is no relationship in the two partial tables. The fourth table shows a spurious non-relationship: A relationship does appear, in the final two tables, when Z is held constant. The latter kind of spuriousness, contrary to popular opinion, is just as dangerous as the former. For instance, in a functionalist interpretation of a spurious non-relationship, one might argue that the non-relationship occurs because the tendency for X (a problem) to cause Y (an adaptation) is cancelled out by the tendency for the adaptation to reduce the severity of the problem. Think of this as an itching-scratching process: If itching causes scratching, and if scratching reduces itching (this can happen), then it is entirely possible that one will observe a spurious non-relationship between these two variables.

The itching-scratching example is a metaphor for a classic form of functionalist theory. If each of the N=334 social entities in the tables of the appendix were behaving according to the sine/cosine process given in Faia (1986: Chapter 6), and if those N entities were placed in the spurious non-relationship format in which no relationships appear, we would lose sight of the prospect that all values of X approaching 1 may be pushing Y values toward 1 (first partial table, lagged X to Y, wherein itching is now presumably activating a process of scratching), while all values of Y approaching 1 may soon begin pushing X values downward (second partial table, lagged Y to X, wherein scratching is now presumably reducing the amount of itching).

Ironically, the Loflands (1984:113) provide an excellent example of precisely this sort of process. In a brief discussion of functionalism, they develop the example of "deviant roles among policemen," pointing out that "... the performance of certain unpleasant tasks is necessary for the maintenance of the police department." As is often the case in functionalist explanations, "maintenance" is not clearly defined. However, it would be appropriate in my SPSS example to define the Y variable as "unpleasant tasks" (i.e., scratching) and the X variable as a problem (i.e., itching) involving maintenance of something within each of N=334 police departments: Arrest records, for instance, adequate to meet legal demands and requiring many person-hours of onerous labor. Within the framework of purely qualitative methods I do not see how the Loflands could have done anything more with this problem than what they did: They more or less defined the key variables. But social science theories involve a lot more than the mere definition of key variables.¹⁰

I maintain that, in demonstrating functionalist theory, no amount of verbal vigor would be as effective as the Lazarsfeldian tradition of playing with partial tables. What is required in this sort of didactic exercise is an appropriate combination of verbal and quantitative insight. When authors and publishers produce books and articles on (purely) qualitative sociology, when undergraduate courses are organized around these proliferating texts, when library references such as Sociological Abstracts classify and code many social science publications as exemplifying (purely) qualitative sociology, we are back to the old high-school bifurcation, the old tracking practices that separated those who wished to take qualitative analysis for their chemistry requirement from those who seemed qualified for the more rigorous and realistic approach. A widely used textbook suggests that quantitative and qualitative approaches to chemistry, ideally, should be "entwined" (Harris, 1987:xiii). Ditto for sociology.
 

(6) Transcending the transcendentalists
 
 

Seidman's [1991] criticisms are nothing compared to the active disregard by theorists in many other fields--literature, feminist studies, politics, anthropology, history, and more--who once looked first to sociology and now hardly notice us at all.

--Charles Lemert (1991:164)
 

This dreadful indifference has been breached in a recent article by Gross (1989), who demonstrates what he considers a proper use of sociological findings: The 1850 United States census can be used to help students understand the lifestyles of the transcendentalist writers of New England. Gross says (1989:504) that one of his major goals as a scholar and teacher is to help reconcile the "two-cultures" syndrome that separates students who excel in the natural sciences from those who excel in the humane disciplines. This is surely a worthy objective, but Gross (1989:512) quickly gets into trouble when he argues that
 
 

By placing the census back into history, my class was led to an important insight into quantitative analysis: numbers are never enough; they must always be restored to the context from which they emerged. Following that intellectual path, we went on to seek in Emerson's biography an understanding of how he and Lidian had fashioned their household over the years ...

The trouble arises because Gross does not make it clear that these same sentences can easily be stood on their head; in other words, he is describing an intellectual process that is capable of running in reverse.¹¹ If Gross's remarks above make sense, then it also makes sense to say that
 
 

by placing biography back into demographic history, my class was led to an important insight into qualitative analysis: biographies--highly detailed case histories--are never enough; they must always be restored to the larger social context from which they emerged. Following that intellectual path, we went on to seek in the census tabulations an understanding of how Emerson and Lidian and other such families had fashioned their households over the years ...

Gross shows us, in brief, that census tables are subject to interpretation. But why not show us also how interpretations are subject to census tables? Reading about the Emersons and their domestic life, a clever sociologist would quickly arrive at the hypothesis suggested by Gross' Table 1 (1989:509): In the days of the transcendentalists a young man was expected to become head of a household, perhaps a very large household, but was rarely able to do so prior to reaching the age of about thirty or thirty-five. Similarly, anybody who reads the feminist literature of the early years of the twentieth century and discerns the close relationship between feminist issues and those of the temperance movement, would probably be delighted to discover a tattered census volume (Bureau of the Census, 1908) that fell into my hands a few years ago. This book contains a mammoth table, continuing for hundreds of pages, that permits the testing of a classic feminist-temperance interpretation: That whenever large numbers of divorces involve male dominance, it is also likely that large numbers of divorces will involve "drunkenness." The method of analysis would be correlational, while the inspiration might well have been biographical.

On the Emerson question: If you had one unit of time and energy for study, and devoted .5 to the 1850 census and .5 to Emerson's family, you would probably learn the equivalent of (.5)(.5)=.25, a reasonable proportion of what you need to know. If you devoted .1 to the one topic and .9 to the other, you would learn (.1)(.9)=.09, a far smaller proportion of what you need to know. (Waldo would love this formulation.)

In conclusion, I believe that sociology should continue to emphasize macrostructures, numbers, systematic measurement, quantitative models, empirical data, and carefully selected samples. But perhaps my view on these matters is perverted, and I must confess that in my "methods" classes and elsewhere I love to defend the following truth: For beauty, for clarity, for elegance, for immense power combined with subtle intricacy, there never was a sonnet written by Petrarch or Shakespeare or Elizabeth Barrett Browning, there never was a poem of any genre nor will there ever be a poem of any language, that will surpass the IRS 1040 Long Form. This glorious set of documents is the ultimate social science instrument: It is highly complex, with elaborate interrelationships among its many parts, and yet it is parsimonious; it is incredibly informative, yet without overtaxing the mental capabilities of intelligent respondents; it is verbally clean and sharp, and it is mathematically clean and sharp; in its careful construction, it embodies the insights and inspirations of a long series of scientific inquiries. It is a work of the social sciences, and it is a work of art.

And one final thought: Given the sophisticated research that it makes possible, the IRS 1040 Long Form expresses our high estimation of the capabilities of the "object" of study and our sense that this particular respondent has a strong commitment to the social contract--if a social contract may be said to exist.

I recently had occasion to express openly my convictions and feelings about the IRS 1040 Long Form. The opportunity came during an academic seminar, a weekend retreat, that involved several colleagues charged with the task of developing an interdisciplinary graduate program in American Studies that will span American history, literature, and popular culture, along with several social science fields. As it turns out, all participants in this program have a lot to learn, especially from one another. To me, for instance, the history of "vernacular architecture" in the United States is a highly esoteric realm, one in which I will probably never feel at home, and one in which I may never have the temerity to undertake a research project. But I was nonetheless mildly shocked to discover that several of my humanities colleagues, when they first heard about the sort of interview that occurs in the NSFG, would react in a way that reminded me of Philip Roth's (1971) account of Richard Nixon's reaction upon learning about the sexual practices of homosexuals. My colleagues could not believe that social scientists would have the bad taste to ask highly detailed questions about the marital experiences, sexual behavior, and fertility histories of a cross-section of American young adult women. It was my feeling that this sort of culture shock must be overcome quickly in a cross-disciplinary program, if it is to succeed.

I should point out that the NSFG came up during a discussion of my introductory seminar on American Studies. The discussion followed a cocktail party, and my interpretations of the discussion may have been clouded by vernacular inebriation. However, it would be no more difficult for me, while sober, to explain sociology to my postmodernist colleagues than it would have been for Rosaldo (1989:3-4) to explain exchange theory to Ilongot headhunters. If only he had been willing to take the time.
 

(7) What do you presuppose, and when did you presuppose it? The Sisyphus of the social sciences
 

Jeffrey Alexander's Theoretical Logic in Sociology is a multi-volume opus in which, among other tasks, the author elaborates his reasons for believing that a major trouble with the social sciences is that they try too hard to emulate the natural sciences; the thesis is developed further in a recent work (Alexander, 1988). Alexander (1982:7-8) is afraid that if we follow the advice of Hans Zetterberg and attempt to set up systems of knowledge replete with axioms, deductions, and theorems, the results are likely to be embarrassing if not disastrous. He believes that the empirical and positivist emphases of contemporary sociology have destroyed any deep interest in theoretical matters of the most abstract nature; he cites (1982:11-12) Robert K. Merton's advocacy of theories of the middle range as a case in point. Although Alexander rejects "human studies" as an alternative to positivist social science (1982:15-18), he believes that we must move toward an "alternative conception of science" in which we recognize, with Alexandre Koyre, that "experimentation is a teleological process in which the goal is determined by theory" (1982:24). As an illustration of Koyre's thesis Alexander cites the case of Einstein and relativity, claiming that the famous Michelson-Morley experiments, contrary to conventional wisdom, were not a major source of inspiration for Einstein (1982:31). And even if social scientists were to clean up their act and begin practicing science a little more honestly (or with a little less self-delusion), they would not necessarily achieve any insight into the larger philosophical issues that interest Alexander.

In Alexander's view, only those scholars "who have thoroughly and explicitly broken from the positivist persuasion will even attempt an exercise in general theoretical logic" (1982:36)--this overstatement recalls Collins' claims (1984:330) about intolerant statisticians. And once having made the crucial break, Alexander's rehabilitated former positivists soon realize that questions of "action and order," which pertain primarily to the place of human agency and free will as causes of social phenomena, "... represent the true presuppositions of sociological debate; they establish a general framework that cannot be subsumed under other kinds of theoretical dispute and, at the same time, they manifest properties that decisively affect sociological thought at every level of the theoretical continuum" (1982:65). Having deepened our understanding of action and order, we shall come to a clearer appreciation of the "nonrational elements in science," realizing that "... science itself gives the lie to conceptions of action which presuppose instrumental rationality" (1982:75).

None of these arguments is new, and none is at all persuasive. In the contemporary social sciences one finds elaborate deductive models--for instance, stable population theory, learning theory, computer simulations of "world systems" or national economic systems or other complex elements of societies or human psyches--that make the Zetterberg approach appear primitive, anachronistic, and to be a mere historical relic.¹² One can easily argue that Robert Merton has been the most important philosopher of science ever produced by American sociology, and that his contributions to "middle-range" theory have not detracted in any way from this accomplishment. When Alexander affirms, as he did at a 1986 symposium, that no significant difference exists between "disaster research" in the social sciences and Shakespearean studies as pursued by English departments, he reveals a surprisingly strong willingness to assimilate the otherwise disparaged modus operandi of human studies. When he cites Koyre's claims about the teleological character of scientific experimentation, he does not bring to our attention the many instances in which social scientists have abandoned major hypotheses on the basis of disconfirmatory evidence--Barber (1987) provides several such instances¹³--nor does he bother to tell us that we could probably rid ourselves of additional untenable notions if we were to emulate more often the natural scientist's routine practice of carrying out replications.

As I've mentioned, our recent repudiation of racism is a case in point, especially when one considers that what social scientists were saying about race, a century ago, would nowadays pass muster with a Kluxer.

On the matter of "general theoretical logic," Alexander does not produce a trace of evidence for his own presupposition that scholars of the "positivist persuasion" show little if any interest in general theoretical issues. My experience, on the contrary, has been that the best social science philosophers tend to be statisticians with a strong empirical and positivist bent and with a scientist's skepticism; Sir Ronald Fisher is my ideal type. Alexander never tries to persuade us of his claim that questions of action versus order are the central presuppositional issue, nor that one's manner of resolving this issue will "decisively affect sociological thought." Again, I have equally compelling evidence that one's beliefs about human agency have virtually nothing to do with one's performance as a social scientist, except in the sense that ridiculous, self-delusional beliefs may sometimes cause us to ignore reality while maintaining excellent intellectual interchanges with those similarly deluded. Finally, the rationality-versus-nonrationality issue, as I tried to show in my earlier discussion of Economic Man, is a red herring: The social sciences focus on both types of behavior and ideally show no favorites. Beyond that, if instrumental rationality emphasizes the means-ends nexus, and if (as is perfectly clear) the means-ends nexus is merely a special instance of the cause-effect nexus, then I must point out that I have learned far more about instrumental rationality from statisticians such as Sir Ronald Fisher, who clarify causal analysis, than from philosophers such as Jeffrey Alexander, who do not.

And one minor point: If Alexander really believes that "... science itself gives the lie ..." to instrumental rationality, then either (1) his claim about a social-science bias toward rationality goes up in a poof or (2) he has to argue that social scientists do not pay attention to their own findings. I lean slightly toward the first option.

Alexander's approach to the social sciences leads us dangerously astray, and I believe that it does so for a very simple reason: It almost never grapples with real, day-to-day, nitty-gritty problems of social science research and analysis. I don't have any objection to the idea that a small coterie of scholars might practice Alexandrian sociology, but it would be most unfortunate if they enticed the rest of us to follow their lead.

Take the case of sex preselection, the process in which parents try to control the sex of their offspring (Bennett, 1983). An ordinary social scientist struggling to understand the nature, determinants, and consequences of sex preselection is far more likely to learn something about presuppositions of a philosophical genre than is Jeff Alexander to learn anything about preselectivity of an infant's gender. We cannot understand the invention of sex preselection methods unless we first ask whether inventions occur more or less by accident or because "necessity is the mother of invention." We cannot understand the willingness of parents to use sex preselection unless we first understand the marketability of the methods, their efficacy, and their diffusion through society. We cannot understand the ways in which potential parents negotiate their preferences for boys or girls without first considering interpersonal power and other aspects of interaction of men and women in marriage (Charles, 1993). And these several questions of human agency are a mere exercise in sound and fury if sex-preselection methods, evaluated objectively, do not work, if they are too costly or cumbersome, or if they have damaging side effects. Finally, we cannot understand the demographic and social effects of sex preselection unless we first develop an appreciation of social structure, recognizing on the basis of etic exercises in social simulation that a society with a large excess of marriageable males (or females), yet valuing monogamy and the one-child family, has a large problem.

The problem, to be sure, has to do with "free will," involving in this instance whether or not we marry the person we prefer at a time we prefer. Whenever the marriage market is tight we experience, under an appropriate operational definition, less free will--it is a matter of degree. Alexander (1982:101), by contrast, tells us that Talcott Parsons once addressed the free-will conundrum in the following either-or language:
 
 

Positivistic ... thought is caught in the "utilitarian dilemma." That is, either the active agency of the actor in the choice of ends is an independent factor in action, and the end element must be random; or the objectionable implication of the randomness of ends is denied, but their independence disappears and they are assimilated to the conditions of the situation, that is to elements analyzable in terms of nonsubjective categories ....

Or, we resolve the dilemma by taking what I shall call the statistician's approach: (1) We know, a priori, that the ends involved in mate selection are neither "independent" of nor entirely determined by "the conditions of the situation"--they contain a measure of both; (2) therefore, we write a multivariate equation that enables us to explain mate preferences as influenced by "nonsubjective categories" such as the sex ratio and (why not?) various subjective categories, and also enables us to isolate those elements of mate preference that we cannot explain, elements that appear to be random; (3) we then analyze the mate-selection market, perhaps using multivariate statistics again to ascertain whether mate preferences can be fulfilled to an appreciable degree; (4) in a situation where there is a strong preference for, say, endogamous marriage, but where we find that ego's characteristics (or preferences) are a poor predictor of mate's characteristics, we ask ego, in a detailed way, whether he or she has a sense of lacking freedom of choice in mate selection. Whenever we receive a large number of affirmative answers, and especially when these answers seem to have a relationship to largely non-subjective conditions such as the sex ratio, we conclude that free will has been diminished--although we might wish to call the process by another name. This, to me, is the only definition of free will that has any utility in the social sciences, and the Parsons-Alexander formulation simply obfuscates. Colleagues who do not like either approach are free to reject them, of course, in a willful act of polymorphous perversity.¹⁴ In any case, an excellent discussion of the statistician's way of approaching questions of free will is to be found in Berger (1963: Chapters 4 and 5).

But perhaps I only say these things because I do not know, nor have I ever known, what Parsons meant by "randomness of ends." Again, a matter for hermeneutics. On the other hand, I do understand what physicist Stephen Hawking means when he says that many questions lie outside the realm of science, and that pursuing such questions is a little like asking what takes place a mile north of the north pole. Refusing to heed Sica's (1988:220) "... need, nay, demand to know unknowables ...," the wise scholar heads south, where he has a prospect of learning something.

Alexander shares Sica's aspiration: He wishes to know unknowables. In many ways he has made himself the Sisyphus of the social sciences: He sets up the goal of becoming a sociologist because this is a way of understanding human action and social order; then, he declares that the goal is unattainable because man is the ultimate mystery. Although it is to Alexander's credit that he shows deference to Heisenberg (Alexander, 1988:80), he generally emulates the various anti-positivists, discussed earlier, who insist upon the complexity of the human and the comparative simplicity of nature:
 
 

Insofar as the objects of a science are located in the physical world outside of the human mind, its empirical referents can, in principle, more easily be verified through interpersonal communication. For social science, the objects of investigation are either mental states or conditions in which mental states are embedded. For this reason, the possibility for confusing mental states of the scientific observer with the mental states of those observed is endemic. This is the social science version of the Heisenberg Uncertainty Principle.

Analyze this statement in detail: It does not make sense. Let the "conditions in which mental states are embedded" be an airliner, or a simulator of an airliner. (It is relevant here that aviation simulations generally put social science simulations to shame.) Suppose that we are studying two-person interaction involving pilots and first officers (co-pilots) under emergency conditions, or under conditions in which a mistake or a distraction could easily create an emergency. We "observe" mental states in the usual ways: We infer them from recorded man-machine interactions, from recorded dyad interactions, from direct observation, from post-experimental de-briefings, and so forth (O'Hare and Roscoe, 1990). When I think about this research setting, I cannot imagine why my mental state, as observer, is at risk of being confused with that of the pilot or first officer. If high risk were demonstrated, we would take the sorts of methodological steps routinely used by psychology: Design double-blind experiments, test for halo effects, and so forth. And remember: These kinds of problems, and these kinds of proposed solutions, are not confined to the social sciences.

In short: A carefully designed experiment creates circumstances under which subjects are likely to fail a "Turing test"; that is, they will not know whether the experiment is conducted by man, woman, or machine. Turing tests, of importance in connection with artificial intelligence, will be discussed in some detail in the next chapter.

The silliness keeps rolling over us. Alexander continues (1988:81): "... while precise empirical measurements of two variable correlations can sometimes be established, it is rarely possible for such a correlation to prove or disprove a proposition about this interrelationship that is stated in more general terms." Of course it is, and this is precisely why we have learned to use some variant of the Lazarsfeld strategy for generalizing findings. In his ensuing discussion of Blau's demographic structuralism, for instance, Alexander points out that rates of "outgroup relation" are said to be influenced by the relative size of various population segments. But the concept of outgroup relations, he says, "... does not have a clear-cut referent" because Blau relies too heavily on exogamy as an indicator of such relations. Again: This is nonsense. Assuming that Blau has not yet falsified his own hypothesis--and even if he has--it is incumbent on the rest of us to design studies that will use indicators of outgroup relations other than exogamy, such as forming business partnerships, attending the same church, etc. That is, we generalize the hypothesis. We then apply it to different populations, different kinds of outgroups, and so forth, thereby generalizing it further. This entire process of generalizing findings is discussed in a later chapter of this book, in relation to social science taxonomies.

Enough! I always felt that Sisyphus' major problem was that he would suffer a crisis of confidence, and allow himself to be run over by very small stones.
 

(8) The meaning of politics and the politics of meaning
 
 

I was once told by a theorist who denied the possibility of correct interpretation that I had not interpreted his writings correctly.

--E. D. Hirsch, Jr.
 

The theorist, of course, probably had misinterpreted Hirsch; this merely compounds the problem. Novick (1988:278) alludes to the same problem when he points out that relativism cannot become an absolute truth, lest it refute itself.

I have no objection to the idea that "meaning" has a large role in the social sciences. Indeed, it would be impossible to develop any science--especially the social sciences--without devoting appropriate attention to semantics, semiotics, interpretation. My quarrel is with inappropriate attention, and with the notion that an emphasis on meaning requires hostility toward science. In many instances human subcranial phenomena have essentially nothing to do with the subject under investigation--and I willingly concede all the necessary Kantian caveats and disclaimers.¹⁵ Michotte's (1963) famous research into how we perceive causation made it clear that causal interpretations are primarily the ways in which our minds impose order on a complicated, recalcitrant world. Any good statistics professor makes the same point time and again, but in making this point one surely does not have to abandon science; in fact, one must continue to have the scientist's proper humility in the presence of a world in which the human species has a small, uncertain, and confused role.

Again, it is the ego-tripping of the anti-science factions that I find especially onerous.

When scholars argue, as they do so often nowadays, about the relative importance of "interpretive" and "structural" phenomena, it is not unusual for the argument to take a political turn. Somebody is likely to proclaim that structuralists are propounding "iron laws," denying free will, insisting strongly and wrongly that the limitations of these laws will never be overcome. Somebody else is likely to counter with the argument that we control our own destiny by recognizing that iron laws do not exist, and that the fundamental realities have to do with our ideas about a world that remains a "blooming, buzzing confusion" until such time as we define or re-define it. An astute structuralist is then likely to reply that interpretations that enable us to ignore objective social and natural realities, whether or not these realities involve iron laws, are nothing more than an opiate of the masses and that the masses will never be free until they stop deluding themselves. The "opiate of the masses" thesis, interestingly, often carries the implication that power elites--the exploitative class--have a clear perception of whatever laws may exist, iron or otherwise, and have learned how to make use of these objective realities. The exploiters, in other words, understand politics. Eventually their emics may become the etics of the rest of us. When Anthony Giddens accuses structuralists of regarding people and their perceptions as "bloody silly" he fails to realize that this is surely not our attitude toward the exploitative classes (cf. James, 1984:119).

The above scenario, by accident, eventually associates structuralism with a sort of "liberalism" and interpretive sociology with a sort of "conservatism." But one could readily design an alternate scenario in which either structuralists or interpretationists--you pick 'em--would be tainted by an association with the extreme left, the extreme right, or the extreme center.

What science seeks to do is to make all of us members of the power elite. It proposes to do so by forcing us to abandon meaningless questions such as the relative importance of interpretive and structural phenomena, the centrality (or lack thereof) of human agency, rationality, irrationality, or nonrationality; as far as I'm concerned, Bourdieu's "constructivist structuralism" (Turner, 1991:508-18) provides a fine resolution of this question. Once having abandoned this futile inquiry, we will necessarily devote less time to the equally futile search for some sort of correlation between one's politics and one's stand on the meaning-versus-structure conundrum. When Parson Malthus tells us, as he did in the several rapidly evolving editions of his famous essay on population, that we had better not run afoul of the laws of population growth but that, paradoxically, we willfully overcome these laws by exercising "moral restraint," we realize that we are in the presence of a scientist whose ideas evolved on the basis of new evidence (Petersen, 1979), who therefore had a reasonable prospect of overcoming his various biases, and whose work cannot be elucidated at all by those who insist upon teasing out of it some sort of stand on meaning-versus-structure, or some sort of political ideology.

The labels traditionally attached to Malthus, political or otherwise, are no more justifiable than those attached to contemporary scholars who make the sad mistake of identifying themselves with particular structuralist or interpretive movements; for this mistake one may deserve a friendly excoriation, but one does not deserve to be labelled as a carrier of easily diagnosed political values. To paraphrase an earlier remark of my own: If every political label can be affixed to Malthus, then no political label can be affixed to Malthus. As Collins implies (1987:182), the relationship between one's politics and one's theories is best characterized by the null hypothesis: Often, there is no demonstrable relationship. As an illustration, Collins points out that "the mixture of positivism and Establishment politics that [existed] in the 1950s and 1960s was particular to that historical epoch. In the 1700s, the situation was exactly the reverse: it was the proscience faction--Voltaire, Condorcet, the Jacobins, Bentham and the utilitarians--who were the revolutionaries and reformers ..." If regress we must, let us go back to the Enlightenment!

As we try, perhaps from a Malthusian perspective, to explain fertility trends in contemporary third-world nations, moral restraint doubtlessly has a role. But if we continue shouting at one another about the a priori importance of moral restraint, or about whether good guys or bad guys emphasize it, and if we continue to do this as a substitute for studying moral restraint scientifically, we shall never understand its role. One of my mentors used to teach his students that society is like a baseball game. He spent hours drawing out the various analogies, and his presentation was most edifying. He never tried to decide whether the distance between the bases was more important, or less important, than Casey Stengel's latest philosophical breakthrough. Beyond that, we learned a lot about his attitude (and Stengel's) toward science, and little about his politics or about what happens a mile north of the north pole.
 

NOTES (Chapter 1)
 

(1) I adopt, in combined form, all of Halfpenny's (1982:11) various definitions of positivism: "Sometimes, to be positivist means no more than to be scientific ... Sometimes, positivist sociology is synonymous with statistical analysis ... Sometimes, to practice positivist sociology is to seek to establish causal explanations, or to search for fundamental laws ..., or to insist upon objective empirical information systematically organised to generate or test hypotheses." Halfpenny (1982:32) implies that these dimensions of positivism have not always gone together. The real power of the positivist tradition, in my view, derives from its having combined these dimensions.
 

(2) MacLaine, however, is not the only scholar thus afflicted. White (1989:27) tells us that "there is no necessity, logical or natural, governing the decision to emplot a given sequence of [historical] events as a tragedy rather than as a comedy or romance."
 

(3) Discussing the popularization of Jacques Derrida in the United States, Lamont (1987:614) suggests that "... like Foucault or Habermas, Derrida offered American humanists a criticism of science that was much needed to promote their own intellectual products." If these humanists compete for resources against social scientists, then it is probably most unwise for us to make remarks such as the following, from a sociologist at Yale University (Stanfield, 1987, xi):
 
 

As we near the end of this century, it is becoming apparent that there are inherent dangers in embracing rationalistic approaches to the human world uncritically whether they are found in the social sciences, journalism, corporate management, or in policy studies.
 

We have been reminded in tragic ways of what happens when we reduce human beings down to path analysis, probability models, and cost/benefit analyses: the escalation of the Vietnam War; the Union Carbide catastrophe in India; in Institute, West Virginia; the destruction of the space shuttle Challenger; and the expansion of an American underclass--disproportionately black and latino--no one seems to know what to do with. In all of these cases, "policy blunders" have been traced back to human mistakes, human coverups, human biases, and human limitations.

As a social scientist, Stanfield is shooting himself in the foot while tying his hands behind his back. But the most interesting aspect of these remarks is the way in which the last sentence contradicts everything that precedes it. Most of the mistakes, coverups, biases, and human limitations that I've come across did not result from rationality, path models, probability theory, and the like. They occurred primarily because human minds and social organizations were not behaving rationally. But given the ambiguity of Stanfield's statement, it is perhaps inappropriate for me to use the words "result from": This author is not proposing a causal hypothesis; rather, he is proposing a way of feeling.
 

(4) In a more easterly direction, sociology is said to be flourishing. See Shelley (1989). The practices and perspectives of the social sciences are now widely diffused; fortunately, it is therefore possible to love sociology without loving sociologists.
 

(5) To be sure, it is possible to create a pseudo-complexity by playing games such as the following (Lukes, 1992:425-26):
 
 

[The authors] ... treat their composite Lockean individualism as a narrowing and distorting ideology. In this vein, they have much of interest to say, in condemnation ... of cost-benefit analysis as applied, among other things, to the value of life, ... as when, for instance, an Environmental Protection Agency economist, asked "What about the theory that human life is priceless?" answered, "We have no data to support that."

Our better seniors, having had a methods course or two, provide a much more compelling answer than does this inept EPA economist: The idea that human life is priceless, they would say, means nothing until we carry out the difficult tasks of defining this idea in sociological terms.
 

(6) I recently had an opportunity to read about and discuss "black sociology" in some detail (Staples, 1976). Black sociology, like feminist standpoint sociology, has much to recommend it, but it is seriously impaired by the way in which it tries to displace mainstream sociology with specialized interpretations that allegedly can be comprehended only by blacks, or by black scholars. Black sociology, if widely adopted as a unique interpretation inscrutable from a general social science standpoint--i.e., "ambiguous" in Levine's (1985) sense--would be a disaster both for sociology and for blacks. On the other hand, if black sociology is primarily an attempt to help the discipline overcome biases and prejudices, then it makes the same contribution to science that I attribute to Rosaldo earlier.

A major objective of black sociology is to attack racism, and a major objective of feminist sociology is to attack sexism. It would not be at all surprising if the style of sociology discussed in the next chapter, that of Michel Foucault, were enlisted for the campaigns against homophobia. One cannot help but notice that these unique orientations and interpretations, despite their strong commitment to social change, have trouble cooperating with one another. Harding (1986, 1991) has made little progress in showing why black women should support feminism, and Sawicki's (1991) study of Foucault's contribution to feminism is singularly disappointing: Mainly, we learn that although the "new reproductive technologies," for instance, may have encouraged American physicians to consolidate control of reproductive behaviors of women (1991:76), it is entirely possible for women to resist this control (1991:80-94). One hopes that those who do resist will not be influenced by the untenable claim (1991:76) that during "... the first decade of this century, physicians waged a successful campaign against midwives ..." and that "... maternal and infant mortality rates actually increased after midwives were eliminated."

Incidentally, I should like to point out, with Jesse Jackson, that if everybody is somebody, the main conclusion is that everybody is somebody.
 

(7) Caveat: "When men want androgyny, they usually intend to appropriate selectively parts of 'the feminine' for their projects, while leaving the lot of real women unchanged" (Harding, 1986:145).
 

(8) I know little about the interpretation of cannibalism from the standpoint of cannibals. From the standpoint of the victims, however, I would surmise that there is only one true story.
 

(9) Regarding novels, the rape-of-nature thesis, androgyny, speaking freely, and Harding's acknowledged problems (Chapter 7) in trying to relate feminism to the needs and epistemologies of the "third world," I note that one of our colleagues, a novelist as well as a sociologist, has written the following description (Greeley, 1990:134) of a situation in which a preacher has just delivered a homily on how to protect "little people," such as crickets, grasshoppers, men, dragonflies, women, and frogs, from the cold of winter:
 
 

Jamie told it with marvelous skill, recounting it as something he had heard from an Indian storyteller and then putting himself into the storyteller's role (a "Keeper of the Talking Sticks") so that he actually became a Chippewa/Cree. ...
 

"So the old man said to the little children, 'Look up at the mountaintops. They are the breasts of Earth, our Mother. She has already taken off her clothes and bared her breasts to the cold so she can wrap us up in her garments and keep us warm during the winter.'"
 

No applause ..., just stunned silence.
 

You don't talk about baring breasts in church--not even if Jesus describes himself as a nursing mother in the seventh chapter of Saint John's Gospel.

It is entirely possible that we feminists will have to learn to embrace the ancient metaphor (or archetype) of earth as mother.
 

(10) Unfortunately, even a few statistics textbooks do not realize the importance of spurious non-relationships. Runyon and Haber (1988:174), for instance, say that "when low correlations are found, we are strongly tempted to conclude that there is little or no relationship between the two variables under study. However, ... the failure to find evidence of a relationship may be due to one of two possibilities: (1) the variables are in fact unrelated or (2) the variables are related in a non-linear fashion." A third possibility, of course, is that the lack of an observed relationship is spurious.
 

(11) Throughout this book, I evaluate arguments by re-stating them in reverse. It is a trick I learned from physicists who experiment with the idea of running time in reverse.
 

(12) For a contemporary illustration of an axiomatic formulation that relies heavily on the Zetterberg approach, see Turner (1986:441-53).
 

(13) See, for instance, Barber (1987:166), which explains how Alan F. Westin's research forced him to change his mind about the impact of computers on the right to privacy. Another excellent example is provided by Fogel (1989:391), who at one time subscribed to a functionalist incompatibility hypothesis that he later was compelled to abandon:
 
 

Between 1968 and 1972 my view of the moral problem of slavery changed considerably. The discovery that slave plantations were more efficient than free farms challenged my beliefs ... The notion of efficient plantations created a dilemma because it was difficult to see how individuals so deeply oppressed ... could nevertheless produce more output per worker than free farmers. It ... seemed to imply a far greater level of labor discipline and a far greater degree of ... acquiescence to the objectives of the planters than I was prepared to entertain.

(14) Twice now--here and in footnote 3--I have adumbrated my feelings about Levine's (1985:15-19,218) hypothesis that one of the functions of ambiguity is "... the bonding of a community through diffuse symbols ..." I have serious doubts about this hypothesis. Levine (1985:14) also expresses agreement with Richard Dewey's claim that sociologists have a "... pervasive indifference ... toward the absence of semantic consensus in the terms they use ..." But this claim is no more convincing than Sica's (1988) insistence that we tend to ignore non-rational or irrational aspects of behavior. The two positions, in fact, are contradictory: If we had Levine's hypothesized indifference toward ambiguity, then by virtue of that very attitude we would be making a strong commitment to irrationality.
 

(15) Rorty, in fact, confirms my own Kantianism when he describes Derrida, with admiration, as a kudzu overgrowing the Kantian edifice (Rorty, 1991:235). We are now engaged in a great battle: cutting back the kudzu. The original edifice looked a hell of a lot better.
 

REFERENCES (Chapter 1)
 

Adler, Mortimer J. 1985. Ten Philosophical Mistakes. New York: Macmillan.
 

Alexander, Jeffrey. 1982. Positivism, Presuppositions, and Current Controversies. Berkeley: University of California Press.
 

Alexander, Jeffrey C. 1988. "The new theoretical movement." Pp. 77-101 in Neil J. Smelser, editor. Handbook of Sociology. Newbury Park: Sage.
 

Andersen, Margaret L. 1987. "Changing the curriculum in higher education." Signs 12:222-54.
 

Andersen, Margaret L. 1988. Thinking About Women: Sociological Perspectives on Sex and Gender. New York: Macmillan.
 

Barber, Bernard. 1987. Effective Social Science: Eight Cases in Economics, Political Science, and Sociology. New York: Russell Sage Foundation.
 

Barzun, Jacques. 1964. Science: The Glorious Entertainment. New York: Harper & Row.
 

Becker, Howard S. 1987. "The writing of science." Contemporary Sociology 16:25-27.
 

Bellah, Robert N. et al. 1985. Habits of the Heart. New York: Harper & Row.
 

Bennett, Neil G., editor. 1983. Sex Selection of Children. New York: Academic Press.
 

Berger, Peter L. 1963. Invitation to Sociology. New York: Doubleday.
 

Bologh, Roslyn W. September, 1990. "Rejoinder on feminist theory." Footnotes (American Sociological Association) 18:6.
 

Bordo, Susan. "The Cartesian masculinization of thought." Pp. 247-64 in Harding, Sandra and Jean F. O'Barr, editors. 1987. Sex and Scientific Inquiry. Chicago: University of Chicago Press.
 

Brookmeyer, Ron. 5 July 1991. "Reconstruction and future trends of the AIDS epidemic in the United States." Science 253:37-42.
 

Browne, Harry. 1973. How I found Freedom in an Unfree World. New York: Avon Books.
 

Bureau of the Census. 1908. Marriage and Divorce, 1887-1906 (Part II, General Tables). Washington, D. C.
 

Cancian, Francesca M. October, 1990. "What is feminist theory." Footnotes (American Sociological Association) 18:5.
 

Charles, Gilbert. 18 February 1993. "Fille ou garçon: la cuisine de Mister Liu." L'Express (No. 2171):22-23.
 

Cohen, Mark Nathan. 1977. The Food Crisis in Prehistory: Overpopulation and the Origins of Agriculture. New Haven: Yale University Press.
 

Collins, Randall. 1984. "Statistics versus words." Pp. 329-62 in Randall Collins (ed.), Sociological Theory. San Francisco: Jossey-Bass.
 

Collins, Randall. 1987. "Looking forward or looking back?: reply to Denzin." American Journal of Sociology 93:180-84.
 

Cook, Thomas D. and Donald T. Campbell. 1979. Quasi-Experimentation: Design and Analysis Issues for Field Settings. Chicago: Rand McNally.
 

Dawkins, Richard. 1986. The Blind Watchmaker. New York: W.W. Norton.
 

De Janvry, Alain and Ann Vandeman. 1987. "The macrocontext of rural development: A second view of the U.S. experience." Pp. 83-109 in Bruce F. Johnston et al., editors, U.S.-Mexico Relations: Agriculture and Rural Development. Stanford, CA: Stanford University Press.
 

Denzin, Norman K. 1989. The Research Act: A Theoretical Introduction to Sociological Methods. Englewood Cliffs: Prentice Hall.
 

Diggins, John P. 1990. "Science and the American experiment: How Newton's laws shaped the constitution." Pp. 3-5 in Schellenberg, Kathryn, ed., Computers in Society. Guilford, CT: Dushkin.
 

Donovan, Josephine. 1985. Feminist Theory. New York: Frederick Ungar.
 

Faia, Michael A. 1986. Dynamic Functionalism: Strategy and Tactics. New York: Cambridge University Press.
 

Faia, Michael A. February, 1990. "The trouble with feminist theory." Footnotes (American Sociological Association) 18:8-9.
 

Flippo, Chris. 1987. When Women Say Yes: Sexual Initiation, Minority Status, and Opportunity Costs. Unpublished master's thesis, College of William & Mary.
 

Fogel, Robert W. 1989. Without Consent or Contract: The Rise and Fall of American Slavery. New York: W.W. Norton.
 

Fogel, Robert W. and Stanley L. Engerman. 1974. Time on the Cross: Evidence and Methods--A Supplement. Boston: Little, Brown.
 

Georges, Michèle. March 11, 1993. "Davis Inlet: ici, on se suicide ..." L'Express (No. 2174):32-36.
 

Gilbert, G. Nigel. 1981. Modelling Society: An Introduction to Loglinear Analysis for Social Researchers. London: Allen & Unwin.
 

Greeley, Andrew M. 1990. The Cardinal Virtues. New York: Warner.
 

Gross, Robert A. 1989. "The machine-readable transcendentalists: Cultural history on the computer." American Quarterly 41:501-521.
 

Halfpenny, Peter. 1982. Positivism and Sociology: Explaining Social Life. London: Allen & Unwin.
 

Hammond, Phillip E. 1964. Sociologists at Work. New York: Basic.
 

Harding, Sandra G. 1986. The Science Question in Feminism. Ithaca: Cornell University Press.
 

Harding, Sandra G. 1991. Whose Science? Whose Knowledge?: Thinking from Women's Lives. Ithaca: Cornell University Press.

Harris, Daniel C. 1987. Quantitative Chemical Analysis. New York: Freeman.
 

Harris, Marvin. 1979. Cultural Materialism: The Struggle for a Science of Culture. New York: Vintage Books.
 

Harris, Marvin and Eric B. Ross. 1987. Death, Sex, and Fertility: Population Regulation in Preindustrial and Developing Societies. New York: Columbia University Press.
 

Harriss, Barbara and Elizabeth Watson. 1987. "The sex ratio in South Asia. " Pp. 85-115 in Momsen, Janet H. and Janet G. Townsend, editors. Geography of Gender in the Third World. Albany: State University of New York Press.
 

James, Susan. 1984. The Content of Social Explanation. New York: Cambridge University Press.
 

Jencks, Christopher et al. 1979. Who Gets Ahead? New York: Basic Books.
 

Koestler, Arthur. 1964. The Act of Creation. New York: Macmillan.
 

Lamont, Michele. 1987. "How to become a dominant French philosopher: the case of Jacques Derrida." American Journal of Sociology 93:584-622.
 

Lazarsfeld, Paul F. 1958. "Evidence and inference in social research." Daedalus 87:99-130.
 

Lemert, Charles. Fall 1991. "The end of ideology, really." Sociological Theory 9:164-72.
 

Lenski, Gerhard, Jean Lenski, and Patrick Nolan. 1991. Human Societies: An Introduction to Macrosociology. New York: McGraw-Hill.
 

Levine, Donald N. 1985. The Flight From Ambiguity. Chicago: University of Chicago Press.
 

Lipset, Seymour M. 1979. "Predicting the future of post-industrial society." Pp. 1-35 in Seymour M. Lipset (ed.), The Third Century: America as a Post-Industrial Society. Stanford: Hoover Institution Press.
 

Lofland, John and Lyn H. 1984. Analyzing Social Settings. Belmont, CA: Wadsworth.
 

Longino, Helen and Ruth Doell. "Body, bias, and behavior: A comparative analysis of reasoning in two areas of biological science." Pp. 165-86 in Harding, Sandra and Jean F. O'Barr, editors. 1987. Sex and Scientific Inquiry. Chicago: University of Chicago Press.
 

Lukes, Steven. 1992. "The rhetoric of thick consensus." Contemporary Sociology 21:425-26.
 

Lundberg, George A. 1947. Can Science Save Us? New York: Longmans, Green.
 

Mackie, Marlene. "Sexism in sociological research." Pp. 1-23 in Tomm, Winnifred and Gordon Hamilton, editors. 1988. Gender Bias in Scholarship. Waterloo: Laurier University Press.
 

Marlowe, Lara. February 22, 1993. "Deadly science." Time 141:64.
 

McCloskey, Donald N. 1985. The Rhetoric of Economics. Madison: University of Wisconsin Press.
 

McKeown, Thomas. 1988. The Origins of Human Disease. New York: Basil Blackwell.
 

Meites, Joseph. 1977. "The 1977 Nobel prize in physiology or medicine." Science 198:594-96.
 

Merton, Robert K. 1957. Social Theory and Social Structure. Glencoe: The Free Press.
 

Michotte, A. 1963. The Perception of Causality. New York: Basic Books.
 

Moore, Barrington Jr. 1973. Reflections on the Causes of Human Misery and Upon Certain Proposals to Eliminate Them. Boston: Beacon.

Morgenthau, Hans J. 1958. Politics Among Nations. New York: Knopf.
 

Morrison, Denton E. and Ramon E. Henkel, editors. 1970. The Significance Test Controversy: A Reader. Chicago, Aldine.
 

Novick, Peter. 1988. That Noble Dream. New York: Cambridge University Press.
 

O'Hare, David and Stanley Roscoe. 1990. Flightdeck Performance: The Human Factor. Ames: Iowa State University Press.
 

Okin, Susan M. 1979. Women in Western Political Thought. Princeton: Princeton University Press.
 

Petersen, William. 1979. Malthus. Cambridge: Harvard University Press.
 

Porter, Theodore M. The Rise of Statistical Thinking: 1820-1900. 1986. Princeton University Press.
 

Rehberg, Karl-Siegbert. 1985. "'Anti-sociology': a conservative view on social science." History of Sociology 5:45-60.
 

Rorty, Richard. 1991. "Philosophy as a kind of writing: An essay on Derrida." Pp. 222-41 in Scott Lash, editor, Post-Structuralist and Post-Modernist Sociology. Edward Elgar Pub. Co.: Brookfield, VT.
 

Rosaldo, Renato. 1986. "From the door of his tent: the fieldworker and the inquisitor." Pp. 77-97 in James Clifford and George E. Marcus (editors), Writing Culture. Berkeley: University of California Press.
 

Rosaldo, Renato. 1989. Culture and Truth: The Remaking of Social Analysis. Boston: Beacon Press.
 

Roszak, Theodore. 1969. The Making of a Counter Culture. New York: Anchor Books.
 

Roth, Philip. 1971. Our Gang. New York: Random House.
 

Runyon, Richard P. and Audrey Haber. 1988. Fundamentals of Behavioral Statistics. New York: Random House.
 

Seidman, Steven. Fall 1991. "The end of sociological theory: The postmodern hope." Sociological Theory 9:131-46.
 

Selvin, Hanan C. 1957. "A critique of tests of significance in survey research." American Sociological Review 22:519-27.
 

Shelley, Louise. "The rebuilding of sociology in the Soviet Union." Footnotes (American Sociological Association Newsletter), November 1989, page 3.
 

Sica, Alan. 1988. Weber, Irrationality, and Social Order. Berkeley: University of California Press.
 

Smith Tom W. and Ruth Fujimoto. 1986. Annotated Bibliography of Papers Using the General Social Surveys. Chicago: National Opinion Research Center.
 

Stacey, Judith and Barrie Thorne. 1985. "The missing feminist revolution in sociology." Social Problems 32:301-16.
 

Stanfield, John H. 1987. Research in Social Policy: Historical and Contemporary Perspectives. Greenwich, CT: JAI Press.
 

Staples, Robert. 1976. Introduction to Black Sociology. New York: McGraw-Hill.
 

Stehr, Nico. 1986. "Sociology: the crisis of success." Contemporary Sociology 15: 365-68.
 

Tenbruck, Friedrich H. 1984. Die unbewaltigten Sozialwissenschaften oder die Abschaffung des Menschen. Graz: Styria.
 

Turner, Jonathan H. 1986. The Structure of Sociological Theory. Chicago: Dorsey.
 

Turner, Jonathan H. 1991. The Structure of Sociological Theory. Belmont, CA: Wadsworth.
 

Tylor, Edward B. 1898. Anthropology. New York: D. Appleton.
 

von Baeyer, Hans C. 1992. Taming the Atom: The Emergence of the Visible Microworld. New York: Random House.
 

Walster, G. William and T. Anne Cleary. 1970. "A proposal for a new editorial policy in the social sciences." The American Statistician 24:16-19.
 

Watson, James D. 1968. The Double Helix. New York: New American Library.
 

Weisskopf, Susan. 1978. Review essay. Harvard Educational Review 48:269-78.
 

Whicker, Marcia L. and Lee Sigelman. 1991. Computer Simulation Applications. Newbury Park: Sage.
 

White, Hayden. 1989. "'Figuring the nature of the times deceased': Literary theory and historical writing." Pp. 19-43 in Cohen, Ralph, editor. 1989. The Future of Literary Theory. New York: Routledge.
 

Whyte, William F. 1986. "On the uses of social science research." American Sociological Review 51: 555-63.
 

Winkler, Karen J. 1985 (June 26). "Questioning the science in social science, scholars signal a 'turn to interpretation'." The Chronicle of Higher Education: 5-6.
 

APPENDIX (Chapter 1)
 

=================================
 

An Adaptation Process Obscured by a Spurious Non-Relationship
 

Y ADAPT by X PROBLEM

X

Count I

Row Pct ILOW HIGH

Col Pct I Row

I -1 I 1 I Total

Y --------+--------+--------+

-1 I 74 I 79 I 154

LOW I 48.3 I 51.7 I 46.0

I 44.2 I 47.9 I

+--------+--------+

1 I 94 I 86 I 180

HIGH I 52.1 I 47.9 I 54.0

I 55.8 I 52.1 I

+--------+--------+

Column 168 166 334

Total 50.3 49.7 100.0
 

Statistic Value

-------------------- ---------
 

Phi .03708

Cramer's V .03708

Contingency Coefficient .03705
 
 
 
 
 
 
 
 
 
 
 

An Adaptation Process Obscured by a Spurious Non-Relationship (cont'd)
 

Y ADAPT by X PROBLEM

Controlling for..

Z LAG Value = 1 X TO Y

X

Count I

Row Pct ILOW HIGH

Col Pct I Row

I -1 I 1 I Total

Y --------+--------+--------+

-1 I 72 I 13 I 84

LOW I 85.1 I 14.9 I 46.1

I 74.2 I 14.6 I

+--------+--------+

1 I 25 I 74 I 99

HIGH I 25.3 I 74.7 I 53.9

I 25.8 I 85.4 I

+--------+--------+

Column 97 86 183

Total 52.9 47.1 100.0
 

Statistic Value

-------------------- ---------
 

Phi .59661

Cramer's V .59661

Contingency Coefficient .51235
 
 
 
 
 
 
 
 
 

An Adaptation Process Obscured by a Spurious Non-Relationship (cont'd)
 

Y ADAPT by X PROBLEM

Controlling for..

Z LAG Value = 2 Y TO X

X

Count I

Row Pct ILOW HIGH

Col Pct I Row

I -1 I 1 I Total

Y --------+--------+--------+

-1 I 3 I 67 I 69

LOW I 3.6 I 96.4 I 45.9

I 3.5 I 84.0 I

+--------+--------+

1 I 69 I 13 I 82

HIGH I 84.4 I 15.6 I 54.1

I 96.5 I 16.0 I

+--------+--------+

Column 71 80 151

Total 47.3 52.7 100.0
 

Statistic Value

-------------------- ---------
 

Phi .80596

Cramer's V .80596

Contingency Coefficient .62752
 

Chapter 2

MICHEL FOUCAULT, MACHINES WHO THINK,

AND THE HUMAN SCIENCES
 
 
 
 
 
 
 
 
 
 
 

The major purpose of this chapter is to have a little fun. Foucault and Foucaultians have challenged the social science disciplines in fascinating ways--some good, some bad, some ugly. This tripartite taxonomy will not satisfy everybody; I suspect, however, that Foucault himself would have liked it and that, in the end, there is little difference between Foucault and Comte.
 

(1) "Man the machine--man the impersonal engine"

(Twain, 1917:5)
 

i
 

During the last few years I've taught a course on the social and psychological impact of computers. Several of the students enrolled in this course last spring were computer science majors, and they were surprisingly skeptical about artificial intelligence, about "thinking machines"; a few had read the Dreyfus brothers' critique of machine intelligence (Dreyfus and Dreyfus, 1986). At the time I had not yet discovered that one of the brothers, Hubert, is among those scholars who have introduced Michel Foucault and his philosophy to English-speaking audiences. I would be most interested in pursuing the hypothesis that scholars friendly toward the Parisian brand of structuralism--or post-structuralism, for ideas move fast in the City of Light--tend to be unfriendly toward machines who may soon have the temerity to affirm, with Descartes, cogito, ergo sum. Still, it is a little disappointing that Hubert Dreyfus's book on Foucault does not contain any reference to machine intelligence, because one who understands the basics of machine intelligence has a good prospect of understanding what is wrong (and right) with Foucault. But this, of course, was not Dreyfus' mission.

As my students and I began arguing more vehemently, I made the slightly platitudinous point that just as airplanes finally were not modelled on birds, so the thinking machine does not necessarily have to be modelled on human brains with all their foibles, let alone the human mind (Wolfe, 1991). The students immediately shot back the slightly less predictable claim that birds do not find it necessary to communicate with airplanes, while computers must communicate with humans--at least, so say we. My response, again somewhat predictable, was to introduce the famous Turing test: If, in a controlled experiment, a human being cannot determine whether he has another human being or a thinking machine on the other end of the line, then it has been established that the machine, presumably, has both brain and mind. Nowadays many chess programs readily pass this test even when there is a possibility that the randomly selected thinker on the other end is a human grand master. I suspect that people (and computers) who make a living by buying and selling stock index futures cannot be certain nowadays whether the competition consists primarily of women, men, or machines. Actually, the computers may have a better prospect than their masters of making an educated guess as to the nature of the recalcitrant being on the other end of a transaction.

At this moment, however, even though things were looking pretty good for my argument, the students were preparing to devastate me by pointing out that my brainless and mindless machine would never realize that the following two grammatical structures, although they have everything in common, are diametrically opposite in meaning:
 

"The chicken is ready to eat."
"The tiger is ready to eat."
 

Before they had an opportunity to toss off another classic conundrum--I expected "the pen is in the box" versus "the box is in the pen"--I delivered what I considered to be the only appropriate response: "Granted," I said, "it is not easy to wreck a nice beach."

Of course they fell for it, hook, line, sinker, body and soul, brain and mind, insisting that the one thing that sets them apart from the lower animals and from the lower forms of humanity and from the inanimate realm is that they have this unique, infinite, God-given capacity to wreck a nice beach, that there never has been nor will there ever be an ugly, disgusting, filthy, greasy, clanking machine that will wreck a nice beach as expeditiously as they have, way back to the days when they were very small children.

The next step in my argument--this is a very mechanical process for a human being, is it not?--was to point out that while it is true that airplanes, when they became successful, were not modelled on birds, it now turns out that birds actually have evolved several features that work to advantage for airplanes. I don't recall where I read it, but recent research on bird locomotion shows that, at least for some species, feathers located toward the wingtips have a way of twisting as the bird accelerates, so that the feathers function as propellers. This discovery leads me to believe that if human beings must communicate with computers, we might try to emulate some of their more benign characteristics. For starters: the immense capacity for clear communication now built into "user-friendly" computer systems.

I then insisted, as always, that my computer science curmudgeons read an excellent classic by Bolles (1972) dealing with one of the clearest and most computer-like forms of human communication, i.e., baby talk. They soon discover, among other things, that
 

Having noticed that telegraph operators are among those who like to say "da-di" a lot, I prepared the following communication--the clincher to my overall argument--and sent it forth not only to my CSc students but as a public announcement over the electronic mail networks that bind my university to the remote outposts of the Western world:
 

Within minutes I had many, many responses either from brains, minds, or machines--sadly, I flunked my own Turing test--and these respondents invariably said yes!, yes!, yes! One CSc major actually quoted James Joyce, "... and his heart was going like mad/ And yes I said yes I will Yes," having realized that Joyce in this instance was describing a fundamental human situation in which binary baby talk prevails. Because many respondents actually said nothing more than "y"--translated from "da-di-da-da," a point my newest baby makes every few minutes--I realized that binary baby talk can be very quick and efficient; it's currently the rage around those parts of campus that I visit, or at least hear about. One exception: a slightly bigoted respondent who said that he would never adulterate his own language, he would never lower himself by learning to speak in the style of computers. I told him that he would not have felt that way a century and a half ago when we first achieved a way of speaking in binary mode, by telegraph, to remotely located human beings. Why can we not make the same concession for computers? I tried to remember the lovely lines from the movie 2001 in which somebody makes the point that we should not discriminate against computers simply because they are made of sand; we, after all, are made of an even baser dust. I also suggested that (1) computers are extremely fast and (2) computers are extremely "dumb," and that for the latter we fault them; if, however, computers were brilliant but plodding, we would still fault them: We would probably call them, in the terminology of some of my students, "greasy grinds." This is a fixed prejudice: And in the eyes of the victim, as always, one sees oneself--this, at least, is the ironical interpretation of the present chapter.

This brings us to the fringes of a very large question: Why is it so easy to translate from English to Morse code and back to English without making mistakes? Steiner (1975:28) tells us that "when we read or hear any language-statement from the past, be it Leviticus or last year's best-seller, we translate," and that nothing is tougher than making a good translation. To me, the answer is that Morse code follows conventional rules in all its details, while most "language-statements" are loaded with subtlety and ambiguity.

Having provided a Morse-code illustration, I feel obliged to provide an illustration of the more complex sort of translation that Steiner has in mind. My example is Mark Twain's (1875:15-34) famous story about the jumping frogs of Calaveras County, California. Twain was intrigued by the French translation of his original story, and set about translating the French back into English in order to determine whether "iconoclasm can further go." One small segment of the original contains the denouement of the story:
 

When Twain brings these lines back from the French, they read as follows:
 

In a sense, Twain fails to make his point: The second version of this incident is indeed funny¹ when compared to the first, but it communicates the original story surprisingly well. Given that the original story is written in a difficult vernacular, given that a French reflexive verb is bound to sound strange when translated literally into English, and given that no serious translator would make such a mistake or write such a mistake into a program, this translation is not at all bad by standards only slightly below Steiner's. But this is nonetheless a tough translation problem. Not as tough as going from Omar the Tentmaker to Edward FitzGerald and then back again to Arabic (old or modern), which is said to be nearly impossible, but much tougher than the English-Morse-English example developed above. The difference arises because there are fewer conventional rules of translation in the Twain example than in the Morse example; in the extreme FitzGerald instance, we would have to proceed sans wine, sans song, sans singer, and--sans rules!²

Take an intermediate case: A piece of journalism originally written (I believe) in English, then translated into Spanish, and finally translated by me back into English so that I might ascertain whether iconoclasm has any limits, whether the original author and I seem to be saying the same things even after applying many subtle rules of translation. In the interest of reflexivity I'll translate a paragraph or two from an interview (McDowell, 1986:38-39) with the famous translator Gregory Rabassa (One Hundred Years of Solitude), so as to clinch my argument both in terms of what Rabassa says and in terms of what I imply in the process of translating his argument, which of course does not necessarily say what I think it says. We'll have to check the original, and find out.
 

So: We don't need the Gestalt before we set to work. This will be a relief to the average supercomputer. We continue:
 

"Tone," I'm afraid, gets messy chiefly because there is so little agreement on what it means. My guess is that it has a lot to do with "unity of diction" (Hall, 1973:123-42; Main and Seng, 1961:13-14), that it is therefore a very large and confusing issue, and that if we ignore it we are at risk of creating travesties such as "The Gettysburg Address in Eisenhowerese" (Hall, 1973:128-30). But, fear not: It is inconceivable that anybody will ever program a computer to think, write, and speak in the style of Dwight D. Eisenhower or, better or worse yet, the style of Casey Stengel, with whom Ike was not even ballpark for obscurantism. Rabassa (McDowell, 1986:39) seems to think that tone has a lot to do with having a "good ear," and this comes as welcome news to intelligent machines who write music. It is also encouraging that one of the world's mightiest translators does not believe that one man's translation is as good as another's--even though it is obvious to me that all the "bad" emic translations are conceived in liberty and created equal.

It is significant that Garcia Marquez himself, interviewed a few years ago in the book review magazine of the New York Times (Simons, 1988), claims that his many translators often send him lists of "... things they have doubts about," and that these lists tend to be the same no matter what the language. The doubts, then, are not an inherent feature of the translation process; rather, they are elements of the original text. If Garcia gives everybody the same reply, all the translations should be reasonable replications of a difficult literary genre.
 

In any case, here is the original English version of the Rabassa excerpt:
 

My own translation, two steps removed from the original and filtered through a knowledge of Spanish inferior to that of the better educated (programmed) machines, does not seem to have missed anything of importance. We learn, through a "content" of secondary significance, that a translator of immense skill does not need to perceive a work as a whole before he proceeds, and that he believes that a complex translation may be constrained by rules that permit only one result. If I ever try to write an AI expert system for Spanish-to-English translation of sociological literature, I will surely consult for days and weeks with Rabassa. The realm of "one result, one true story," of course, is the delightful realm of replication. And it is therefore, in spirit, the realm of science--and of the edifying technique known as mad-libbing, to be discussed below.
 

ii
 

This entire exercise has illustrated nothing more than what social psychologists have understood for many years as the rumor process (Allport and Postman, 1958; Bartlett, 1958), but with this important proviso: Translations distort, but some kinds of translations distort far more than others. The advocates and exemplars of Foucault's brand of structuralism, I believe, urge social scientists to devote most of their effort to the more esoteric problems of translation such as that represented by the Twain and FitzGerald examples; this is what the emic perspective is all about. But the central task of a social scientist, on the contrary, is to develop a language that translates more in the manner of the English-Morse-English example or, realistically, the Rabassa example, so that when we speak of something called "Pygmalion effect" (Rosenthal and Jacobson, 1968) or something called a "specific deterrent effect of arrest" (Sherman and Berk, 1984; Sherman et al., 1992), there is little if any confusion as to what these phrases mean and there is little if any confusion as to how social scientists might try to replicate the alleged effects. When we have replications that come close to the English-Morse-English example, we are achieving science. When we have a corpus of knowledge that a good AI program could translate, say, from English to Russian and back again without significant distortion, we are achieving science.

Foucault's interpreters can hardly help us with this task, because they insist that we concentrate on Twain-Smiley and his unique, emic perceptions. As one probes more deeply into the literature by and about Foucault, one realizes that the best way to understand his work is to read books that explain the interpretation of poetry. Main and Seng (1961:19-20), for instance, define poetic license as
 

Poetic license has its functions, but it should not tempt us toward the popular notion that poetry never demands precision: Main and Seng (1961:7-10) speak of "the indispensable dictionary" as merely one aspect of this precision. It appears, however, that a special sort of license has been granted to Monsieur Foucault, a sort of license that never can be given a large role in scientific analysis and discourse. I have trouble developing enthusiasm for an author about whom admirers say this (Lemert and Gillan, 1982:ix): "He has been accused of willful obscurity. But this complaint runs counter to the fact that, even when one cannot be sure what he is saying, it is evident that he is saying it well."³ This reverential attitude toward rhetoric goes to the essence of my complaint against Foucault and his supporters. Well-spoken obscurantism is an oxymoron. In scientific discourse, if you cannot be sure what somebody is saying, you cannot be sure that he is saying it at all well, and either you or the author or both must learn the appropriate language. One who abandons this principle is abandoning the idea of science--social or otherwise. For those of us who accept the feasibility of a science of society, to be enticed by Foucault's (and/or his interpreters') brand of obscurantism is to be enticed by a quick trip to the edge of a flat earth.

Here is an instance: I am not sure what Foucault is saying, and I am convinced that he is not saying it at all well (Foucault, 1978:92-3):
 

To my mind these endless abstractions, at best, are the grindstones of the garrulous; at worst, they are the word salads of the mentally deranged. And it is not simply that I have lifted this excerpt out of context: Nothing that precedes or follows it will clarify what it is trying to accomplish. It is dead-level abstraction of the sort that general semanticists used to deplore, and not only do I claim that science does not work this way: I claim that the human mind does not work this way. If Papert (Bass, 1990) is correct in saying that one of the best ways to get children to learn algebra is to allow them to fondle differential gears, then the problem with Foucault is that a good greasy gearbox is almost never in evidence. In this excerpt, Foucault clarifies the concept of social evolution in the same way that Professor A. Parker Nevin once clarified the concept of "justice" in a famous nonsense speech that was almost always treated, by the frumiously gullible, as if it contained important information (Starmer, 1955).

Apropos of children and their games, the game called "mad libs" provides a sort of Turing test for the presence of Foucaultian rhetorical spin. In this game, new words are substituted randomly for those found in the original text--nouns for nouns, verbs for verbs, etc.--and the results are often hilarious. In the above excerpt, I played a trick: I substituted the phrase "social evolution" for every instance in which Foucault, in the original, used the word "power," and the mad lib sounds at least as plausible as the original. When one does not discern that the game is being played, one has failed a variant of the Turing test and one is almost certainly in the presence of Foucaultian rhetoric. In another place, where Foucault (1984) discusses the relationship between writing and death, I achieved far more insight (and pleasure) by mad libbing "life" in place of "death" and by making a few additional alterations as appropriate. Similar exercises, involving psychoanalytic writings, appear in Faia (1977).⁴

At this moment, I feel compelled to make a confession: Recently I found myself mad-libbing my own work. Using the SPSS statistical software for loglinear analysis, I had replicated Swafford's (1980) estimates, based on a national survey, of ways in which respondents' attitudes about equal employment opportunity were influenced by education, by region (south versus non-south), by the passage of time from 1946 to 1963, and by interaction among these variables, e.g., the possibility that the effects of region may have been different in 1963 than in 1946, etc. I then carried out another analysis, based on the National Survey of Family Growth (NSFG), of ways in which the experience of unwanted pregnancies--and, presumably, the resulting demand for abortion--were influenced by education, by use of contraception in general, by use of condoms for "safe sex" purposes, and by interaction among these variables, e.g., the possibility that the effects of "safe sex" practices on abortion demand may have been different among poorly educated respondents than among others, etc.

Not surprisingly, the mad-libs essentially wrote themselves. It is therefore crucial to my argument that I state my reasons for believing that my self-reflexive mad-libs are superior to my Foucaultian mad-libs, as a set of scientific findings. Here are the major reasons:
 

Finally, a dialectical insight: The irony of Foucault(ians) is that, while obscurantism cannot properly be translated into English (or any other language) by human intelligence or machine intelligence, it would be easy for a computer to write mad libs based on the more esoteric Foucaultian texts. Such a computer could easily masquerade as an exemplar of artificial intelligence; for one thing, it could probably meet the Turing criterion, even if those on the other end of the line were Foucaultians. If the machine tried to fake a loglinear analysis of the NSFG, a good statistician would catch it. I have 5,000 francs supporting this contention.
 

(2) Good, bad, or ugly?
 

--Michel Foucault
 

In another part of their book Lemert and Gillan (1982:9) quote Foucault on the topic of Foucault's quotes: "... I quote Marx without saying I am, without quotation marks, and because people are unable to recognize Marx's texts I am considered to be someone who doesn't quote Marx." If the ordinary rules of intellectual inquiry were not already suspended, this confession would open up a strong possibility of (1) a pervasive carelessness about documentation or (2) plagiarism. Yet Lemert and Gillan tell us not to worry, that while it is "... futile to talk of his sources" (1982:10), "Foucault could not more closely identify himself with Marx. Nor could he more clearly and perversely distance himself from contemporary French Marxism. This is typical of Foucault's method. Neither Marxist nor anti-Marxist, yet both" (1982:9). Now, I generally travel pretty far along this pike: I appreciate Korzybski's (1933) and Ogden and Richards' (1930) ancient warnings about relying too heavily on the Aristotelian notion that an entity must be "either A or not-A"--Foucault, in more ways than one, brings back the basic imagery of general semantics. Further, I think I fathom the subtle ways in which Saddam Hussein sees victory in defeat. But, eventually, the dialectical pike runs into a dead end: There is no excuse for failing to explain the sense in which one is both Marxist and anti-Marxist, and there is no excuse for incorporating into one's own work the scholarship of others without acknowledgement.

It is outrageous that at least one publisher encouraged Foucault (1970:viii) to continue his bad habits regarding documentation, by promising that the editorial staff would try to do the essential work for him--such are the powers of the marketplace. And in case editorial staffs cannot deliver, Lemert and Gillan (1982:32) suggest that the "impatient reader" is expected to take on the responsibility herself. Foucault and friends may believe that social norms both exist and do not exist, but I for one do not believe that Foucault should have been exempted from the fluffy norm that asks us to acknowledge our intellectual debts; furthermore, science is an elaborate, time-consuming communication process, and most of us do not have time for Easter-egg hunts.

This inexplicable suspension of the rules causes endless difficulty (Eribon, 1991:141-42). When Lemert and Gillan (1982:110) tell us that the theoretical kin of Foucault include Poulantzas, Miliband, Giddens, Parkin, Gouldner, Offe, Habermas, Althusser, O'Connor, Wallerstein, Skocpol, et al., but that Foucault "...gives little evidence of having read these others," we must ask: What, in such an instance, would constitute evidence? What would constitute theoretical kinship? And when these same authors (1982:121) say that Foucault "... is aware, as is practically everyone else, of the futility of positivism," we must ask again: What constitutes the evidence? Who is "everyone else," and has anybody actually read a reasonable sample of their work, or at least interviewed them? Picked up their vibes? Turned up their eggs?

I wish Foucault were still alive, so that he could play Mark Twain in ordre inverse. I would be especially keen on seeing how the following excerpt from Foucault,⁵ quoted by Dreyfus and Rabinow (1982:31), would fare if subjected to a three- or four-step rumor process involving word-of-mouth among social philosophers, and then translated back into French by the original author:
 

I'd love to ask my anti-positivist colleagues--i.e., "practically everyone"--to participate in a multiple-translation process, but they always seem to know exactly what I'm up to despite their difficulty in reading one another. If the experiment were actually conducted, Foucault, I believe, would soon exclaim, "je te l'avais bien dit!"

To be sure, Foucault had serious misgivings about the human sciences, and his many disciples feel the same way. Most of his arguments, however, turn mainly on tired issues already addressed: Is it appropriate for the human sciences to emulate the natural sciences? Is it possible for man to study himself reflexively without being caught up in those infinite convolutions created by the fact--or the belief--that behind every representation there is nothing more than another representation? Where do we find the divine spark in the human sciences? I've proposed to answer these questions, respectively, by saying yes, yes, and I don't know. These answers should suffice, and I hate to dwell further; however, Foucault and his interpreters leave me no choice.

Cousins and Hussain (1984:63) claim, with Foucault (1970:357), that "obviously the social sciences are and always have been riven with methodological controversy. This is largely because of their parasitic relation to other forms of knowledge from which there has been a perpetual import-led boom of conceptual activity." But this remark is not valid, because we happen to be dealing with an ex-im business: When Darwin and Wallace took ideas from the Reverend Malthus and from Adam Smith, was this not the greatest act of "parasitism" in the history of the sciences?⁶ When natural scientists use the logic of the life table and the stable population model, are they not stealing from Edmund Halley, John Graunt, and Alfred Lotka--all of whom were at least part-time social scientists? When the great Lotka moved from biology to demography, was he parasitic on himself? Can we borrow a biological term for reflexive parasitism, or should we get ready to lend another neologism to the biologists? Science is the process of learning, "no holds barred," in a way that permits public verification; if these criteria are met, it does not matter a whit who borrows from whom, and intellectual property remains communal. (It is, however, a fine courtesy to reveal one's sources, and perhaps Cousins and Hussain are merely projecting a latent sense of guilt for the moral turpitude of the master.)

Cousins and Hussain (1984:29) tell us that Foucault's concept of meaning is that there are representations of "things," but that these things are themselves merely representations of other things, and so forth ad infinitum. I don't have any problem with this notion: It merely reminds us that we are wise to arrive at operational definitions of "things." In my research, for instance, "unintentional pregnancy" is merely a representation of other representations that occur during an NSFG interview, representations in which a given woman tries to remember and reconstruct her sexual, marital, and fertility history. This does not vitiate the demonstrable fact that unintentional pregnancies have a large impact on the 1.5 million abortions that occur in this nation each year,⁷ and that these abortions are a fundamental and formidable Durkheimian chose.

When we begin to appreciate that representations represent other representations onward to infinity, we soon enter the realm of mysticism. When I try to enter this realm in my capacity as a social scientist, I discover immediately how James Thurber must have felt when he was informed by a self-improvement manual that his major problem in life was his own large ego: He was forever using the first-person singular--it was always I, me, mine, to me, for me, myself, etc. Thurber tried for weeks to break out of this ugly syndrome, and finally declared in obvious despair: "Tried it. Am tied in knots. No can do." When I read Foucault on sexuality, I love the mysticism (1970:25-26) about how we are "discontinuous beings," about the sacred character of eroticism, about the meaning of the sperm cell and the ovum. But whenever I read these particular passages I quickly end up at Thurber's dead end, and I cannot help but think of the old story about how the French social welfare system is the world's most advanced, looking after the citizenry not merely from cradle to grave, not merely from womb to tomb, not merely from sperm to worm, but--with tender loving care and all deliberate speed--from erection to resurrection. And that's enough mystery for me; beyond that, no can do.
 

(3) Mitigating circumstances
 

Despite his alleged anti-positivism, Foucault consistently spins arguments in a way that aids and abets those who remain strongly committed to the human sciences. He wrote excellent sociological treatises on prisons, mental institutions, and the history of sexuality; in these and other works, he made important contributions to major theoretical traditions having to do with "labelling" as a self-fulfilling prophecy, deterrence as a means of social control (Cousins and Hussain, 1984:182), the nature of social norms (Dreyfus and Rabinow, 1982:81), and structure-functional analysis (Foucault, 1970:168-74; Cousins and Hussain, 1984:196). He had deep insights into the nature, development, and uses of scientific taxonomies (Foucault, 1970). Rather than rejecting an important role for mathematics in the human sciences, he clarified that role in a way that would be warmly endorsed by contemporary positivism (Cousins and Hussain, 1982:58-59; Foucault, 1970:349). He warned us, as we often have warned ourselves, against the temptation to explain major historical events with reference to phenomena such as "Luther's anal retentiveness" (Cousins and Hussain, 1982:64). He introduced us (1970: Chapter 1) to Borges' metaphysical conceits in a way that encourages us to use what C. Wright Mills called the sociological imagination. He traced the evolution from anatomy to physiology (1970:226-32,266), the evolution from normative economics to economics as the study of institutional dynamics, and the development of modern linguistics in a way that should inspire all human scientists who wish to see their own disciplines achieve the successes of biology, economics, and linguistics.

Those who share Collins' (1984) worries (and conclusions) about the dialectic of statistics versus words will find little encouragement in Foucault--although his anti-science interpreters appear to be capable of finding anything, anywhere, anytime. In their treatise on Foucault's major works, Cousins and Hussain (1982:58-59) discuss the association of positivism with mathematization, mentioning Foucault's belief that
 

I'm a statistician, and this is good enough for me. But Foucault himself (1970:349) says it even better:
 

Again: Good enough for me, good enough for Weber, and good enough for vulcanology. If these remarks can be taken to mean that we no longer have to read Lazarsfeld, then they can be taken also to mean that we no longer have to read Poulantzas, Miliband, Giddens, Parkin, Gouldner, Offe, Habermas, Althusser, O'Connor, Wallerstein, Skocpol, et al. In short, we institutionalize Comte's cerebral hygiene.

Even when Foucault seems to criticize quantitative social science, he is encouraging. For instance (1970:351):
 

Again I say, Precisely! Not long ago I supervised a master's thesis (Flippo, 1987), based on the 1982-83 NSFG, in which the major finding was a pair of beta weights: One is .34, the other is .31. These numbers are clear; they are untroubled, transparent; they have a scientific style, form, and justification. The justification is that they remind me and my colleagues about a vast number of things: about the tremendous care that must be taken, throughout the research enterprise from sample design to coding and analyzing data, in order to isolate a pair of beta weights; about the venerable hypothesis and related theories that hold that, for many plausible reasons having to do largely with an enhanced ability to defer gratification, the sexual conduct of high-status minority women in the United States is different from that of high-status majority women--a hypothesis that our beta weights have led us to question. The beta weights are nothing more than the Torah: They stand alone in their simple beauty, in their elegance, and they sustain an infinitude of commentary and insight.

Sometimes, when I attack Foucault, I realize that I am going for the capillaries; the arteries can look pretty good. For instance (1970:353):
 

For those who subscribe to the "cultural materialist" perspective (Harris, 1979), it will surely come as a surprise that biology (e.g., fertility, mortality, migration), economics, and linguistics are not at the heart of the social science endeavor. To Foucault, they are not fundamental; what is fundamental is man's capacity for knowledge. But there are many excellent social scientists who believe that (1) man's perceptions and beliefs are not the central focus of the social sciences and (2) insofar as they are a topic of interest, they may be studied in the same way in which we study organisms of the phylum Platyhelminthes. It may strike Foucault as the "peculiar configuration" of the human sciences that it is possible to create "human sciences of human sciences" (1970:354-55), the psychology of psychology, the sociology of sociology, etc., but the sociology of sociology is no more peculiar than the anthropology of chemistry, the psychology of physics, the chemistry of biology, or the physiology of platyhelminthic worms. Foucault (1970:361) drifts toward the ecological point of view when he concedes that functions can be performed, conflicts can be developed, and signification can "... impose its intelligibility, without passing through the stage of explicit consciousness ... And now, is it not necessary to recognize that the peculiar property of the norm in relation to the function it determines, of the rule in relation to the conflict it regulates, of the system in relation to the signification it makes possible, is precisely that of not being given to consciousness?" If I read this passage correctly--and Foucaultians are invited to read it as they wish--it comes down strongly on the side of cultural materialism, of human ecology. To me, this enhances the scientific value of Foucault's work.

Dreyfus and Rabinow (1982:66) have a similar impact when they try "to make Foucault's structuralist view plausible" by providing the example of the university. (Never mind Foucault's claim [1970:xiv] that "half-witted 'commentators' persist in labelling me a 'structuralist'" and that the correct label is something that "I have been unable to get ... into their tiny minds ..." It is possible, I suppose, that Foucault is both a structuralist and an anti-structuralist.) The functioning of universities, say Dreyfus and Rabinow, depends on economic, political, familial, institutional, architectural, and pedagogical factors, but these factors coalesce into a university only when we have "the idea of the university." If we ask about the relative importance of the factors and the idea, we are digging into terrain that lies considerably more than a mile north of the north pole--the question is unanswerable. Dreyfus and Rabinow seem to sense this, but they soon revert to mysticism when they say that the concept, i.e. the "idea," "is itself a 'secondary relation' conditioned by something else." But this "final unifying factor" cannot be described in objective or mentalistic terms; it is a "way of talking" that is taken seriously in the domains of higher education. The rules of talking "... ultimately 'effect' or 'establish' university life as we know it" (1982:66). Aside from the obscurantism regarding talk that is not amenable to objective or mentalistic understanding, there is nothing new about this formulation; it is mainstream social science.

In large measure Foucault's The Order of Things is about taxonomies, how they are constructed, and how they evolve. Apparently, the book was inspired by Borges:
 

Foucault (1970:xv) is delighted by the "...wonderment of this taxonomy ..." In a few moments, he asks "... what is the ground on which we are able to establish the validity ..." of a classification (1970:xix)?--a question that he does not answer satisfactorily. In fact he makes big mistakes, like the suggestion (1970:147,268) that the transition from anatomy to physiology destroys the prospect of a workable scientific taxonomy. He should have asked a contemporary Chinese biologist how it is that he knows what the word "Platyhelminthes" means. When Foucault tells us (1970:27) that there may be a "sympathy" (as the translator says) between aconite and our eyes he is offering a metaphysical conceit worthy of John Donne, who once wrote a poem in which the mingling of a man's blood with a woman's blood in the tiny body of an insect becomes the tenuous basis for a seduction, a sort of flea-bitten fait accompli. But in calling attention to metaphysical conceits Foucault is not telling scientists anything that we do not already know, especially if we have read Arthur Koestler (1964) or have become familiar with the recent software development known as "hypertext."
 

(4) Oodles of Boodles
 

Foucault begins The Order of Things (1970) with a detailed description of Las meninas, a painting by Velazquez. This is an extraordinary document, the best of the genre; if I ever have an opportunity to view this painting I'll be able to gaze upon it with a new capacity for vision. But when Foucault claims that this painting demonstrates for us an epistemology that encompasses scientific developments of the era of Velazquez, I have problems. To those who disagree with me, I offer this challenge: Hire a dozen French art critics, and have each of them write a 6,000 word essay--the length of Foucault's chapter--on Las meninas. Translate these works into English, and compare them. Then, support a dozen different research teams in replications of the domestic violence-police intervention experiment conducted by Sherman and Berk (1984), and compare the results of these replications.⁸ Finally, compare the two processes, thereby obtaining what will surely look like representations of many other representations, and you will see the essential differences between non-science and science. Regarding the second process, one would immediately realize that the scientists are talking about the same sort of object; one could readily design another replication. In the Las meninas example, if the name of the artist and the painting were removed one would hardly be able to identify the object of discussion, and there would surely be no conclusion that one could try to replicate. Show me somebody who denies my claim and I'll show you somebody who believes that art criticism is a science.

Foucault was charmed by the idea that, because the knower and the known are the same (1970:366-67), a science of the social is difficult if not impossible. "Man," he says (1970:318), "in the analytic of finitude, is a strange empirico-transcendental doublet, since he is a being such that knowledge will be attained in him of what renders all knowledge possible." This does not vitiate the fact that there is no difference between an American and a Chinese scientist working in the realm of helminthology, or an American and a Chinese conducting an experiment to ascertain whether human-computer interaction is facilitated by an efficient keyboard, a "mouse," a "joystick," or a light pen. And this is not a trivial problem: The joystick and its ilk clearly have a fundamental impact on human wellbeing; in fact, they may exceed helminthic worms in this regard. There may be a rationale underlying the reflexivity argument, but I've searched in vain. Those who accept it could perhaps convince me if they were to explain how it is that the knowledge created by vulcanology is superior to the knowledge created by the United States Census Bureau, an example I select primarily because it had a strong appeal to Lundberg (1947:21,37).

I like to point out to my CSc students that certain tasks are "two-sip" tasks: Two sips of one of my infamous macromartinis--I happen to be a Boodles man--and the task cannot be performed well if at all. Writing an AI program in Prolog while operating in a multi-tasking computer environment, or landing an airplane while announcing one's position and maneuvers to controllers and to other aircraft, and simultaneously worrying about a funny engine vibration, are very good examples.⁹ In the case of Foucault, however, I usually cannot begin to understand him until I have consumed a triple Boodles. This is one of the chief reasons why I enjoy Foucault so much; he helps me to see that while science is sober, non-science and anti-science are informed by a lovely little buzz.
 

(5) Why Foucault needed Lindroth, and why Lindroth needed Foucault
 

The major argument of this chapter is captured perfectly in a story told by one of Foucault's biographers (Eribon, 1991:83-84). As a young scholar at the University of Uppsala, Foucault explored the possibility of taking an advanced degree under the supervision of Professor Stirn Lindroth. He asked Lindroth to read an early draft of his work on the history of mental illness and its treatment. Lindroth, a "... dyed-in-the-wool positivist and not particularly open to grand speculation ...," was "frightened" by the pages he examined; he wrote to Foucault and expressed his extremely unfavorable impressions. Foucault replied, somewhat defensively it would seem, as follows:
 

These scholars had very different intentions, and I hope they continued talking to one another. In any case, the problem did not go away (Eribon, 1991:217).
 

NOTES (Chapter 2)
 

(1) In general, translating humor is at least as difficult as translating Omar Khayyam, and machines may never do much more than help humorists discover good jokes--here, I think, machines will have an important role. Humor, however, not only makes exorbitant demands on machines; it also operates at the limits of human capability. During the December, 1987 summit meeting between Reagan and Gorbachev, Reagan quipped that he had made a disarmament proposal that was "disarmingly simple." Gorbachev, as I recall, gave a courtesy laugh, but I suspect that it required at least fifteen minutes for a good translator to explain to him the significance of Reagan's remark. And if Reagan could have seen the resulting elaborate translation, he probably would not have agreed with it or even recognized it. If one were to make a detailed study of translation processes at summit conferences, one would come away with a profound understanding of what the field of social psychology is all about.
 

(2) A critic of an early draft of this chapter claims that my argument about Morse code is
 

Whenever I hear that I'm dead wrong, I come alive with powerful counter-arguments; this is part of my aviation training. I think that it is a mistake to assume a sharp qualitative difference between transliteration and translation: These processes vary by degree, and even my critic makes a large concession regarding clicks, gutturals, etc.

In the languages of cultures with phonetic alphabets there is a scale of complexity that ranges, typically, from the elements of alphabet characters--such as dots within matrices or Morse dot-dash sequences--through the characters themselves, words, sentences, paragraphs, paragraphs organized into sections, sections organized into chapters, entire volumes, multi-volume works, etc. At each level there is meaning, although the complexity of these meanings increases exponentially at higher levels. When considered across languages, letters of the alphabet cannot simply be transliterated, because they may convey relatively complex meanings, such as an "F" in the grading system, the letter "X" as a way of indicating negation, or the letter "Z" or "Zulu" in reference to clock time at Greenwich, England. Airplanes have crashed because pilots failed to listen to three-letter Morse identifiers of navigational radios. We are all taught to "copy" letters, words, phrases (e.g., the standard platitudes), and many sentences, although I doubt that anybody has ever been accused of plagiarism without copying at least a paragraph or two: The smaller units are simple, and cannot sustain a claim of unique authorship. It is perhaps not coincidental that the most impressive computer writings that I've ever come across used short paragraphs as the most complex unit.

The point of my Morse example is that it involves a relatively simple and unambiguous way of communicating, that scientists try to accomplish comparable simplicity and clarity at all levels of meaning, and that we probably have (and should have) a bias toward simplicity of meaning structures: Single-sentence hypotheses rather than full paragraphs, articles or monographs rather than books, etc.
 

(3) Apparently, Derrida had the same set of bad habits. Lamont (1987:591) says that "Derrida, like other French intellectuals, is renowned for writing in a sophisticated and somewhat obscure style ..." Later in the same paper (1987:604-5) she says that
 

If these rude behaviors can inhere in the social facts of an entire national university system, this is all the more reason for refusing to emulate them. The recurring references to sophisticated obscurantism suggest the presence of an institutionalized oxymoron.
 

(4) About a week after I wrote the preceding paragraphs, I came across a paper by Ross (1990:262-65) that contains an extraordinary coincidence: Ross quotes Foucault's definition of power, discusses it for a page or two, and then presents a mad lib in which he substitutes technology for power. His results are at least as edifying as mine. Recognizing that Ross is generally friendly toward the Foucaultian style, I found that the scales fell from my eyes as I realized further that mad libbing may be a central feature of the Foucaultian intellectual repertoire. A grammar serves as one's airframe: One fills it with passengers and then sees whether it will fly! It usually does, up to a point.

Jon Stratton (1990:227) has also discovered flying. He quotes Foucault on power and sex, and then tells us that "for our purposes ... the implication ... can be best ascertained by substituting truth for sex in the above quotation." When I integrate Ross and Stratton, simultaneously substituting technology for power and truth for sex, I experience immediate detumescence although I still have a good time.

Mad libs are a delightful intellectual device. Substitutions generally should not be random within parts of speech, as in the children's game. A given noun, let's say, should consistently be substituted for another. A second non-random form of substitution involves reversals: Substitute antonyms. Note how the following translation of remarks by Rosaldo (1989:224) elucidates the original. First, Rosaldo:
 

And my mad lib, not motivated by the rage in grief:
 

(5) Almost any excerpt from Rubenstein (1988) would serve equally well.
 

(6) The biologist Ernst Mayr (1991:85-6), asking "how great was Darwin's debt to Malthus?" answers that it was very great indeed. Much was borrowed, including the idea that we live in a "... pessimistic world: there are ever-repeated catastrophes, an unending, fierce struggle for existence, yet the world essentially remains the same." This is ironic: As is usually true of those who cite Malthus, Mayr shows familiarity with the first edition of Malthus' essay on population but not with the truly scholarly editions of 1803 and later. In the more scholarly work, the pessimism disappears. I urge my colleagues in biology to take delivery on another sociological import. And if we can hold down the tariffs, I'd like to import state-of-the-arts computer graphics on human physiology.
 

(7) This number has dropped in recent years.
 

(8) Several replications of this experiment appear in the American Sociological Review, volume 57, October, 1992.
 

(9) Dear FAA: I've only tried the first of these experiments.
 

REFERENCES (Chapter 2)
 

Allport, Gordon W. and Leo J. Postman. 1958. "The basic psychology of rumor." Pp. 54-65 in Eleanor E. Maccoby, Theodore M. Newcomb, and Eugene L. Hartley, editors. Readings in Social Psychology. New York: Henry Holt.
 

Bartlett, Frederick C. 1958. "Social factors in recall." Pp. 47-54 in Eleanor E. Maccoby, Theodore M. Newcomb, and Eugene L. Hartley, editors. Readings in Social Psychology. New York: Henry Holt.
 

Bass, Alison B. 1990. "Computers in the classroom." Pp. 34-42 in Kathryn Schellenberg, editor. Computers in Society. Guilford, CT: Dushkin.
 

Bolles, Edmund B. 1972. "Innate grammar of baby talk." Saturday Review 55:52-55.
 

Collins, Randall. 1984. "Statistics versus words." Pp. 329-62 in Randall Collins (ed.), Sociological Theory. San Francisco: Jossey-Bass.
 

Cousins, Mark and Athar Hussain. 1984. Michel Foucault. New York: St. Martin's.
 

Dreyfus, Hubert L. and Paul Rabinow. 1982. Michel Foucault: Beyond Structuralism and Hermeneutics. Chicago: University of Chicago Press.
 

Dreyfus, Hubert L. and Stuart E. 1986. "Why computers may never think like people." Technology Review 89:42-61.
 

Eribon, Didier. 1991. Michel Foucault. Cambridge: Harvard University Press.
 

Faia, Michael A. "The Mastectomy Complex." 1977. Worm Runner's Digest XIX: 109-10.
 

Flippo, Chris. 1987. When Women Say Yes: Sexual Initiation, Minority Status, and Opportunity Costs. Unpublished master's thesis, College of William & Mary.
 

Foucault, Michel. 1973. The Order of Things. New York: Vintage.
 

Foucault, Michel. 1978. The History of Sexuality. Volume 1. New York: Pantheon.
 

Foucault, Michel. 1984. "What is an author?" Pp. 101-20 in Paul Rabinow, editor. The Foucault Reader. New York: Pantheon.
 

Hall, Donald. 1973. Writing Well. Boston: Little, Brown.
 

Harris, Marvin. 1979. Cultural Materialism: The Struggle for a Science of Culture. New York: Vintage Books.
 

Koestler, Arthur. 1964. The Act of Creation. New York: Macmillan.
 

Korzybski, Alfred. 1933. Science and Sanity. Lancaster, PA: Science Press.
 

Lamont, Michele. 1987. "How to become a dominant French philosopher: the case of Jacques Derrida." American Journal of Sociology 93:584-622.
 

Lemert, Charles C. and Garth Gillan. 1982. Michel Foucault: Social Theory and Transgression. New York: Columbia University Press.
 

Lundberg, George A. 1947. Can Science Save Us? New York: Longmans, Green.
 

Main, C. F. and Peter J. Seng. 1961. Poems. San Francisco: Wadsworth.
 

Mayr, Ernst. 1991. One Long Argument: Charles Darwin and the Genesis of Modern Evolutionary Thought. Cambridge: Harvard University Press.
 

McDowell, Edwin. 1986. "Mago de las palabras." Americas [Sp. ed.] 38:36-39.
 

McKelvey, Bill. 1982. Organizational Systematics: Taxonomy, Evolution, Classification. Berkeley: University of California.
 

Ogden, C. K. and I. A. Richards. 1930. The Meaning of Meaning. New York: Harcourt, Brace.
 

Rosaldo, Renato. 1989. Culture and Truth: The Remaking of Social Analysis. Boston: Beacon Press.
 

Rosenthal, Robert and Lenore Jacobson. 1968. Pygmalion in the Classroom: Teacher Expectation and Pupils' Intellectual Development. New York, Holt, Rinehart and Winston.
 

Ross, Stephen D. 1990. "Power, discourse, and technology: The presence of the future." Pp. 255-72 in Gary Shapiro, editor. After the Future: Postmodern Times and Places. Albany, NY: State University of New York Press.
 

Rubenstein, Diane. 1988. "Food for thought: Metonymy in the late Foucault." Pp. 83-101 in Bernauer, James and David Rasmussen, editors. The Final Foucault. Cambridge: MIT Press.
 

Sawicki, Jana. 1991. Disciplining Foucault: Feminism, Power, and the Body. New York: Routledge.
 

Sherman, Lawrence W. and Richard A. Berk. 1984. "The specific deterrent effects of arrest for domestic assault." American Sociological Review 49:261-72.
 

Sherman, Lawrence W., Douglas A. Smith, Janell D. Schmidt, and Dennis P. Rogan. 1992. "Crime, punishment, and stake in conformity: Legal and informal control of domestic violence." American Sociological Review 57:680-90.
 

Simons, Marlise 1988. (April 10). "The best years of his life: an interview with Gabriel Garcia Marquez." New York Times Book Review.
 

Starmer, Garrett L. 1955. "And shun the frumious jabberwock." Pp. 325-28 in William G. Leary and James S. Smith, editors. Thought and Statement. New York: Harcourt Brace.
 

Steiner, George. 1975. After Babel. New York: Oxford University Press.
 

Stratton, Jon. 1990. Writing Sites: A Genealogy of the Postmodern World. Ann Arbor: University of Michigan Press.
 

Swafford, Michael. 1980. "Three parametric techniques for contingency table analysis: A nontechnical commentary." American Sociological Review 45:664-90.
 

Twain, Mark. 1875. Sketches New and Old. New York: Harper and Brothers.
 

Twain, Mark. 1917. What is Man? New York: Harper and Brothers.
 

Van Dijk, Teun A. 1985. Handbook of Discourse Analysis, Volume IV: Discourse Analysis in Society. New York: Academic Press.
 

Wolfe, Alan. 1991. "Mind, self, society, and computer: Artificial intelligence and the sociology of mind." American Journal of Sociology 96:1073-96.

Chapter 3

IN PRAISE OF THE NULL HYPOTHESIS:

THE MYTH OF "THE VALUE-FREE MYTH"
 
 
 
 
 
 
 
 
 
 
 

--Fernand Braudel (1990:16)
 

(1) The nature and extent of bias in scientific research
 

Back in the 'fifties, when I first began serious study of the social sciences, it seemed to be taken for granted by most of my mentors that intellectual biases, although very subtle, seductive, and dangerous, were subject to a modicum of control. One of the best ways of controlling bias, they believed, was to learn to use the scientific method, especially those aspects of it that had to do with critical analysis (and replication) of the work of others, i.e., those aspects that had something to do with consensus-seeking. I found these arguments to be highly persuasive. Among the latest generations of scholars, however, an entirely new attitude has been emerging: It is commonly asserted nowadays that there is nothing special about scientific methods and scientific discourse as a way of controlling bias, and that the "knowledge" of the sciences--and this surely includes the social sciences--is deeply and inescapably distorted by the prejudices and presuppositions of scholars.

Writers who have major misgivings about the claims of scientific method--Agger (1989:373), for instance, who writes magnanimously that "I want to get positivism off the backs of those of us who claim science for ourselves"--are wont to reject causal analysis, path diagrams, graphic representation of data and hypotheses, and other such devices, but this attitude should not deter us from insisting that hypotheses about intellectual biases are typically causal hypotheses: They state that a given intellectual production is influenced by (1) historical factors, and/or (2) one's social class background, sex, race, etc., and/or (3) other elements of one's intellectual development, e.g., religious beliefs, and/or (4) one's personality characteristics, life experiences, neurotic conflicts, etc. Levy (1989:3) makes exactly the same point when he says that "the idea that an observer's values and concepts influence what he or she observes and concludes is nothing but a special case of the general proposition that an empirical explanation exists for anything empirically observable."

Moreover, it is also possible that a given intellectual production would be looked upon as a causal agent in itself, influencing, say, historical processes, or the further psychic development of individuals. In a moment, for instance, I shall explore the question whether the Comtean positivism of the Mexican political elite, during the decades prior to the revolution of 1910-20, had any appreciable impact on the development of recent Mexican history.

Admittedly this sort of causal hypothesis is hard to state in specific detail, and the stipulated variables are hard to define and measure, but we must nevertheless recognize that any causal hypothesis should give way to the null hypothesis--i.e., a finding of no (causal) relationship--unless those who advance the causal hypothesis have provided convincing evidence. There is a burden of proof, in other words, on those who allege bias. In general, recent detractors of scientific method as a bias-control technique have not provided such support.

In regard to the following series of statements, collected informally over many years of conversing with or reading the works of many colleagues, I do not reject the null hypothesis of no relationship; that is, I do not believe any of the following statements. They illustrate many of the alleged interactions among intellectual productions, historical factors, socio-economic status, sex, race, religious beliefs, personal conflicts and neuroses, etc. I shall explain in considerable detail why I do not believe statements (1) and (2); the remaining assertions will be touched upon briefly, at least by implication. Although I make a few references to pertinent discussions found in the literature, I do not wish to imply that cited authors believe these statements any more than I do. Nor do I wish to imply that statements of this genre are always wrong. I only wish to suggest that they are wrong far more often than we realize.

Here, then, are the statements:
 

Harding (1986) presents many variations on (13).
 

(2) Quashing the indictment: Can a Comtean rule a country?
 

I take this as my major operating premise: If one wishes to demonstrate that positivism in the social sciences, over the last century, has not been the demons' dance that it is sometimes said to be, then Mexico during the years of the Porfirian dictatorship (1876-1910) presents the challenge of being a worst-case scenario, or at least a poor-prospects scenario. If the positivist philosophy can be exculpated in this instance, it can be exculpated in virtually any instance. If it receives a large share of the blame for the Porfirian dictatorship, then it is a blameworthy philosophy indeed.
 

(2.1) Positions
 

I am reluctant to say that the positivists "ran Mexico" during the Porfiriato, but I am convinced that Mexican positivism provides a rare instance in which a coherent non-Marxist social science theory has been placed in the service--for good or bad--of major political power. "Placed in the service" is a phrase that is subject to interpretation: It is very difficult to determine the degree of Comtean expertise achieved by the Diaz positivists, it is equally difficult to discover the degree to which governmental policies were derived from the positivist premises, and it is nearly impossible to determine whether Comteanism, in the Mexican context centering on the turn of the century, made much progress in achieving its various goals. Nor does Leopoldo Zea (1974:20,21) help when he says that "... neither positivism nor any other doctrine causes social well-being or social misery, but rather this social well-being or social misery expresses itself in a doctrine," and that "to say that one thing is theoretical positivism and another thing is practical positivism implies a lack of intellectual responsibility ... It is like hiding the hand that throws the rock. The purpose of every theory is to put that theory into practice. The theory is responsible for the practical effects of the theory." As we shall see, Zea cannot have it both ways: A close reading shows that he, along with Carleton Beals, prefers the first of these two theses, seeing in Mexican positivism a sort of post-facto rationalization for policies set in motion by non-ideological factors.

Beals (1932:323) describes an elaborate public occasion, just after the turn of the century, in which a high official of the cabinet of Porfirio Diaz, dictator of Mexico, seems to be practicing racism against Mexican Indians and mestizos in order to impress European guests of the Diaz government. Jose Yves Limantour, the perpetrator of this incident and himself an ardent Comtean positivist and evanescent heir to the Mexican presidency, exemplified the racism of the times. This was the racism of positivistic organicism, Ratzenhofer, Gumplowitz, the social darwinists, various eugenics movements, the American Protective Association, Samuel George Morton, Edward B. Tylor, Lewis Henry Morgan, and contemporary editions of the Encyclopedia Britannica. The intellectual portmanteau of Mexican positivism contained the racist paraphernalia found elsewhere in the academic settings of the western world. Beyond that, Zea (1974:xx) believes that, in part, the hidden agenda of the Mexican positivists was "... an attempt, as Justo Sierra picturesquely describes it, to Saxonize Mexico, that is, to convert the Mexicans into 'Yankees of the South,'" and one can readily surmise that this attitude, this strong xenocentrism, was related to the prevalent racism.

Mexican positivists read their Comte with care and diligence. Aside from its social darwinist commitments and its racist undertones, Mexican positivism seemed to have a characteristically Comtean notion of ineluctable historical stages--theological, metaphysical, and scientific. Mexican positivism was a variety of positivistic organicism (Martindale, 1960:65), and therefore it tended to look upon expanding international commerce as an historical inevitability and a very good thing indeed. The belief that theology and metaphysics must give way to science may have added to anti-clerical traditions already firmly implanted in Mexico. And when the Mexican positivists considered the level of domestic tranquillity then established in their nation, they must have agreed readily with Benito Juarez, the greatest of Mexican national heroes¹, when he claimed to see in positivism a way of containing anarchy (Zea, 1974:40-41). It was Juarez who in 1867 asked Gabino Barreda to collaborate with a national commission in the writing of a new plan of education (Zea, 1974:39). Barreda had been in direct contact with Comte in France, and was deeply immersed in Comtean philosophy. Octavio Paz surely does not miss the irony in the fact that a nation that had just succeeded in expelling French colonialism would find itself flirting with--indeed, carrying on a courtship with--French social philosophy.

The Asociación metodófila, established in 1877 and largely inspired by Barreda's educational programs, "... was to show how a group of men in different fields of specialization could understand each other and unite by means of certain basic principles, and by means of a method of interpretation that could be applied uniformly to diverse problems." The program, in other words, contemplates the level of erudition, the synthesizing and generalizing of other scientific disciplines, that has always been the central tenet of Comtean sociology. The Asociación was given to the practice of consensus politics--not so much because it was politic to pursue this practice but because it was good science. Philosophical issues such as the role of human freedom--the free-will controversy--were handled with a deftness that characterizes modern sociology textbooks. That which is taken to be free--say, a body in "free fall"--is behaving in accordance with discernable natural laws, moving in this instance with a strict inertia interrupted only by a predictable acceleration of thirty-two feet per second per second. Whether such a body is truly free, again, can be ascertained just as surely as we ascertain what is transpiring a mile north of the north pole, and just as surely as we ascertain whether Anatole France was speaking of human freedom when he said that both the rich and the poor are free to live under bridges.

So much for metaphysics.
 

(2.2) Correctives
 

And so much for Agger's (1989:21) "unalterable world" as a summing up of the positivist Weltanschauung. Mexican positivism demonstrates the clear untenability of the charge (Agger, 1989:21) that "... sociology suppresses its narrative imagination precisely to have its reflection of invariant social patterns acted out uncritically." Those members of the Diaz government who called themselves positivists were not a uniformly admirable group, but they had a far greater flexibility in the application of their social philosophy than is implied by the label "unalterable world." Racist philosophy as a respectable academic calling dried up within the Mexican elite just as surely as it has dried up elsewhere. Even though Mexican positivism may have insisted on the anachronistic character of theology, religion, and specifically Catholicism, it eventually developed a surprising tolerance toward religion, an attitude that Beals (1932:340-41), no admirer of the positivists, regarded as highly conciliatory. Perhaps most importantly, there was substantial positivist opposition to the higher immoralities of the Diaz dictatorship (Cockcroft, 1968:59-60; 165), and several positivists eventually found themselves serving in the cabinet of the revolutionary government of Francisco Madero (Cockcroft, 1968:205). In brief, although the positivists emphasized consensus-seeking, they were by no means committed to an unalterable consensus. There were inexorable confrontations of thesis and antithesis: The Mexican political elite, in Beezley's (1987:13) view, "... subscribed to a loose sense of progress, based on Comtean positivism with individual touches of Catholicism or anticlericalism, of Indianism or anti-Indianism, and of greater or lesser doses of the Liberal belief in the efficacy of property."
 

(2.3) Motives: A series of acts contrary ...
 

As I implied earlier, Zea had trouble deciding whether Mexican positivism was a collection of "ideas that matter," or merely a set of post-facto rationalizations--more cynically, a manipulative ideology of rule. This issue cannot finally be resolved, but those who agree with Agger's contention that the positivist philosophy is a mischievous monkey that must be gotten off our backs would do well to consider the many mitigating circumstances that apply in the Porfirian instance--an instance in which the positivist philosophy is clearly at high risk of high culpability. After cavalierly dismissing the científico version of positivism as "drivel," Beals (1932:324) takes great pain to show that Mexican positivism was little more than a post facto rationalization for what amounts to systematic social banditry on the part of Diaz' major political henchmen. His chapter entitled "profits of positivism" (Beals, 1932: Chapter 43) makes the case in detail, and the chapter is not vitiated by the fact that it is titled wrongly--positivism itself produced no profits, but merely served as an element of the legerdemain of very clever social bandits. Zea makes the same case, although far more convincingly. It was not so much positivism per se, he says, that was attacked by its many detractors (Zea, 1974:15-16), but rather the expression of positivism as transformed by the científicos:
 

The positivist Jose Torres, cited heavily and with strong approval by Zea (1974:17-18,221),
 

Just as Talcott Parsons appears to have had an attachment to political stability that had nothing to do with his theoretical predilections (Alexander, 1983:186), so the Mexican positivists may have pursued political goals and personal interests that had little if anything to do with their own unique brand of positivism, and much to do with "... one basic idea, to steal, much, often and scientifically" (Beals, 1932:334). Regarding matters of politico-economic policy, Beals (1932:343-44) points to an intellectual tabula rasa: "... Diaz had no economic plan; Goddess Fortune directed it entirely."

In closing the abridged English translation of his inquiry into Mexican positivism, Zea (1968:221) concludes that "Jose Torres was right when he asserted that positivism was not to blame for the evils of Mexico. The positivist ideal was one thing and the reality called Porfirismo was another. ... Positivism provided the weapon by which to justify a series of acts contrary to the positivist ideal." In the Mexican edition of 1968, in Zea's Spanish, this final sentence is italicized (Zea, 1968:229).
 

(3) Quashing the indictment: The case against functional analysis²
 

In the long and tortuous history of functionalist theory, consensus usually has been no more evident than in the case of Mexican positivism. At any given historical moment functionalists of a given genre are struggling valiantly against those of a discernably different point of view, or perhaps against a vociferous school of non-functionalists or even anti-functionalists; anti-functionalism has a history at least as elaborate as that of anti-positivism. Eventually, one sees the wisdom of Ritzer's claim (1980:201-208) that paradigm disputes probably do more harm than good. On the other hand, this may be an auspicious time to rewrite these histories, recognizing (and perhaps insisting) that entire paradigms, in a holistic sense, are rarely what is at stake (Glymour, 1980: esp. Chapter 5). Positivistic organicism, i.e., scientifically-oriented functionalism, has shown an extraordinary capacity for correcting itself piecemeal, as in the case of its repudiation of racism.

In Harris' excellent chapters on the subject (1968), it becomes clear that positivistic functionalism, with its constant internecine warfare, damaged itself in several ways: First, its early overindulgence in the organic analogy, social darwinism, racism, sexism, and unilinear evolutionary theory (parallelism); second, its subsequent over-reliance--a sort of compensation for earlier excesses--on synchronic (cross-sectional) analysis, its occasional repudiation of the idea of structural survivals, its occasional indulgence in hopelessly abstract Hegelian idealism, its apparent conservatism, and its alleged willingness to tolerate questionable research practices. These various elements do not necessarily fit together into a coherent package, and we encounter here the same sort of hodge-podge of ideas that we encountered in the case of Mexican positivism. The Mexican positivists showed a similar high talent for picking and choosing among the great historical fallacies.

While I consider most functionalists to be guilty on at least a few of Harris' several counts, I believe that all sorts of extenuating circumstances now require clemency and perhaps even amnesty. In any case, regardless of how severely we castigate functionalists, we must recognize that the litany of charges against them does not necessarily tell us anything about the strengths and weaknesses of the central premises of functional analysis. We must avoid the synecdochic fallacy: A small part of a theory does not represent the whole, and may prove to be misleading and highly expendable. The occurrence of an occasional illness reveals little if anything about the general condition of the organism; what is important is the capacity for making a recovery. Again, as Glymour (1980) suggests, it is possible to reject part (or several parts) of a large theory without rejecting its totality.
 

(3.1) From Spencer to Weber
 

Martindale (1960:65) gives Herbert Spencer credit for having provided the philosophical groundwork for positivistic organicism, and a detailed examination of Spencer's organicism makes it clear that, once we have stripped away a few quaint, inessential, and perhaps outrageous notions no longer taken seriously, it subsumes or surpasses all the central premises of modern functionalism, including "neo-functionalism" with its revival of social evolutionism. Martindale (1960:67) summarizes Spencer's organic analogy as follows:
 

If the sociology of the twentieth century had been content to pursue the Spencerian premises to their logical conclusions as a way of understanding social dynamics, rather than pointing to the inevitable limitations of the organic analogy and to the many faults, foibles, and egregious claims of Spencer and his followers, contemporary sociology probably would be in a much stronger position than it is, and would not be worried about an alleged tendency to fall into the doldrums. A perusal of Spencer's sociology--never mind his many contributions to philosophy, biology, and psychology--shows that it is, as Lenin once said of Clausewitz' writings on war and diplomacy, a splendid collection of fine points.

In his discussion of "social metamorphoses," for instance, it becomes clear that Spencer's positivistic organicism has a strong sense of the sometimes discontinuous character of social evolution--that is, a sense of social change that has nothing whatsoever to do with "unalterable worlds":
 

Major ecological events, then, such as those subsumed under the "hydraulic hypothesis" (Harris, 1979:104-5), continually create massive social dislocations. And warfare, to Spencer, is another powerful engine of social evolution:
 

Here we have an extraordinary irony: The man who coined the phrase "survival of the fittest" (Harris, 1968:128) brings forth a theory that implies a social metamorphosis in which past warmaking either loses its functionality or evolves in such a way as to take on new non-warmaking functions. Spencer, then, clearly envisions the occasional disassociation of structure and function, and makes imaginative use of that disassociation as a mechanism of social transformation--"unalterable worlds," again, have no place. And if Spencer's functionalism does not have any ultimate commitment to social conflict, perhaps more recent versions of functionalism do not, or need not, have any ultimate commitment to social integration, cooperation, and consensus. Our politics--as in the case of the Mexican positivists--may be largely a matter of personal predilection, not theoretical necessity.

It is arguable, then, that the worst nineteenth-century snafus of positivistic organicism--starting, say, with the cranial-capacity indexes of Samuel G. Morton--resulted from the fact that biology was very much in the air, and nobody was exempt. Similarly, the stigmata attaching themselves to contemporary North American versions of positivist functional analysis may have been created by the prevailing political climate of recent decades. The following comment was written more than thirty years ago:
 

There is a political aura, but the aura does not emanate from the object that it surrounds. I doubt Martindale's claim that ideology--the current political climate--is now less important than heretofore, primarily because I see so little evidence of its past significance as an inescapable determinant or consequence of one's theoretical strategies. I have yet to discover evidence that the political conservatism of a scholar such as Spencer was a necessary consequence of his positivistic organicism, I have grave doubts that the arrival into the functionalist camp of "liberal" social behaviorists had a large political impact on either field (cf. Faia, 1974), and I do not see any reason to believe that Martindale is calling our attention to anything more than "accidental correlations," as he himself surmises. Accidental correlations, by definition, occur when the null hypothesis prevails, and it is the thesis of this chapter that, in most instances where "ideological bias" is the leading theoretical hypothesis, the null hypothesis will continue to prevail.

On the matter of Spencer's celebrated distrust of government, his consequent laissez-faire predilections, and the relationship of these to his sociological theories, Barnes (1948:129) has provided the definitive alternative hypothesis. Spencer, he says, had an "anti-authority complex" exacerbated by various experiences of childhood, especially his "domination by male relatives, and his confirmed neurotic tendencies." We must also remember that Spencer came from a "... dissenting family and was reared in that atmosphere." It seems, says Barnes, "that his attitude toward government must have had a deep-rooted emotional foundation, since it diverged materially from some of the vital premises of his general philosophy." This style of psycho-history explains Spencer's politics far more plausibly than does his advocacy of evolutionary positivistic organicism. And nota bene: Barnes is explaining Spencer's politics with reference to his personal neurotic conflicts; Spencer's positivistic organicism is not implicated either as cause or effect.

If anybody ever returns a plausible diagnosis of neuroticism in the case of Talcott Parsons, the thesis that political values are sharply constrained by theories, or vice versa, will receive yet another devastating setback. In the meantime, Sica's (1988) book on Max Weber is replete with hypotheses of a psycho-historical nature, and the assumed psychodynamics, again, usually have nothing demonstrable to do with Weber's "general philosophy."

Even when the psychodynamic explanation of a general philosophy has a surface plausibility, the evidence presented generally leaves a lot to be desired, and again one has no recourse but the null hypothesis: Nothing has been demonstrated, and the theories in question have to be tested and evaluated on the assumption that they have been derived through legitimate scholarly efforts, not through idiosyncratic psychic disturbances. Sica's (1988:221) description of Weber's alleged neurotic interactions with his parents, then, appears to be little more than an exercise in phrenology:
 

Needless to say, these expressions were not obvious to this reader.
 

(3.2) Pareto
 

I have always believed that many of the laissez-faire philosophers were merely warning us against grabbing the flywheel of the social engine, when the appropriate tactic is to seek out the ignition switch or some way of leaning out the fuel mixture. I therefore believe that Harris (1968:534-35) goes too far when he draws a strong analogy
 

Perhaps this claim can be made against Malinowski and (more debatably) against Radcliffe-Brown, but it can hardly be sustained against Spencer or Pareto, neither of whom ever insisted upon universal functionality and whose laissez-faire proclivities cannot possibly have been derived from the anti-evolutionism of later British anthropologists. Even Malinowski and Radcliffe-Brown did not play down the evolutionary themes because of some latent commitment to know-nothingism or laissez-faire politics, but rather because they wished to disassociate themselves from the ethnocentric evolutionary theories of some of their predecessors, and to escape the recrudescence of social darwinism, racism, and sexism that had stained the escutcheons of Spencerian theory. In other words, they overreacted. Harris himself (1968:54) points to one of the positive consequences of the self-regulating laissez-faire model when he says that, as exemplified in the works of Jeremy Bentham, James Mill, David Ricardo, and John McCulloch, it contributed to "... the perpetuation of the scientific approach"--in modern times, in the midst of a major assault against positivism, we realize that this function is of the utmost importance.

As for Pareto, perhaps we should conclude by questioning the ugliest of the political accusations against him, viz., that he somehow aided and abetted the Italian fascist movement. In Lopreato's assessment (1980:xix-xx), Pareto realized by 1922
 

(3.3) Malinowski and Radcliffe-Brown
 

Malinowski's major contribution to positivistic organicism was to raise poignant questions about the unilinear predilections of earlier functionalists, about their use of the comparative method as a way of comprehending the distant history of human societies, and about the use of survivals as keys to the past. Malinowski's critics have accused him of adopting a horribly constraining synchronic approach to cultural anthropology, but Malinowski himself (1944:16,175-76; 1982:60) makes it clear that his position is far from simplistic:
 

Malinowski equivocates over the meaning of "origins," but I do not discern here an attitude of total hostility toward history, and there is certainly no commitment to "unalterable worlds"; perhaps Malinowski was guilty only of taking seriously Pareto's contention (1980:77) that "in general the unknown has to be explained by the known, and the past is therefore better explained by the present than the present is by the past"; teach history, in other words, with the calendar running in reverse.

As for the Malinowskian emphasis on social equilibrium, there is for Radcliffe-Brown (1952:43) nothing magical or inevitable about it:
 

Perhaps Gregg and Williams (1948) have a point in accusing Malinowski of being charmed and seduced by the notion of "functioning harmonious wholes," but they cannot make a similar charge against Radcliffe-Brown (Faia, 1986:159-60), and this leads us to surmise that the charge against Malinowski merely reflects a personal quirk.
 

(3.4) Parsons
 

For something like thirty years, starting around 1940, Parsons' work provided and reflected a dominant social science paradigm. In Alexander's view (1983:7), the strength of Parsonian thought derives mainly from its multidimensionality, its systemic features; that is, from its tendency to focus on the interactions of biology, personality, society, and culture. The "cybernetic continuum," operating through various "exchange media," makes possible a degree of integration of highly disparate social phenomena, so that society is seen as comprising the ultimate form of Spencer's "coherent heterogeneity" and clearly falls under the rubrics of positivistic organicism. Parsons' predilection for systemic perspectives made him a strong advocate of paradigm integration in the social sciences.

For my present purposes it is important to show the untenability of the conventional claim that Parsonian theory involves an inescapable, albeit tacit, commitment to social stability. Those who make this claim must find a way to explain the following anomalies: First, Parsons has a clear imagery of structural strain and its ability to produce adaptive efforts or to disrupt stability. He occasionally interprets social systems from the standpoint of what I call the "GM analogy" (Parsons, 1937:31, 230), and he never regards reciprocating engines as inherently smooth-running--if abused, they tend to blow up. Second, his insistence that utilitarian theory is built upon the mistaken notion of "randomness of ends," and that this venerable theory therefore fails to explain ends adequately (1937:59-60, 231), forces us to examine reciprocal interaction between ends and other aspects of the social system. Non-random processes, by definition, are subject to explanation. If explanation implies causation, and if causation implies manipulation of causal agents, and if manipulation of causal agents implies change, then non-randomness of ends implies change. Third, when Parsons (1937:346) points to Durkheim's thesis on the "anteriority" of social facts, implying that many social facts therefore do not have utilitarian value for given actors, he inadvertently introduces the important concept of structural survivals and the possibility of a disassociation of structure and function. This disassociation is an essential element of the functional analysis of social change, and Malinowski's work, as we have seen, was surely not aided by his denial of it. Fourth, despite his commitment to requisite analysis, Parsons (1951:167-77) has a clear recognition of the possibility of functional alternatives; selection among alternatives implies change, and it also implies the creation of new survivals.

The modal accusation against Parsons, weak though it may be, is that he had political predilections toward social stability. Many readers who have made a careful analysis of The Social System while remaining a little cavalier in their treatment of other Parsonian works, have had understandable problems with the statement that "... it is essential to the conception of the interaction process ... and of the theorem of the institutional integration of motivation ... that the stabilization of the processes of mutual orientation within complementary roles is a fundamental 'tendency' of interaction" (1951:481). This sounds like an a priori postulate of social stability. Parsons does not help the situation--in fact, he aggravates it--when he tells us that the stability postulate is a theoretical assumption and not an empirical generalization (1951:481), as if there were a sharp discontinuity between the theoretical and empirical levels of analysis. Parsons further argues that social systems, having properties of equifinality, are often able to tolerate "... fluctuations in the factors of the environment" while maintaining "certain constancies of pattern, whether this constancy be static or moving" (1951:482). Never mind the fact that Parsons' "moving" constancy is not precisely an instance of stability, or that for every instance of equifinality one can easily identify a departure from equifinality, as Radcliffe-Brown suggests above. If hunting-gathering bands are running out of food, they had better forget about constancies of pattern.

Equifinality implies disturbances of equilibrium in the same sense that the landing of an aircraft implies an earlier takeoff. And my flight instructors have always insisted that the total number of my landings (L) equal the total number of my takeoffs (T). If T - L ever equals 1 or more, I will have experienced profound change.

Perhaps the best instance of an unreasonable adherence to stability and inertia in Parsons' work is his theory of sex roles and their allocation within nuclear families. These writings do seem to exemplify what feminists have called a "functional freeze" (Friedan, 1963; Parsons, 1951:193, 241, 503-505; Parsons et al., 1955: Chapter II). My own work on functionalist theory, however, has persuaded me that the alleged functionalist crystallization of sex roles was the result of an inappropriate application of requisite analysis, and that requisite analysis should long since have been jettisoned as an element of functional analysis.

Consider, for instance, a recent work by Levy (1989). This book seeks to demonstrate the universality of the nuclear or quasi-nuclear family; it seeks to demonstrate the centrality of mothers and mother-substitutes as agents of infant and child socialization in virtually every realm of cultural activity; and it purports to explain these phenomena, these social structures, by means of requisite analysis. A "structural requisite," as defined in the glossary of this volume, is "a pattern or observable uniformity of action or operation necessary for the continued existence of the unit with which it is associated ..." (Levy, 1989:228). Very early on (Levy, 1989:xv,9) the author invokes the concept of structural requisites, stating that this concept will be fundamental to his method of theoretical inquiry. And yet, one reads the entire volume without finding a single instance in which a social structure has been shown to be necessary for the continued existence of any unit with which it is associated. It is not shown, for example, that the centrality of motherhood is necessary for the continued existence of the nuclear or quasi-nuclear family, for the continued existence of adequate socialization of infant and child, or for the continued existence of the larger social system and its many components. Nor do we encounter any attempt to explain which requisites are met, if any, by the nuclear or quasi-nuclear family.

On the other hand, given that Levy himself (1989:39) agrees that current family structures will not necessarily continue indefinitely, what happens to his brand of requisite analysis? And why does he argue, in spite of what he has just said, that the current situation will persist?³ I contend that there is a presupposition in requisite analysis that any social structure, considered abstractly, that is a universal or quasi-universal must be an inescapable requisite for something. But for what? And how do we demonstrate this relationship? Given the fact that functional analysis always involves circular causation, there are a few additional questions to be raised here: How could the current (universal?) family structure be changed? And why are the factors capable of producing such changes never activated, so that the current situation persists? If specific causal hypotheses about requisites were being stated and tested, these questions could be addressed. The fact that they cannot be addressed within Levy's perspective implies strongly that the hypotheses have not been developed. I made precisely the same point several years ago (Faia, 1986:169).

After presenting a large part of his argument, Levy (1989:123) arrives at an inescapable impasse: He confesses that neither he nor anybody else, confronted by the ubiquity of the nuclear family and the centrality of motherhood, can explain these structures adequately. I would argue, therefore, that this is a descriptive study, and not a traditional functionalist study of requisites. When Levy realizes that social scientists are not prepared to answer the questions raised above, he seems to fall back upon a vague hope for some sort of reductionist bail-out by sociobiology, stating at one point (Levy, 1989:26-8) that
 

And within a few pages (Levy, 1989:28) we encounter the example of mother cats inducing peristalsis in their kittens by massaging their bellies; and a second cat story, this time involving male and female lions and the nature of their hunting-gathering activities, occurs in a later context (Levy, 1989:86-87). In these commentaries we see what amounts to a prayer for sociobiology to the rescue, despite the fact that sociobiology has had few empirically substantiated successes in explaining the interaction of human biology and social behavior. Phenylketonuria (PKU) and Tay-Sachs disease, involving readily identifiable genes and the prospect of cultural interventions that head off their damaging effects, are among a short list of prime examples ("The Telltale Gene," July 1990:483-84). I remain skeptical about Desmond Morris' (1967: Chapter 2) famous speculation that face-to-face sexual intercourse, with its massive social implications, was an evolutionary breakthrough that occurred when the frontal nudity of females began to replicate--perhaps simulate is a better word--their rearward nudity. The biosociology of face-to-face intercourse must be at least as edifying as the sociobiology of this important invention (cf. Jencks, 1992:92-119).

Levy's best discussion of causal dynamics, in fact, occurs in chapter 10 where he argues that various dimensions of human relationships, within the realm of kinship or elsewhere, are highly correlated with each other, e.g., "arational" relationships tend to be "particularistic," "intimate" relationships tend to be "responsible," etc. Such correlations are summarized efficiently in Levy's (1989:174-76) special notation. There is also in this part of Levy's presentation a diachronic dimension: A rational relationship, for instance, typically evolves toward being arational, individualistic relationships evolve toward being responsible, etc. Presumably, combinations of "relationship aspects" (as he calls them) that are not highly correlated with each other are selected against by various social factors. Further, those relationship aspects that tend to occur at the end of social-evolutionary processes seem to receive favorable selection. But Levy's brand of requisite analysis, again, remains silent as to what specific social factors might be implicated in these social-selection processes.

Social relationships that are rational and universalistic, but also functionally diffuse, intimate, responsible, and non-hierarchical appear somewhere on Levy's list of logical possibilities, but relationships with this combination of qualities, in Levy's view, would be highly improbable. Logically, his Table 1 (Levy, 1989:136) permits 2⁶ or 64 distinctive types of social relationships, although Levy considers only two of these types to have a high probability of occurrence. And of these two types, the first has a tendency to evolve into the second.

Most of the other 62 forms, then, appear to be more or less untenable, for reasons never specified beyond noting that they are forms that, in Levy's estimation, tend to be rare. But regarding the argument that absence (or rarity) implies untenability, we must note Dawkins' (1986:9) remark that in the realm of natural history there are many more ways of being dead than of being alive, and that we know far less about ways of being dead than we know about ways of being alive--which admittedly is not much. The problem, then, is that many ways of being dead probably would work well if a resurrection could be arranged in an appropriate context, or ecological niche. Why, then, do such resurrections tend not to occur? In biology, the answer is that nobody knows how to resurrect extinct species; they are infinitely more complex than what is implied, say, by Levy's list of 64 forms of social relationships. Yet, it would appear that in the social realm we have a relatively good prospect of resurrecting extinct structures or creating new structures from the lists of logical possibilities supplied by the likes of Levy. And the best news is that there is nothing about Levy's brand of requisite analysis that rules out the various forms of social engineering implied by his unique set of pattern variables, or by the similar taxonomies produced by other scholars.

Levy's experiments with pattern variables, of course, are a Parsonian penchant, and his analyses of specific instances of social change do not generally insist on a tendency toward stability. The Social System presents an array of conditions and processes capable of producing change from within a society, including not only selection against untenable combinations of the pattern variables (1951:152)--the matter just discussed in some detail in connection with Levy's work--but also several forms of structural strain (1951:179-200), deviant behavior (1951:206), the extreme malleability of the human infant (1951:336, 508), a general strain toward consistency in values (1951:383), new religious ideas, genetic change (1951:493), and the changeability of functions performed by given structures (again, a prospect of survivals), as in the example of horses (1951:511).

Perhaps Parsons' best statements about external sources of change are found in the extraordinary series of essays on the historical evolution of Germany and Japan, and on the prospects of social engineering in these societies after 1945 (Parsons, 1954).

In short, the various labels affixed to Parsons are nearly as variegated and inexplicable as those applied to Durkheim (Levine, 1985:55-6):
 

I don't know what these labels tell us about the eye of the beholder, but they tell us next to nothing about Durkheim or about Durkheim's contributions to sociological theory. Many of these labels have the binary, dialectical, contradictory qualities of folk wisdom, comparable to informing us that "birds of a feather flock together" while "opposites attract," that "absence makes the heart grow fonder" and "out of sight, out of mind," or that "a bird in hand is better than two in a bush," but "if at first you don't succeed, try, try again."
 

(4) Overturned on appeal: Research on political power
 

Procedurally, the case looks strong (Walton, 1970:443):
 

This inquiry, in other words, does not tell us much about power structures in American communities, but it may give us insight into the ways in which methods and "substantive" factors relate to what we think we know about community power. That is, it may give us insight into our biases.

And there are, indeed, some fascinating findings. Most importantly, it appears that the "reputational" method usually identifies pyramidal power structures, while those methods that focus on actual decisionmaking processes are more likely to discover factionalized and coalitional power structures (Walton, 1970:452). Regarding the "substantive" factors it was found, for instance, that communities with high population growth tend to have more "concentrated" power structures (Walton, 1970:452).

In evaluating Walton's work as a study of investigator bias, we should first realize that the "substantive" findings do not necessarily tell us anything about bias: It is entirely possible, for instance, that the statement about population growth and power concentration may be an accurate, unbiased description of the realities of various types of American communities. The case for bias must be built on findings of the first kind: When the selection of one's method seems to have a strong influence on one's findings, we begin to suspect that those who have a liking for certain findings will consciously or unconsciously select their methods accordingly. Clearly, whenever a relationship exists between methods and results, the grand traditions of scientific skepticism must come into full force.

In the present instance, however, inexorable skepticism leads to the realization that the reputational method tells us primarily how (presumably) knowledgeable members of a community perceive local power structures. This method produces findings that have more to do with the sociology of knowledge than with the sociology of social classes. Interestingly, the judges used in reputational studies typically perceive elitism; this is in contrast to what is usually an early lesson in our social stratification courses, to the effect that American public opinion tends to deny the existence of an elaborate system of social classes, and certainly to deny elitist dominance of the political process. Apparently the reputational studies have often found judges of high sophistication: Perhaps many of these judges have taken our courses.

In any case Walton's study, which at first appears to be an inquiry into the nature and causes of bias, may actually be--thanks to a dialectical one-eighty--an inquiry into the processes of controlling biases.
 

(5) A new hypothesis
 

Without comment (Aldous, 1991:661):
 

(6) Conclusion: Science as bias control
 

For nearly thirty-five years I have been misinforming students (and more than a few colleagues) about Malthus: Until recently I subscribed to the belief--a commonplace among demographers--that the original version (1798) of Malthus' theory of population, that of 1798, made sense for the centuries preceding 1800 or thereabouts, but became invalid for subsequent times. Over the last few years, however, I've found it necessary to change my mind. I now believe, with Harris and Ross (1987), that Malthus' famous postulate about the "passion between the sexes" as a source of constantly high fertility throughout human history was probably incorrect: The human race has had a considerable aptitude for controlling fertility by means that include feticide, infanticide, pedicide, and harsh treatment of women--the latter having a powerful anti-natalist impact.

On the matter of Malthus' early bias against contraception and other means of fertility control, I now realize that it must have been easy for him eventually to control this bias and to support "moral restraint" as a means of reducing fertility, for modern populations. It was easy for him because, first, the data suggesting fertility reductions among European populations during his lifetime could not be ignored, and, second, there was nothing in his version of functionalism that made such a change difficult. Malthus' functionalism did not distort his socio-demographic analysis any more than Parsons' functionalism created his alleged commitment to social stability as an operating postulate. The major reason why Malthus' modified theory of population--the 1803 version emphasizing moral restraint--is far more interesting and valuable than his first formulation is that it introduces new variables, new instabilities, and the prospect of feedback and control, and it thereby transforms itself into a much more sophisticated functionalist theory than it had been originally. Still, it did not capture the earlier historical patterns, as Harris and Ross make clear.

As I suggested earlier, it is often claimed that biases tend to occur because scholars are subject to the limitations of historicism: They wear the intellectual blinders characteristic of their times. When Marxists criticize Malthusians, however, they usually allege that Malthusians have deluded themselves into believing that they have a socio-demographic theory that is valid for all times and for all populations. This claim makes Marxists very uncomfortable, because they have a strong commitment to the notion that every historical epoch produces its own typical distortions. It is ironic, of course, that Marxists themselves have the same intellectual aggressiveness, or arrogance, that they see as the bane of Malthusianism, when they suggest that dialectical materialism is the master process of all human history.

As a Marxian functionalist with a strong admiration for Malthus, I experience infinite agonies and ecstacies from all three traditions, but what I find particularly compelling and comforting about Malthusian, Marxist, and Mertonian functionalism (Merton, 1957) is that each of these lines of analysis has the courage to claim sociological laws of wide applicability (Levy, 1989), laws that have a built-in capacity for protecting us not only against the biases of historicism but also against the sorts of biases to which I alluded earlier. What these three traditions have in common is a commitment to the scientific method, and in my estimation the scientific method itself has proven to be among the most powerful and universal of sociological laws. This set of laws clearly has the power to protect Harvard from Hoechst.
 

NOTES (Chapter 3)
 

(1) Juarez was the constitutional president of Mexico from 1857 until his death in 1872, and played a major role in the expulsion of the French forces supporting Maximilian. Maximilian was emperor of Mexico from 1864 to 1867. See Fehrenbach (1973:432-38).
 

(2) An earlier, more detailed version of section 3 appears in Faia (1986:144-65).
 

(3) Spencer's claim (1867:485-6) that modern women have a "... deficiency of reproductive power ..." due to an "... overtaxing of their brains ..." is a sexist claim in the sense that it assumes that reproductive roles are biologically linked to traditional sex roles that allegedly rule out heavy mental demands on women. And yet, Spencer's analysis consistently regards sex roles as highly changeable. It should be remembered that his speculations about declining reproductive power were an answer to Malthus: Spencer believed that modern women were indeed entering mentally demanding occupations in large numbers and that this continuing trend would relieve them and modern societies of the Malthusian curse. In other works by Spencer there are similar extenuations. It is shockingly sexist to say, for instance, that mental differences between men and women arise from an "... arrest of individual evolution..." in women, a physiological retardation that seems to make women especially fit for reproduction (Spencer, 1910:340). Yet, the same discussion anticipates Michel Foucault by claiming that women, adapting to male dominance, have developed excellent manipulative skills and an impressive ability to disguise their feelings, apparently with minimal risk of overtaxing mental faculties. In another work, Spencer argues that "... it would ... be impossible to reduce the theory of unequal rights to practice. We should yet have to find a rule by which to allot these different shares of privilege" (Spencer, 1851 [1969]:108). This thesis would seem to imply equal rights regardless of sex.
 

REFERENCES (Chapter 3)
 

Agger, Ben. 1989. Socio(onto)logy: A Disciplinary Reading. Urbana: University of Illinois Press.
 

Aldous, Joan. 1991. "In the families' ways." Contemporary Sociology 20:660-62.

Alexander, Jeffrey C. 1983. Theoretical Logic in Sociology, Volume IV: The Modern Reconstruction of Classical Thought: Talcott Parsons. Berkeley: University of California.
 

Barnes, Harry Elmer. 1948. An Introduction to the History of Sociology. Chicago: University of Chicago Press.
 

Barzun, Jacques. 1964. Science: The Glorious Entertainment. New York: Harper & Row.
 

Beals, Carleton. 1932. Porfirio Diaz: Dictator of Mexico. Westport, CT: Greenwood.
 

Beezley, William H. 1987. Judas at the Jockey Club. Lincoln: University of Nebraska Press.
 

Bennett, Neil G., editor. 1983. Sex Selection of Children. New York: Academic Press.
 

Braudel, Fernand. 1990. The Identify of France. Volume I. New York: Harper and Row.
 

Clark, Ronald W. 1971. Einstein: The Life and Times. New York: World Publishing Co.
 

Cockcroft, James D. 1968. Intellectual Precursors of the Mexican Revolution, 1900-1913. Austin: University of Texas Press.
 

Faia, Michael A. 1974. "The Myth of the Liberal Professor." Sociology of Education 47: 171-202.
 

Faia, Michael A. 1976. "Secularization and Scholarship Among American Professors." Sociological Analysis 37: 63-73.
 

Faia, Michael A. l982. "Willful, Deliberate, Premeditated and Irrational: Reflections on the Futility of Executions." State Government 55: 14-21.
 

Faia, Michael A. l986. Dynamic Functionalism: Strategy and Tactics. Cambridge University Press, 1986.
 

Faia, Michael A. February, 1990. "The Trouble With Feminist Theory." Footnotes (American Sociological Association) 18: 8-9.
 

Fehrenbach, T. R. 1973. Fire and Blood: A History of Mexico. New York: Collier.
 

Friedan, Betty. 1963. The Feminine Mystique. New York: Norton.
 

Glymour, Clark. 1980. Theory and Evidence. Princeton: Princeton University Press.
 

Gregg, Dorothy and Elgin Williams. 1948. "The dismal science of functionalism." American Anthropologist 50:594-611.
 

Harding, Sandra. 1986. The Science Question in Feminism. Ithaca: Cornell University Press.
 

Harre, Rom. 1986. Varieties of Realism: A Rationale for the Natural Sciences. New York: Basil Blackwell.
 

Harris, Marvin. 1968. The Rise of Anthropological Theory. New York: Crowell.
 

Harris, Marvin. 1979. Cultural Materialism: The Struggle for a Science of Culture. New York: Vintage Books.
 

Harris, Marvin and Eric B. Ross. 1987. Death, Sex, and Fertility: Population Regulation in Preindustrial and Developing Societies. New York: Columbia University Press.
 

Jencks, Christopher. 1992. Rethinking Social Policy: Race, Poverty, and the Underclass. Cambridge: Harvard University Press.
 

Leuba, James H. 1916. The Belief in God and Immortality. Boston: Sherman, French.

Levine, Donald N. 1985. The Flight From Ambiguity. Chicago: University of Chicago Press.
 

Levy, Marion J., Jr. 1989. Our Mother-tempers. Berkeley: University of California Press.
 

Lopreato, Joseph. 1980. "Introduction." Pp. xiii-xliv in Vilfredo Pareto, Compendium of General Sociology. Minneapolis: University of Minnesota.
 

Malinowski, Bronislaw. 1944. A Scientific Theory of Culture. Chapel Hill: University of North Carolina.
 

Malinowski, Bronislaw. 1948. Magic, Science and Religion. New York: Doubleday Anchor.
 

Merton, Robert K. 1957. Social Theory and Social Structure. New York: The Free Press.
 

Morris, Desmond. 1967. The Naked Ape. New York: McGraw-Hill.
 

Pareto, Vilfredo. 1980. Compendium of General Sociology. Minneapolis: University of Minnesota Press. (Abridged by Giulio Farina.)
 

Parsons, Talcott. 1937. The Structure of Social Action. New York: Free Press.
 

Parsons, Talcott. 1951. The Social System. Glencoe: The Free Press.
 

Parsons, Talcott et al. 1955. Family, Socialization, and Interaction Process. Glencoe: The Free Press.
 

Radcliffe-Brown, Alfred R. 1952. Structure and Function in Primitive Society. Glencoe: Free Press.
 

Ritzer, George. 1980. Sociology: A Multiple Paradigm Science. Boston: Allyn and Bacon.
 

Schrecker, Ellen W. 1986. No Ivory Tower: McCarthyism and the Universities. New York: Oxford University Press.
 

Sica, Alan. 1988. Weber, Irrationality, and Social Order. Berkeley: University of California Press.
 

Spencer, Herbert. 1851 [1969]. Social Statics. New York: A. M. Kelley.
 

Spencer, Herbert. 1867. The Principles of Biology. Vol. 2. New York: Appleton.
 

Spencer, Herbert. 1910. The Study of Sociology. New York: D. Appleton and Co.
 

Spencer, Herbert. 1969. Principles of Sociology. [Edited by Stanislav Andreski.] Hamden, CT: Archon Books.
 

"The Telltale Gene." July 1990. Consumer Reports 55: 483-88.
 

Walton, John. 1970. "A systematic survey of community power research." Pp. 443-64 in Michael Aiken and Paul E. Mott, editors, The Structure of Community Power. New York: Random House.
 

Zea, Leopoldo. 1968. El positivismo en Mexico: nacimiento, apogeo, y decadencia. Mexico: Fondo de cultura economica.
 

Zea, Leopoldo. 1974. Positivism in Mexico. Austin: University of Texas Press.
 

Chapter 4

SOCIAL SCIENCE SYNTACTICS AND EMPIRICS:

ALLEGED PROBLEMS ...
 
 
 
 
 
 
 
 
 
 
 

This chapter is a detailed critique of three published articles. What these articles have in common is the fact that they express serious misgivings about quantitative structuralism and its positivist underpinnings. In my view these papers also share a large problem: Most of their claims are untenable.
 

(1) Three papers in search of an argument
 

Let me begin by saying that I do not share Andrew Abbott's (1988:169) pessimism: I do not believe that there is a "growing chasm" between sociological theory and sociological research despite the fact that a few scholars, such as Sica (1988:20,40,78-9, 94-5,118, 136-41,259), grasp at every opportunity to show that the contemporary program of social theory is badly hampered by such alleged inanities as empirical research and statistical analysis. The traditional rift, if anything, has become smaller in recent decades. It has become smaller, I believe, because more and more social scientists have come to the realization that those who are weak in methodology cannot fully comprehend sociological theory, that those who achieve little understanding of theory have equally limited prospects in the methodological realm, and that the most precious intellectual abilities have to do with "translating models" (Moreno and Glassner, 1982) that may seem to involve incompatible ways of thinking, but which in reality do not. Abbott claims support in the writings of Peter Abell (1987), but Abell's method of "comparative narratives," beginning with the type of idiographic story analysis that Abbott finds so alluring, ends by combining story analysis with traditional "variable-centred" modes of social research. Abell believes that theory and research should be complementary. He also believes that we should take advantage of "triangulation," the use of multiple methodologies--a theme to which I shall return.

The relationship between theory and method, then, involves precisely the sort of reciprocal causation that, ironically, will often become a matter of controversy in the pages that follow. If we could somehow make this nexus serve us more effectively, we could rise above the theory-methods rift by creating a magnificent structure, a structure that would be as wondrous, imposing, and unforgettable as the huge arch bridge that spans the New River Gorge of West Virginia. As we begin building the foundations, we must ask the likes of Stephen Turner, Andrew Abbott, and Randall Collins to cease and desist from the practice of placing dynamite charges deep within the escarpments of granite.
 

(1.1) Turner (1987): The causal conundrum
 

Turner's article, I believe, is wrong in its basic thesis despite the fact that it makes some very clever points. The central thesis is that theory is "underdetermined" by data, and that statistically-oriented social scientists and our many mentors among professional statisticians do not realize this. I disagree with the argument for several reasons: First, any science progresses insofar as it is able to reduce (not eliminate, as Turner implies) the underdetermined character of theories, and the social sciences have made considerable progress in this regard. Second, the most capable statisticians--in particular Blalock, who is used as a whipping boy by both Turner and Abbott (1988)--are deeply aware of the problem of underdetermination and constantly warn that statistical procedures cannot substitute for other elements of the theoretical enterprise. Third--and most importantly--Turner opens his article by referring to an alleged lack of "progress" in sociological theory. Yet after he arrives at what he considers to be an adequate diagnosis of our problems--the fundamental tension between social theory and statistical methodology--we are left without any suggestion as to what sort of theory would be appropriate for the contemporary social sciences, what sort of theory would be capable of meaningful progress, or what sort of theory would escape the various pitfalls of underdetermination as Turner understands them. Turner may share Jeffrey Alexander's belief that there is little difference between the Department of Sociology and the Department of English; indeed, he may even believe that the various fields of the natural sciences have so much in common, and that the various social science/humanities fields have so many features that distinguish them from the natural sciences, that the universities should reorganize themselves immediately into two large departments: The Department of Natural Sciences and the Department of Natural Languages--the latter phrase is an artificial intelligence usage.
 

i
 

I am pleased that Turner mentions Blalock as one of the major malefactors of the 'sixties and 'seventies, because Blalock's popular undergraduate statistics textbook (1979) provides a perfect illustration of my second claim above. As the 1979 edition of this text was being prepared, McGraw-Hill asked me (with no assurance of or request for anonymity) to evaluate the manuscript; there were about a dozen places where I had a discernable impact on the final text. In several of these instances I merely asked Blalock to illustrate the underdetermination thesis more compellingly. The best example is found on page 470, where it is stated that in a simple three-variable causation problem, there are 64 (4 to the 3rd, not 3 to the 4th, as the typo says) different causal diagrams that one could propose after making simplifying assumptions about linearity of relationships. Allowing nonlinearity, the possibilities become indefinitely large. But note that Blalock presents only ten of these models on the following page, and then discusses how one might eliminate various possibilities from this already short list.

The remaining 54 models are not simply ignored, as Turner often implies. An attempt is made to eliminate many of them through statistical tests, and many are eliminated through triangulation, i.e., through information obtained outside the framework of a given study. Although Turner does not seem to subscribe to Collins' mistaken conviction (1984:336) that one cannot use statistics to test the appropriateness of statistical procedures, he makes many passing references to these tests without telling us, for instance, that the rules governing the transition from ordinary least squares (OLS) to generalized least squares (GLS) have been among the more impressive accomplishments of mathematical statistics. I presume that it is these rules--the Gauss-Markov principles (Hanushek and Jackson, 1977: Chapter 5)--that Turner has in mind when he says (1987:175) that "with different and equally plausible assumptions, one will derive different estimates." Or when he alludes (1987:175) to "... the kinds of assumptions needed to make causal inferences ..." Or when he refers (1987:177) to "the problem of the status of assumptions about correlated error terms ..." Or when he says (1987:180) that "... mathematical results, such as multiple correlations, can be fixed only by making additional kinds of assumptions" for which "the values are not arbitrary, but they are determinate only under assumptions for which there are ... valid alternatives." Or, finally, when he warns us (1987:182) that there is no simple meaning to the ceteris paribus clause, "... for there are many plausible assumptions about 'extraneous variables' which might be made ..."

Turner creates a world of chaos, a blooming, buzzing confusion, a place of darkness without form and void, a place where only the null hypothesis has a right to exist. But in fact the Gauss-Markov postulates and other comparable postulates enable us to eliminate large families of plausible models describing, say, the causal dynamics among a given set of variables. Turner himself shows awareness of the power of this selectivity: In his discussion of the problem of linearity and departures therefrom (1987:176), he concedes that "... classes of alternative hypotheses which could not be readily excluded now can, and it is consequently more difficult to devise plausible counterinterpretations to established findings."

When one realizes that ways of categorically rejecting families of hypotheses have been invented and improved upon many times in the history of statistics, one also realizes that Turner, in the remark above, is on the verge of abandoning his central argument. When he tells us (1987:176) that "... these [selective] devices increase the cognitive weight of the hypotheses by reducing the weight of other alternatives," he presages an argument that will be developed below, by Collins: The argument that the null hypothesis is a weighty and meaningful theoretical statement. Turner soon regresses, however, describing situations in which, allegedly, there is not "... any sort of decisive victory for one model or another" (1987:178); this, despite the fact that many such victories are reported routinely in the research literature. Although the use of statistical controls is a standard means of eliminating large numbers of alternative hypotheses, controlled scores are disparaged by Turner as "fictional" (1987:181). But controlled scores are no more fictional than non-controlled scores. If I say that "my prestige score is 60, and this score is 10 points higher than the average prestige score for those who have the same level of education I have," why is the 10 less real than the 60? As a matter of fact, since the 10 can be interpreted in relation to an expected ("controlled") value for those at a given level of education, it strikes me that the 10 is less fictional than the 60. When we see that ideas in science are "fictionalized" (and de-fictionalized) to various degrees, we perceive the error of Agger's (1989:21) either-or decision to "... abandon the distinction between science and fiction ..." Fictionalization is a variable.

In each of these instances Turner speaks of alternative methods of estimation as if these methods were selected arbitrarily. They are not. One should apply the appropriate rules--e.g., the Gauss-Markov postulates--before one makes a choice. Some choices are better than others--this is known through mathematics, through simulations, and through various kinds of practical experience. On the other hand, it is also known that statistical methods have a surprisingly high degree of robustness, so that even when researchers select the wrong model--say, OLS when they should use GLS corrections--they tend to end up with a model that has a good prospect of revealing something about social realities. Taylor (1987) provides a fine example: Her argument teaches us something about status attainment in socialist nations despite the fact that her dependent variable violates Gauss-Markov. One day, somebody will test her most interesting hypotheses using a dependent variable that does not commit the indiscretion of violating Gauss-Markov.

Aside from the Gauss-Markov postulates there are many other properties, inherent in statistical procedures, that protect us from a fatal underdetermination of theories. For instance, although Turner mentions (1987:178) that measurement error alone may make it impossible for us to avoid a "... large element of 'unexplained' variation in most empirical social science models ...," he neglects to inform us that (1) biased errors of measurement that merely add constants to a series of scores will have no impact on the correlational structure among variables; (2) random errors throughout the measurement process will tend to reduce the level of all correlations among several variables, thereby increasing the likelihood that the surviving model will be our most conservative, low-risk "theory"--the null hypothesis; (3) indices are often constructed in such a way that systematic biases cannot distort them at all, as, for instance, when we define the crude birth rate as total births divided by total population: An underenumeration involving, say, ten percentage points for each variable will introduce no distortion into the resulting rate.

These controls and cross checks have always been the state of the arts in statistics, far more than in verbal formulations. Along with the fail-safe devices just enumerated, it should be noted that the null hypothesis, which receives considerable deference from Collins (see below) as a theoretical formulation, is in the final analysis a way of protecting ourselves from producing large numbers of underdetermined theories--underdetermined in the sense of arising solely through type I error. A powerful convention insists that we maintain a low risk of type I error, and suggestions have been made that a similar convention be established for the complementary error, called type II error, in which we fail to discern patterns, structures, or regularities in the social realm that have sufficient importance that they should not be overlooked (Faia, 1966; Cohen, 1982; Moore and Gledhill, 1988).

To his credit, Blalock proceeds in his statistics textbook and elsewhere (Blalock, 1989) as if we were trying to eliminate models that must be falsifiable, not as if we were trying to prove some model beyond a reasonable doubt; his attitude is strictly Popperian. This is a distinction of fundamental importance, and Turner tends to obscure it. I think it is plainly incorrect to assert, as Turner does, that statisticians of Blalock's ilk are out to "prove causation." I could provide any number of illustrations of ways in which good statisticians maintain the appropriate scientific skepticism, seeking to eliminate hypotheses rather than trying to corroborate some preconceived point of view. I doubt that Turner can name a serious sociologist who believes that "... at the end of inquiry ... data would determine a unique theoretical result" (1987:173).

In my experience, few social scientists have taken such a naive view of the nature of theory construction, and it therefore strikes me as a classic loaded question for Turner to ask (1987:173) "why have the ... successes of statistical methods in 'bringing evidence to bear on theoretically important ideas' ... been so modest?" He seems to assume that the frustrating search for a unique theoretical result "... is not ordinarily a practical problem in the natural sciences" (1987:173), but just a few sentences later he refers to Pearson's claim that physical science theories--for instance, the kinetic theory of gases, so important in the development of statistics--are little more than "idealizations," a label that surely implies the same level of underdetermination that is supposed to characterize the social sciences. I hope that the second of these claims is the closer to the truth, because it seems to me that if natural scientists have learned to live with underdetermination, then social scientists should be able to do so.
 

ii
 

At several places Turner (1987:173-174,178) refers to scholars who feel that we ought to try to get our coefficients of determination--"explained variance"--close to 100 per cent, but I do not believe that there are many social scientists who entertain this aspiration or who are terribly frustrated when we fail to attain it. "Saturated models" used in loglinear analysis "explain" all variance, but Halli and Rao (1992:120,128) regard such models as "exploratory" rather than "explanatory," and suggest "... that any log-linear analysis begin with an examination of the saturated model," in order to obtain "... a good idea of which interactions are good candidates for deletion in a more parsimonious treatment of the data." I find myself more likely to be irritated by research workers who seem to be satisfied if their theoretical model explains, say, 15 per cent of the variance in some social phenomenon. On the other hand, a student and I just completed an MA project on fertility in which we account for about 25 per cent of the variance in our dependent variable, and I believe that this work has considerable theoretical import partly because it makes use of elegant multivariate methods. This study provided a classic instance of tension between statistical significance and "substantive" significance, and we opted for the latter. The playing down of statistical significance was the major philosophical outcome of the strident "significance test controversy" of the 'fifties and 'sixties (Morrison and Henkel, 1970), and I believe that this outcome undercuts Turner's basic argument and perhaps Collins' argument also: It teaches us that statistical outcomes should not be taken as absolute, and ordinarily have not been so taken.¹

Incidentally, we happen to have many models in which we do "explain" essentially all the variance (Faia, 1989). Simulation programs, nowadays widely available, provide an excellent example. Elaborate simulation packages such as the Florida Interactive Modeler (FIRM) make possible a clearer understanding of the implications of our hypotheses about, say, the factors that influence various social attitudes. A more compelling example would be Bongaarts and Potter's empirically grounded studies (1983) showing that fertility trends and differentials can be decomposed into their components, with these components being selected from the standard Davis-Blake (1956) "intermediate" determinants of fertility. Bongaarts and Potter are creating very good theory in this instance, and their efforts are informed by an excellent, widely used taxonomy (Davis-Blake) as well as considerable statistical acumen. There are many edifying examples of the use of deterministic models in the social science disciplines, including the determination of size, distribution, and composition of populations by fertility, mortality, migration, and vertical mobility, and the determination of the sex ratio at birth by parity level, the sex composition of children already born, and various methods for implementing boy/girl preferences on the part of parents. In addition, it has been recognized for many years that when we aggregate data on social behavior into rates, ratios, proportions, etc., we maximize our prospects of explaining variance.

Turner (1987:174) worries unduly about the fact that when we add new variables to a regression equation we often fail to explain any substantial amount of additional variance. To me, this alleged defect has its advantages, especially if the original equation seems to be doing a reasonably good job: What Turner calls "redundancy" is one of the few ways in which nature encourages us to be satisfied with parsimony. Beyond that, it may well be that the best discussions of redundant variables have come from sophisticated methodologists addressing the fact that measures of social mobility, status discrepancy, cohort effects, etc., typically contain no information beyond that contained in their components. Vast amounts of verbal speculation have been devastated by the simple algebraic demonstration of this redundancy, or by the fact that our computers immediately send out error messages when we try to tell them, for instance, that prejudice may be influenced by parental socioeconomic status (SES), filial SES, and by social mobility defined as the difference between filial and parental SES. None of the three causal variables, taken alone, contains any information not contained in the other two.
 

iii
 

Turner (1987:177,180-82) cites an earlier paper by himself and Wilcox (1974) that deals with the idea of axiomatic theory as developed by Costner and Leik (1964), Zetterberg (1965), and others. In essence, the argument of these theories is that if A is related to B, and B to C, we can deduce a relationship between A and C. This is proposed as a straightforward use of the transitivity assumption or the pure conditional argument, but it must be taken with the proviso that any qualifications that appear in the premises of a syllogism must be retained in the conclusion. In other words, if the premises involve probable relationships or weak relationships, then the conclusion must retain these qualifications. In a study involving data from Textor (1967), to be summarized in a moment, I found that this argument, the transitivity argument, has considerable utility. But Turner's basic claim (1987:180) is that there is no logical way of testing the propositions of axiomatic theories by means of correlation and regression analysis.

The rationale is found in Turner and Wilcox (1974), a fascinating paper that provides an opportunity for me to demonstrate the essence of my argument against Turner. In presenting their "A Calculus," as opposed to more complex B and C calculi, they enumerate (in their own notation) several syllogisms (1974:574) that represent Zetterberg-type arguments. One of these arguments, for instance, may be translated as follows:
 

They then point out that, because the qualifications mentioned above are not made explicit, this argument does not necessarily lead to a true conclusion. If, however, we allow the syllogism to take the form of the modus tollens, a form in which the negation of a conclusion leads us to deny at least one of the premises thought to generate it, then we create a situation in which premises are subject to being eliminated, to being falsified. The test involves the valid argument known as denying the consequent rather than the invalid argument of asserting the consequent. If, in the Turner-Wilcox example, we find through observation that x is not positively correlated with w, we reject the premise that x is related to y and the premise that y is related to w. The fact that we do not retain any qualifications--that is, we do not acknowledge that it is logically possible for these premises to be true despite the fact that the conclusion is false--is non-problematical: This is a highly conservative decision rule, comparable to our having a strong bias toward the null hypothesis.

A work by Textor (1967) provides many opportunities to construct syllogisms of the form described above. Based on the Human Relations Area Files and written primarily by computer, this large tome defines nearly 500 variables ("finished characteristics" or FC's), classifies 400 societies according to each of the variables for which ethnographic data have been provided, and then presents a computer printout that shows how the variables are interrelated. The format is simple: Each variable is a dichotomy, and all tables are bivariate two-by-two crosstabulations. Below each table is a chi-square value or a Fisher exact test, a phi coefficient provided as a measure of strength of relationship, and the level of significance. A simple, computer-written paragraph describes the relationship. In order to appear in the printout, tables had to attain significance at the .10 level or better.

Under my sampling protocol, I ended up with 155 syllogisms that had the Zetterberg-Turner-Wilcox form. For example:
 

The code numbers 44, 36, and 51 are Textor's identifiers for a set of fixed characteristics, the signs between the FC's indicate whether a relationship is positive or negative, and the line segment should be read as "therefore." Using the slightly artificial language of Textor's computer-written propositions, in which verbs such as "tend" or "tilt" reflect the level of statistical significance, the syllogism would be described as follows:
 

Already, then, we have a nicely structured theory linking three ecological features; our next task would be to search for the "intervening variables" that encouraged early agriculturalists to remain sedentary (Lenski, Lenski, and Nolan, 1991:158-60). The syllogism representing the theory contains a conclusion with a phi coefficient that is both high and in the anticipated direction, and we therefore conclude that the premises have not been falsified by the many field studies that contributed to the Textor dataset.

Among the 155 syllogisms constructed from Textor's data, 117 or 75 per cent produced truthful conclusions. The 38 negative instances occurred either because conclusions were false or because the number of societies classified on both of the variables contained in conclusions fell below 100, which I used as an arbitrary minimal N. What, then, is the probability of producing a truthful conclusion when at least one premise is false? Fortunately Textor (1967:54-60) answers this question by providing the cleverly conceived "whiskers" test. The whiskers FC's are ten random dichotomous variables that assign whiskers of various colors to the men of varying numbers of societies. As random variates, these FC's should be related to other FC's only by chance, by the occurrence of type I error. Because there are ten whiskers FC's, and because each of these variables could potentially be related to about 250 other FC's used as predicates, there are potentially about 2500 tables to be printed out for the whiskers variables. However, only 126 significant relationships actually appeared, about 5 per cent of the potential number. Therefore, the likelihood of deriving a truthful deduction from demonstrably false premises is probably less than .05. These findings seem to vindicate the axiomatic format as a way of generating theories, and to demonstrate that the falsifiability or "bootstrap" criterion (Glymour, 1980:291-321) is an appropriate strategy for testing this type of theory.

The falsifiability criterion is widely applicable, and it is not limited to the type of generalization developed by Turner, Wilcox, and Zetterberg. I shall argue in a later chapter that in developing and testing theories it would be advantageous to make use of taxonomies with hierarchic properties (Faia, 1991): These hierarchies provide additional opportunities for falsification, as we shall see.
 

iv
 

I cannot agree with Turner's recurring suggestion that the more highly developed sciences, due to the nature of their subject matter, are free of the ambiguities of statistical interpretation with which social scientists must struggle. For instance, geology and its subdivisions such as seismology and vulcanology, because of their deep dependence on field observations, can hardly avoid a serious risk of underdetermination of theories. When animal ecologists use ANOVA and related multivariate designs to test the notion that rapid, "explosive" evolutionary changes may be brought about by positive feedback--i.e., reciprocal causation--among (1) sexual ornamentation, (2) mate selection favoring the genes that make for ornamentation, and (3) mate selection favoring the genes for preferring certain kinds of ornamentation (Dawkins, 1986: Chapter 8), we encounter the full range of causal dilemmas that, following Turner, one might consider unique to the social sciences. In this instance, then, social scientists are not at a disadvantage. And on a related matter, animal ecologists have little prospect of obtaining random samples that approach the quality of those designed by, say, the U.S. Census Bureau for the current population surveys (CPS). Nor do animals and plants seem to encourage the emic option, always available to social scientists.

In a personal letter, Turner refers to the difficulty we have in convincing "students or anyone else outside sociology that sociology has anything to say." I find this claim bothersome because, although I believe that we have an etic discipline that should not devote excessive attention to the gathering of emic data, I nevertheless realize that we must impress a broader audience than just ourselves. Physicists couldn't care less whether the public at large reads Physical Review Letters, but they have all sorts of Carl Sagans, Freeman Dysons, and Steven Hawkings in the field busily translating findings for popular consumption, in ways that are most impressive. I believe that we in the social sciences must do this sort of translation, or we may not survive. It is encouraging that Footnotes, the newsletter of the American Sociological Association, now shows a little deference to sociologists who make contributions to the mass media, but this effort will hardly suffice. We should do infinitely more.
 

(1.2) Abbott (1988): The transcendence of triangulation
 

Abbott has a way of pulling punches. His concessions are so large and so frequent that one cannot help but doubt his central arguments. He says, for instance (1988:170), that it is commonly assumed by statistical workers that causal relationships do not vary by social context, but he is quick to point out that "there are, of course, ways of relaxing some of these assumptions." Indeed there are! For many years the literature on contextual analysis has been large enough to support introductory textbooks (Boyd and Iversen, 1979; Iversen, 1991). I suspect that researchers who understand contextual analysis in the Boyd-Iversen sense are better prepared than most of us to do the sorts of causal analysis suggested by Abbott. Again, while the basic intention of Abbott's article is to persuade the reader that the general linear model (GLM), along with the corresponding delusional system that he calls "general linear reality" (GLR), have achieved few if any successes, he says early on (1988:171) that he intends no "derogation" of the "very great successes" of the GLM. In truth, Abbott's article is highly derogatory in that it says next to nothing about ways in which the more subtle uses of the GLM--those suggested by Boyd and Iversen, for instance--might help us to close (or avoid) the "growing chasm" that constitutes the opening allegation. To Abbott, the GLM has few if any redeeming virtues. It is most extraordinary that he believes that his aim in making these arguments "... is not controversial" (1988:183).

I cannot, for instance, accept the suggestion (1988:175) that the GLM somehow precludes our hypothesizing that a given causal agent may have multiple effects. I do not believe (1988:173) that those who apply the GLM are so committed to "monotonic causal flow" that they automatically acquiesce in the notion that "a given cause is equally relevant at all times ..."; I've never encountered a sociologist this naive. I cannot agree (1988:173) that the GLM assumes that "cause can never flow from small to large, from the arbitrary [?] to the general, from the minor event to the major development." I do not recall (1988:179) having read anything by or about Max Weber that suggests that worries about multicollinearity among variables will preclude our appreciating Weber's elective affinities or his ideal types, or will make it difficult for us to address the issue of "elitists against pluralists." Finally, I cannot accept the argument (1988:172) that those who study a profession such as accounting by means of the GLM do not realize that, over the last several decades, many such occupations have been "reshaped by technology," by competition, or by other factors.
 

i
 

On the question of multiple effects, all one must do to undermine Abbott's argument is to locate a few classic path diagrams--the "weapon on the wall" of the GLM--and show that these diagrams have at least a few causal factors that have more than a single arrow flowing from them to some effect. Take, for instance, an influential paper by Duncan (1966) that introduces the basic logic of path analysis. Duncan's Figure 1, a path diagram, proposes that social background may influence ambition, class values, and intelligence; that ambition, in turn, may influence class values and intelligence; and that the socio-economic rating of schools may influence ambition, class values, and intelligence. Figure 2(b) and Figure 4 present the same sorts of causal hypotheses, replete with multiple outcomes. Similarly, Alwin and Hauser (1975:38) present a causal diagram in which a man's education is thought to influence both his occupational prestige and his income--a multiple-effects hypothesis that has been tested scores of times in the status-attainment literature. The major purpose of the Alwin-Hauser article, in fact, is to explain how to separate direct effects, indirect effects, spuriousness, and other distortions in complex causal processes in which there are many variables with multiple effects or "sequence effects" (Abbott, 1988:177). In brief, then, those who construct analyses based on the GLM (and, presumably, its associated GLR) do contemplate the possibility that a given causal agent may have a multiplicity of subtle effects. Users of the GLM/GLR occasionally may miss these subtleties, but if they spend as much time worrying about making proper "specifications" (Boyd and Iversen, 1979: Chapter 2) as they do about multicollinearity, they may have a better eye for subtleties than those who do not have the benefit of the GLM/GLR worldview. The presumption that this worldview creates a general lack of insight must give way, at least for the moment, to the null hypothesis.

Perhaps, however, Abbott is addressing another type of multiple-outcomes situation. I recently came across fascinating findings (James, 1987a, 1987b) regarding social class variations in the sex ratio of children at birth. At present these variations are small, but at some future time they may become much larger due to the introduction of cheap, safe, and effective methods of influencing the sex of embryos prior to conception. If there is a strong preference for male offspring within, say, the lower strata of a given society (Williamson, 1976:54-65), one would expect that the introduction of sex-preselection technology would tend to raise the sex ratio at birth (SRB) within these strata, thereby creating an inverse relationship between SES and the SRB. On the other hand, if there is a tendency for technological innovations to be adopted earliest within the higher socioeconomic strata, and assuming that the higher strata have at least a moderate preference for boys, we would anticipate that for some time this same technology would tend to raise the SRB within the higher strata. In one instance, then, the SES SRB causal arrow would seem to carry a negative sign; in the other instance, a positive sign. How do we handle this kind of situation?

For one thing, it is clear again that those who have a commitment to the use of the GLM are not at all handicapped in approaching this type of problem. In the case at hand, the typical procedure would be to find an index of sex preference and an index of willingness to adopt innovations, declare these indexes--in classic Lazarsfeldian style--to be intervening variables between SES and the SRB, and then decompose the SES-to-SRB relationship into these two elements. I do not see how this kind of problem could be handled any more deftly by those who insulate themselves against the GLM, by those theorists who for inexplicable reasons "may reject" (Abbott, 1988:176) an emphasis on both antecedent and intervening variables. Abbott would probably argue, as we shall see, that it is important to tell detailed idiographic "stories" about how particular couples would decide the sex of a child, how they would decide about whether to use fertility-related innovations, and how they would decide several other such questions. I agree strongly. But once again I fail to see how a commitment to the GLM/GLR worldview precludes an interest in these kinds of stories or undermines our ability to tell them well. At one point (1988:177), in fact, Abbott credits GLM practitioners with telling a good causal story based on a proposed path diagram; clearly, these stories are a wonderful source of new specifications. Ward's (1986) book on the American railroads presents a fascinating collection of such stories, which we examine in a later chapter.

There is a generic multiple-outcomes story that occurs in several academic fields, including sociology, economics, and psychology. This story assumes that various kinds of needs, problems, or disequilibria may be so compelling that there are many different means of satisfying them, and that the denial of one set of means will usually increase the likelihood that an alternative set of means will appear. The title of this story is "substitutability," and its themes have been captured in many renditions by GLM practitioners and by others. In abnormal psychology, for instance, it is often assumed that one set of psychiatric symptoms, or a given adjustment technique (direct or indirect aggression, identification, projection, compensation, etc.), may be substituted for another. In economics, one might entertain the hypothesis that, in a developing nation such as Mexico, meat will be substituted for other foods when the opportunity arises, or that poultry will be substituted for beef among well educated consumers (Contreras and Cifuentes, 1987: Table 5). In sociology, one might suggest that the denial of access to one type of weapon will create a strong demand for other types of weapons (Wright et al., 1983). In all these instances we are in the presence of multiple outcomes, with "ends" presumed to give rise to "means" not through teleology but through causation. And because it is assumed that these multiple means, whether substituted for one another or not, may have an impact on the need, problem, or disequilibrium that generated them in the original instance, we are also in the presence of functionalist hypotheses. The GLM, of course, would be applicable to all these examples.
 

ii
 

Abbott argues that the GLM makes it difficult for social scientists to evaluate a given causal relationship within several contexts, or to entertain the possibility that causation may flow from "small" events to "large" events. But one cannot reconcile these claims with Boyd and Iversen's demonstration (1979) that the GLM helps us to explain variations in individual behavior from one structural context to another, or their demonstration that we may also reverse this process by "explaining aggregate relationships in terms of individual behavior" (Boyd and Iversen, 1979: Chapter 2, part III). Once again, I do not see anything about the GLM that reduces the likelihood of our perceiving these opportunities, and in fact it appears to me that careful study of the Boyd-Iversen texts (and related literature) would enable us to exploit the GLM effectively to facilitate the perception of important contextual effects or important causal flows from the microsocial level to the macrosocial level. Statistical models, in other words, help us to have these important theoretical perceptions.²

Boyd and Iversen, then, make it clear that causation may flow from small causes to large effects. One easily finds scores of illustrations, and one could readily invoke the GLM to study most of them. When tens of millions of individuals release small amounts of chlorofluorocarbons (CFC's) into the atmosphere, these are small events that may have large effects on global ecology (Schneider and Londer, 1984:197). When President Ford made a "misstatement" about Poland during one of his debates with Jimmy Carter, this was a small cause that had a large impact. The meetings between Carter, Begin, and Sadat at Camp David were a small event that had large consequences (Carter, 1982:319-403), and the same may be said for the meetings between Truman and MacArthur during the Korean war (Spanier, 1965). Michel Foucault had little enthusiasm for the idea that Martin Luther's "anal retentiveness" explains the Reformation (Cousins and Hussain, 1984:64), but Isaac Asimov's brand of psychohistory (Kanfer, 1988:82) has been taken seriously enough by scholars that there is at least one academic journal devoted to the Great Man and the Not-So-Great Man theory of leadership and social change; surely, the psychohistorical orientation involves small causes and large effects. Social scientists interested in what Abbott calls "elitists against pluralists" generally regard elite networks as a servomechanism in which small amounts of energy are used to control large amounts of energy--this, to me, is a good definition of elitist power. Large, continent-wide systems of production, such as Anheuser-Busch breweries, are controlled by just such servomechanistic elites. Finally, although "catastrophe" (or "chaos") theorists tend to focus on large events, even they may concede that "thou canst not stir a flower without troubling of a star" (Woodcock and Davis, 1978:67).³

In essence, Abbott accuses GLM devotees of having a bias against reductionism, i.e., against the strategy of explaining large phenomena with reference to small phenomena. On the contrary, it may well be that our established predilections have made us rely too heavily on reductionism, and that therefore any tendency for the GLM to sustain the opposite bias may serve as a necessary corrective. Again, however, I see nothing inherent in the GLM that would preclude our testing reductionist models adequately, should we choose to do so. Small insects and small rodents, for instance, do not look like major historical actors capable of troubling stars, but Zinsser (1935) argues that these tiny vectors indeed may have a stupendous impact, influencing the course of important events such as wars. Or, consider AIDS. Studies of the micro-interaction involved in perpetuating the AIDS epidemic (a macrosocial event) clearly involve small causes combined with large effects. I suspect that the GLM is the modal method in studies currently underway in this area.
 

iii
 

The essential problem with Abbott, however, has little to do with causation: The problem is that he has a philosophical commitment to idiography, to what he calls storytelling. He develops a pair of stories about Kenya and Nigeria (1988:181), and quickly brings us the distressing news that it is virtually impossible to generalize about the developmental dynamics of these two nations. All we can do in attempting to understand social development across Africa is to tell "... thirty particular stories" in detail. I certainly do agree with Abbott that if we can do nothing more than tell these thirty stories, if we cannot generalize about Kenya and Nigeria and other nations with comparable characteristics, then indeed we do not have much of a prospect of creating a science about social development or about any other aspect of social systems. However, I do not agree with Abbott's premise. I don't think he agrees with it either, in the sense that he could probably not tell Kenya/Nigeria stories for any length of time without proposing some sort of generalization.

I don't believe that one can tell detailed sociological stories without generalizing, any more than one can think randomly: My students are forever confident that they can readily conjure up a set of random numbers from their heads, and they invariably lose (i.e., they fail to win) a good three-dollar cigar when I show them that their proposed numbers depart significantly from randomness. (Once on a fluke I lost the bet, and I had to suffer the indignity of trying to cut their cigar into 34 segments of random size, using a straight-edge.) Abbott would have a chance to get his storytelling program under way by one means only: He would have to devote himself to poetry. He would have to thrust himself into the type of idiographic exploration defined by Jacques Barzun in his critique of modern science (1964:198): Law and poetry, says Barzun, are highly exact because they are highly idiographic, laying out a particular case in the finest detail. Science, in contrast, is very inexact because it is extremely abstract much of the time. The major trouble with science, says Barzun (1964:202), is that the scientist often cannot see the trees for the forest. A scientist with this affliction, he continues, is likely to try something like scientizing Tennyson (1964:129). Or simonizing her grandma.

Incidentally, although I do not propose to get into a detailed argument with Barzun, my instincts tell me that poets and legal scholars, despite their passion for idiographies, do not work for long without making generalizations. Nor did Michel Foucault, despite his insisting to his students that "general psychology, like anything general, does not exist" (Eribon, 1991:139.)

Mass murderers are fascinating, and Newton's Mass Murder (1988) is filled with the details of their hideous lives; but to a sociologist the best part of the book is the place where Newton generalizes about research into the kinds of social conditions that create these lives (1988:3-21). Again, college novels tell fascinating stories (Kissiah, 1969); but the most interesting features of these stories have to do with their ability to say what is right and what is wrong with higher education in the United States. The apparent suicide of Marilyn Monroe makes a gripping story; but far more interesting than Monroe's life and career is the prospect that mass media reports of her suicide may have inspired imitative suicides--think of Tarde's and Le Bon's principles of crowd contagion--across the nation, especially in places where media coverage was highly detailed (Phillips, 1979). LeMasters' (1975) intensive account of the social life of a working-class tavern is a fascinating case study; yet, virtually every chapter of this work begins with a set of generalizations drawn from the research literature, and the work is enriched by the way it deepens our understanding of these generalizations. In each of these instances the stories are a means of arriving at better generalizations. The stories are essential, they may be excellent poetry or excellent legal briefs or both, but they alone cannot constitute a social science. At best, they lend themselves to Abell's method of comparative narratives. And in this case, best can mean very, very good.⁴
 

iv
 

In large measure Abbott criticizes the GLM unfairly, by invoking the law of plenitude. As the philosophical basis of scientific skepticism, this law states that the subtlety and complexity of nature are infinite, and that therefore nature tends to flamboozle scientists most of the time. But the law of plenitude applies to every scientific method in the same way that it applies to the GLM. Therefore, showing that the GLM is subject to the law of plenitude is not a critique of the GLM; rather, it is a simple reaffirmation of the principle of plenitude. When Abbott (1988:170) tells us that we cannot trust our momentary, untested intuitions about "causal time," we must immediately concede the point and seek alternative methods that would allow us to approach issues of causal lags and periodicities more effectively. When we use several approaches to the same question we take advantage of Percy Bridgman's admonition about science as a process of "doing one's damnedest with one's mind, no holds barred." Thus, Phillips (1979) not only uses empirical data and the GLM to test the notion that imitative suicides are induced over a matter of days, or hours; he also invokes the theories of Gabriel Tarde in order to strengthen the argument that, if one does not have daily rates for these kinds of suicides, one probably cannot conduct the study at all.

Gottman (1981), in an excellent treatise on time-series analysis, tells us that unless we experiment with different assumptions about the timing of cause-effect relationships, we risk arriving at conclusions that are precisely the opposite of the truth. His best illustration has to do with the way wagon wheels in motion pictures often appear to be rotating in reverse: The "methodology"--in this instance a camera with a particular shutter speed--relates to the rotation of the wagon wheel in a way that grossly misleads us. But all we have to do in order to overcome this distortion is to run a series of experiments, using variable shutter speeds. When we combine intuition, theories, multiple observations, data reduction, and instrument experiments, we are making use of triangulation. We are also conducting an analysis--albeit a statistical analysis--to which Agger (1989:90) contributes precisely nothing with the remark that the trouble with suicide research is that it accepts the Durkheimian presupposition "... of a world that makes people want to kill themselves." Phillips shows us that, on some happy days, hardly anybody wishes to kill himself.

Triangulation, in this sense, is essential for a seismologist trying to locate the epicenter of an earthquake, and it is a matter of routine for a good biologist. Dawkins' treatise on Darwinism (1986:99-100) provides illustrations as compelling as those of Phillips and Gottman. He asks, for instance, why it is that some types of cicadas have a lengthy 13-year cycle, while other types have an even longer 17-year cycle. His tentative hypothesis: These two numbers are primes, not divisible by any number other than unity and themselves; therefore, predators are not likely to have cycles that correspond to cicada rhythms, and cicada periodicity may serve to protect this organism from potential predators. Using a similar theory-based triangulation, a social scientist might propose that Presidential cycles of four years, combined with Congressional cycles of two years, create a situation in which Congressmen are slightly more likely to be beaten than cicadas are to be eaten (Tufte, 1974).

It turns out, however, that not all social scientists share my admiration for Dawkins' clever use of numbers, prime or otherwise.
 

(1.3) Collins (1984): The contours of quantiphobia
 

--Morris Shamos (1990:92)
 

Collins (1985:41) makes critical comments about Adolphe Quetelet, a scholar of renown and of central importance in the history of social statistics (Porter, 1986). He claims that because Quetelet's statistical laws involve no more than "a few simple probabilities" and a reasonable prospect of predicting "rates of population change or crime," interest in Quetelet soon "fell away as the statistics failed to live up to the claims ..." (1985:41). This alleged failure has never been demonstrated, and the claim is surely vitiated by the fact that statistics, a relatively new academic discipline, has had immense successes over the last century or so. Furthermore I suspect that this claim, because it contains an obvious contradiction, never will be demonstrated: Nothing is more delightful for a scientist than the discovery of laws that can be stated in "simple" (i.e., parsimonious) ways, or the discovery that an event can be predicted (and perhaps explained) through the application of such laws. A good scientist pursues parsimony as resolutely as she respects plenitude.

An equally untenable stand--one that goes far beyond his relatively harmless remarks about Quetelet--is taken by Collins (1984) in a paper revealingly entitled "Statistics Versus Words." Although I do not agree with most of the arguments of this paper, it makes several excellent points: First, that the null hypothesis--the idea that many social processes occur by chance, or seem to occur by chance because we do not understand them, or both--should be treated as a theory per se. The null hypothesis as theory would imply, for instance, that Wall Street is the world's foremost gambling establishment in the sense that the factors that determine outcomes seem to elude everybody except (I gather) Ivan Boesky, Michael Milken, and their ilk. Susan Strange's Casino Capitalism (1986) is an excellent introduction to this topic, and in reading this book one has a clear sense that we have learned little about global financial systems, beyond the extent of our ignorance. Second, Collins argues that there are ways of testing (or formulating) theories that do not involve the gathering of statistics; the periodic table of the elements, in Collins' example, was arrived at through "induction." Third, he points out that it is difficult to study small numbers of large organizations as compared to large numbers of small organizations, with the result that historical sociology, often dealing with small numbers of whole societies, is at risk of being idiographic.
 

i
 

Collins' ideas about the null hypothesis deserve far more attention than they have received thus far. First of all, it is clear that if the null hypothesis is actually an important theoretical formulation, then Collins' larger arguments about quantitative methods are undercut by the fact that it is difficult to make decisions about null hypotheses without understanding statistics. But Collins' argument is strengthened when one realizes that the null hypothesis, as theory, could more accurately be called the "debunking" hypothesis. As understood by Berger (1963:38-43), debunking might involve the demonstration that if the protestant ethic gives rise to the spirit of capitalism, this result is not necessarily intended (cf. Thompson, Ellis, and Wildavsky, 1988: Weber section, p. 19). In general, members of society do not anticipate future events accurately, so that when they try to explain their behavior with reference to some plausible cause-effect scenario, they are usually just winging it. They do not really understand as much as they think they do about cause-effect relationships, and it is the task of the null hypothesis--that is, the debunking process--to make this limitation clear. When the null hypothesis is a prominent theoretical construct, it becomes difficult for us to believe our own propaganda; Berger suggests (1963:41,42) that sociology flourishes when accepted interpretations of society become "shaky." When Sica (1988) urges us not to lose sight of the irrational aspects of human behavior, I believe that he is arguing essentially for the debunking function.

Wildavsky (1988), for example, argues that the process of searching for safety needs to be debunked. He believes that we try too hard to conduct our lives on a basis of "trial without error," and that this attitude so diminishes our risk-taking aptitudes that it actually becomes maladaptive, keeping us from discovering better (and safer) ways of doing things. Wildavsky believes that the usual justification for trial without error is the unthinking conviction that we have such complete a priori knowledge of what is safe and what isn't, what works and what doesn't, that we need not attempt serious experimentation. He continues (1988:57):
 

But most of the time, he fears, our reliance on conventional wisdom is so strong and uncritical that we do not correctly anticipate either costs or benefits. Therefore, the appropriate thing to do is to admit that (1) the null hypothesis prevails in most instances and (2) we discover opportunities for rejecting the null hypothesis, for learning new ways of doing things, only through serious forms of investigation involving trial and error. And it should be noted that many of the experiments he cites, such as those involving the Salk and Sabine vaccines against polio (1988:55), make use of the general linear model and other statistical paraphernalia in epidemiological and human-ecological studies, and that the attitude of "statistics versus words" would be a serious impediment to research that treats the null hypothesis as a major theoretical formulation.
 

ii
 

In contrast to the incisive discussion of the null hypothesis, Collins gets into deep trouble in other parts of his article. He claims that "... statisticians are the worst offenders in imposing a narrowly positivist orthodoxy on the field" (1984:330), but deigns not to offer a trace of evidence. The example of William H. Sewell (Collins, 1986:1341) does not convince us any more than does the unsubstantiated claim by Agger (1989:29) that the science's have an "... imperial urge to control the entire cognitive map, to banish the nonquantitative ..." And although Lazarsfeld is often numbered among the worst offenders, he seems to be vindicated by his admission that "I always believed in the interdependence of quantitative and qualitative work ..." (1972b:xvii); he admits, however, that he tended to drift toward the quantitative. In any case, it is hard to think of Lazarsfeld as a major malefactor, imposing a rancorous orthodoxy on students and colleagues alike.

Historical sociology, says Collins, "... can proceed best without inroads from the statistical side" (1984:331)--a finding that will surely come as a shock to the average cliometrician, if one may speak of the average. When he writes that statistical procedures create a bias toward "... our theory, the one that is being tested" as opposed to "chance" (1984:331), Collins is simply wrong: In fact statistical procedures, based on a "minimax" strategy that tries to hold down the magnitude of our biggest mistakes, create a strong bias in favor of the null hypothesis, in favor of chance as an explanation. When the null hypothesis itself is disparaged as "empty" (1984:335), the preceding three pages of text--the best part of the article, as I've already tried to show--are contradicted as we lose sight of the need for debunking, that is, for protecting ourselves against our dangerous tendency toward Kantian overinterpretation of social realities, for protecting ourselves against the opportunity costs that so worried Wildavsky.

In asserting that "there is no way to test a statistical model statistically; one simply demonstrates that the pattern of data is consonant with it" (1984:336), Collins leads one to suspect that (1) he does not clearly understand the reciprocal and symmetrical relationship between statistical models and data, the ways in which, to return to an earlier example, a model such as ANOVA may relate to theories such as natural selection; and (2) he hasn't given serious attention to Monte Carlo simulations, which literally do test statistical models statistically and show us, happily, that these models have considerable "robustness" (Hanushek and Jackson, 1977:77).

Robustness is a term that critics of quantitative sociology might well ponder: It means that even when the real world--that is, the plenitude of nature--forces us to violate the basic mathematical assumptions of statistical models, the models still give us reasonably accurate approximations of what is going on.

Collins claims wrongly that social mobility studies have ignored mobility that is "forced" by changes in the occupational structure (1984:346,348)--note that Abbott (1988:179-80) discusses this distinction in some detail. He then tells us that "log-linear analysis ... gives us an array of numerical coefficients that mean nothing except from the point of view of the statistical analysis proposed" (1984:357), and brings us to the consummating clincher of the argument (1984:350-53): that his perusal of assorted mathematics journals shows that mathematicians, after all, speak and write English or French or German or Swahili or some other natural language, even while doing mathematics. This finding is supposed to tell us something about the prospects of quantification in the social sciences--but, alas, the argument is a non-sequitur: Surely one may concede that natural languages will continue to have a large role in social science analysis and in mathematical analysis, without also conceding that the role of mathematics must therefore be cut back in the social sciences.

In any case, criticizing statistics by citing the "worst offenders" is a little like criticizing written English--the "words" approach--by citing the most horrendous undergraduate bluebooks.

Finally, when Collins says that "the work of Erving Goffman, for example, ... would not have been improved by carrying it out under a rigid program of statistical measurement and hypothesis testing" (1984:340) and that "... someone following such methods would not have been able to do it at all," one must reply that the work of Christopher Jencks or James S. Coleman or William H. Sewell or Fogel and Engerman, for example, would not have been improved by carrying it out under a rigid program of quantiphobia, and that a quantiphobe would not have been able to do the work at all. Again, Collins' argument is a non-sequitur, proving nothing. Actually, I believe that a good statistician has a better prospect of becoming a good Goffmanian than a quantiphobe has of becoming a good cliometrician. But the anticipated failure by either party would make our fundamental problem clear.

The fundamental problem is clarified further by Mackie (1988:4-5,10-13). If, as she says, the quantitative, "agentic," "controlled reality" style of sociology prevails among male sociologists while qualitative, "communal" approaches are more likely among female sociologists, and if it is deemed essential that the discipline move away from its allegedly male-dominated theories, methods, and social structures, we are at risk of producing a self-fulfilling prophecy: Those who wish to repudiate traditions of male dominance at a structural level by creating, for instance, a more just distribution of career opportunities and rewards in the academic world, and who also repudiate particular theories and methods thought to be associated with maleness and male dominance, may ultimately strengthen whatever small association currently exists between theories, methods, and the sex of scholars. As a consequence, increasing numbers of both male and female scholars will make the sound of one hand clapping.

In sum: What we must all strive for is an androgynous methodology.
 

(2) Solicitous gatekeepers, #2
 

Without comment:
 

...,

...
 

NOTES (Chapter 4)
 

(1) Recently Turner and Turner (1990:114-18,131-32) have claimed that the "... enormous and obscure ..." literature of the 'fifties overlooked the fact that null hypotheses were "... usually known to be literally false or highly improbable." These authors are aware of the possibility of using confidence intervals to assess the strength of statistical relationships; yet they insist that the trouble with the practice of testing hypotheses is that "... it confused, or encouraged confusion over, the relationship between 'substantive' and 'statistical' significance."

Most of the Turner/Turner polemic is highly questionable: What, for instance, makes this body of literature "obscure"? If it was usually known a priori that null hypotheses were false, has this situation corrected itself given the fact that nowadays so many null hypotheses are not rejected? If so, don't statisticians deserve some credit for correcting bad habits? If we tend to ignore the strength of relationships in instances where null hypotheses have been rejected, why does virtually every contemporary statistics textbook devote a large amount of attention to measures of strength of relationship? Again, I would argue that the major conclusion of the 'fifties debate was that statistical significance is only one means of attaining substantive significance, and I believe that anybody who re-examines this literature will agree that these authors are discussing it in an oversimplified way.
 

(2) Notice, for instance, that contextual analysis is closely related to analysis of variance (Boyd and Iversen, 1979:7-12), and that analysis of variance, as Porter says (1986:316), constitutes "a theory of heredity." Actually Porter understates the case: Analysis of variance is better understood as a theory of genetics and natural selection, as suggested by Dawkins' claim (1986:41) that "[genetic] mutation is random; natural selection is the very opposite of random." ANOVA was made for precisely this sort of reality, this sort of GLR, and it provides help of inestimable value in our efforts to perceive such realities whenever and wherever they occur.
 

(3) In folklore, there is another rendition of the same idea: "For want of a nail ...," etc. This story entails two questions: (a) Can a nail cause the loss of a kingdom? and (b) What became of the nail that lost a particular kingdom? The first is eminently answerable; the second--without a lot of luck--is not. Here we have another instance in which it is not impossible to explain large effects through small causes. How would one use the "want of a nail" explanation? In the usual way: One would have an appropriate theory, and one would test it by means of standard social science methods. (For starters, one would set up a health survey of the horses maintained by what may turn out to be a careless cavalry. It would also be important to investigate one of the most important of all military inventions, the stirrup--see Volti, 1988:172-73.) How would this inquiry relate to the claim of "chaos" theorists, among meteorologists, that an asymmetrical sea gull in Singapore eventually can cause a hurricane in South Florida? I'm not sure, but I can guarantee that we shall never apprehend any sea gulls implicated in Hurricane Andrew (1992).

Incidentally, Spencer's (1910:295) term for situations in which small causes produce large effects was "fructifying causation," a term that could readily have been used by Ogburn in his discussions of inventions and their way of producing "many results from a single cause" (Ogburn, 1946:55).
 

(4) The aircraft accident reports prepared by the National Transportation Safety Board, and by similar boards of inquiry in other nations, are highly detailed case studies, idiographic analyses par excellence. Yet most of these studies would not be undertaken except for the fact that the results are eminently generalizable, typically serving to warn pilots and others against the sort of error that led to an accident in the case at hand. The recent crash of a Scenic Air Tours flight at Maui, Hawaii, for instance, was due to continued flight into bad weather. The NTSB (1993:v) generalizes that
 

REFERENCES (Chapter 4)
 

Abbott, Andrew. 1988. "Transcending general linear reality." Sociological Theory 6:169-86.
 

Abell, Peter. 1987. The Syntax of Social Life. Oxford: Clarendon.
 

Agger, Ben. 1989. Socio(onto)logy: A Disciplinary Reading. Urbana: University of Illinois Press.
 

Alwin, Duane F. and Robert M. Hauser. 1975. "The decomposition of effects in path analysis." American Sociological Review 40:37-47.
 

Barzun, Jacques. 1964. Science: The Glorious Entertainment. New York: Harper & Row.
 

Berger, Peter L. 1963. Invitation to Sociology: A Humanistic Perspective. Garden City: Doubleday.
 

Blalock, Hubert M. 1979. Social Statistics. New York: McGraw-Hill.
 

Blalock, Hubert M. 1989. "The real and unrealized contributions of quantitative sociology." American Sociological Review 54:447-60.
 

Bongaarts, John and Robert G. Potter. 1983. Fertility, Biology, and Behavior: An Analysis of the Proximate Determinants. New York: Academic Press.
 

Boyd, Lawrence H. Jr. and Gudmund R. Iversen. 1979. Contextual Analysis: Concepts and Statistical Techniques. Belmont: Wadsworth.
 

Carter, Jimmy. 1982. Keeping Faith: Memoirs of a President. New York: Bantam Books.
 

Cohen, Patricia. 1982. "To be or not to be: Control and balancing of Type I and Type II errors." Evaluation and Program Planning 5:247-53.
 

Collins, Randall. 1984. "Statistics versus words." Pp. 329-62 in Randall Collins (ed.), Sociological Theory. San Francisco: Jossey-Bass.
 

Collins, Randall. 1985. Three Sociological Traditions. New York: Oxford University Press.
 

Collins, Randall. 1986. "Is 1980s sociology in the doldrums?" American Journal of Sociology 91:1336-55.
 

Contreras, C. Cartas and Eduardo Cifuentes. 1987. "Prospects for Mexico-U.S. trade in livestock products." Pp. 201-18 in Bruce F. Johnston et al., U.S.-Mexico Relations: Agriculture and Rural Development. Stanford, CA: Stanford University Press.
 

Costner, Herbert L. and Robert K. Leik. 1964. "Deductions from 'axiomatic theory'." American Sociological Review 29:819-35.
 

Cousins, Mark and Athar Hussain. 1984. Michel Foucault. New York: St. Martin's.
 

Davis, Kingsley and Judith Blake. 1956. "Social structure and fertility: an analytic framework." Economic Development and Cultural Change 4:211-35.
 

Dawkins, Richard. 1986. The Blind Watchmaker. New York: W.W. Norton.
 

Duncan, Otis Dudley. 1966. "Path analysis: Sociological examples." American Journal of Sociology 72:1-16.
 

Faia, Michael A. 1966. "A Proposal for the Standardization of Type II Error in the Replication of Social Research." Sociology and Social Research 51:87-93.
 

Faia, Michael A. 1989. "Cultural Materialism in the Functionalist Mode." American Sociological Review 54:658-60.
 

Faia, Michael A. 1991. "Sociolog(omach)y: An Open Endorsement." Perspectives 14:4-5.
 

Glymour, Clark. 1980. Theory and Evidence. Princeton: Princeton University Press.
 

Gottman, John M. 1981. Time-Series Analysis: A Comprehensive Introduction for Social Scientists. New York: Cambridge University Press.
 

Halli, Shiva S. and K. Vaninadha Rao. 1992. Advanced Techniques of Population Analysis. New York: Plenum.
 

Hanushek, Eric A. and John E. Jackson. 1977. Statistical Methods for Social Scientists. New York: Academic Press.
 

Iversen, Gudmund R. 1991. Contextual Analysis. Newbury Park, CA: Sage.
 

James, William H. 1987a. "The human sex ratio. Part 1: A review of the literature." Human Biology 59:721-52.
 

James, William H. 1987b. "The human sex ratio. Part 2: A hypothesis and a program of research." Human Biology 59:873-900.
 

Kanfer, Stefan. 1988. "The Protean Penman." Time 132 (December 19):80-82.
 

Kissiah, Herman C. 1969. The College Student in the American Novel: 1930-1939 and 1964-1967. Ph.D. dissertation, Michigan State University. Ann Arbor: University Microfilms 70-15066.
 

LeMasters, E. E. 1975. Blue-Collar Aristocrats. Madison: University of Wisconsin Press.
 

Lenski, Gerhard, Jean Lenski, and Patrick Nolan. 1991. Human Societies: An Introduction to Macrosociology. New York: McGraw-Hill.
 

Mackie, Marlene. 1988. "Sexism in sociological research." Pp. 1-23 in Winnifred Tomm and Gordon Hamilton, editors, Gender Bias in Scholarship: The Pervasive Prejudice. Waterloo, Ontario: Wilfrid Laurier University Press.
 

Moore, Dan H. II and Barton L. Gledhill. 1988. "How large should my study be so that I can detect an altered sex ratio?" Fertility and Sterility 50:21-25.
 

Moreno, Jonathan D. and Barry Glassner. 1982. Discourse in the Social Sciences: Strategies for Translating Models of Mental Illness. Westport, CT: Greenwood.
 

Morrison, Denton E. and Ramon E. Henkel, editors. 1970. The Significance Test Controversy: A Reader. Chicago: Aldine.
 

National Transportation Safety Board. 1993. Aircraft Accident Report. Washington, D.C. NTSB/AAR-93/01.
 

Newton, Michael. 1988. Mass Murder: An Annotated Bibliography. New York: Garland.
 

Ogburn, William F. 1946. The Social Effects of Aviation. Boston: Houghton Mifflin.
 

Phillips, David P. 1979. "Suicide, motor vehicle fatalities, and the mass media: Evidence toward a theory of suggestion." American Journal of Sociology 84:1150-74.
 

Porter, Theodore M. 1986. The Rise of Statistical Thinking: 1820-1900. Princeton University Press.
 

Schneider, Stephen H. and Randi Londer. 1984. The Coevolution of Climate and Life. San Francisco: Sierra Club Books.
 

Shamos, Morris. 1990. "The lesson every child need not learn." Pp. 90-93 in Schellenberg, Kathryn, editor. Computers in Society. Third Edition. Guilford, CT: Dushkin.
 

Sica, Alan. 1988. Weber, Irrationality, and Social Order. Berkeley: University of California Press.
 

Spanier, John W. 1965. The Truman-MacArthur Controversy and the Korean War. New York: Norton.
 

Spencer, Herbert. 1910. The Study of Sociology. New York: D. Appleton and Co.
 

Strange, Susan. 1986. Casino Capitalism. New York: Basil Blackwell.
 

Taylor, Patricia A. 1987. "The celebration of heroes under communism: On honors and the reproduction of inequality." American Sociological Review 52:143-54.
 

Textor, Robert. 1967. A Cross-Cultural Summary. New Haven: Human Relations Area Files Press.
 

Thompson, Michael, Richard Ellis, and Aaron Wildavsky. 1988. Foundations of cultural theory: Part II: A historical and conceptual analysis of functional explanation. [Westview Press, forthcoming.]
 

Tufte, Edward R. 1974. Data Analysis for Politics and Policy. Englewood Cliffs: Prentice-Hall.
 

Turner, Stephen P. 1987. "Underdetermination and the promise of statistical sociology." Sociological Theory 5:172-84.
 

Turner, Stephen P. and William C. Wilcox. 1974. "Getting clear about the 'sign rule.'" The Sociological Quarterly 15:571-88.
 

Turner, Stephen P. and Jonathan H. Turner. 1990. The Impossible Science. Newbury Park: Sage.
 

Volti, Rudi. 1988. Society and Technological Change. New York: St. Martin's.
 

Walker, Henry A. and Bernard P. Cohen. 1985. "Scope statements: Imperatives for evaluating theory." American Sociological Review 50:288-301.
 

Ward, James A. 1986. Railroads and the Character of America, 1820-1887. Knoxville: University of Tennessee Press.
 

Williamson, Nancy E. 1976. Sons or Daughters: A Cross-Cultural Survey of Parental Preferences. Beverly Hills: Sage.
 

Woodcock, Alexander and Monte Davis. 1978. Catastrophe Theory. New York: Avon.
 

Wright, James D., Peter H. Rossi, and Kathleen Daly. 1983. Under the Gun: Weapons, Crime, and Violence in America. New York: Aldine.
 

Zetterberg, Hans. 1965. On Theory and Verification in Sociology. Totowa, NJ: Bedminster Press.
 

Zinsser, Hans. 1935. Rats, Lice, and History. Boston: Little, Brown.
 

Chapter 5

... AND HAPPY PROSPECTS
 
 
 
 
 
 
 
 
 
 
 

--Bernard Barber (1987:155-56)
 

(1) Introduction
 

This chapter begins with a debate, turns to a discussion of applied social science, arrives at a distinction among "reverse" regression models, "forward" regression models, and natural experiments, and then attempts to bring it all together within the framework of Henshel's (1976) argument regarding the prospects of sociological prognostication. The chapter implies that one of the advantages of a federated nation-state, such as the United States, is the presence of many jurisdictions able to carry out social experiments. This implication leads into the final chapter, which argues that the spatial areas that make up nation-states usually form a hierarchy, and that we should routinely make sure that all of our concepts have hierarchical organization, i.e., that they fit into taxonomies. Another implication, carried forward from the preceding chapter, is that while we are already very good in the realms of induction and syntactics, we should seek ways of making ourselves much better.

Social scientists should be conducting systematic experiments in large numbers, and we should introduce the results of these experiments into our best social simulations, post-facto designs, and projections of future events. I believe--and this belief is highly testable--that we have few experiments of this sort in progress, and that we therefore lose opportunities for the prestige-enhancing activities advocated by Henshel (1976). Social experimentation is costly, and it is risky. But as Wildavsky (1988) argues, when searching for safety precludes trial and error (or "learning by doing") solely on the basis that errors hurt, it also tends to preclude safety. And as Barber implies, Earth should be a hotbed of social science experimentation.
 
 
 

i
 

A few years ago I had the honor of an argument with Marion J. Levy. Levy (1988) had written a critique of an earlier work of mine (Faia, 1986), and I did my best--as I'll show in a moment--to make an adequate reply to his more compelling points. If I had been afforded space for extended remarks, I probably would have written something along the lines of the present chapter.

Levy (1988:244) has the opening allegation: "Nowhere in his book," he says, "can I find Faia's definition of the term functionalism or a forthright declaration that for his work that concept is an undefined predicate." At first blush, I feared that Levy was correct. But then I reexamined the pages in which I discuss Kingsley Davis (Faia, 1986:4-5), and I soon realized that Davis' provisional definition, as I present it, should have sufficed--for functionalism, dynamic functionalism, and Marxian functionalism. Each of these modes involves an attempt to relate parts of society to the whole and an attempt to explain social structures with reference to their consequences.

Admittedly a little vague, this definition was nonetheless a fine basis on which to start a book. By the time I got deeply into this volume (Faia, 1986:128ff.) I was prepared to offer a highly detailed definition of functionalism: It is a form of analysis that scores a 4 on my "hypothetical Guttman scale of sociological propositions" in that it involves multivariate causal analysis with time-series data and makes use of feedback and circular causation, with feedback and circular causation having an impact on the survival of an organization (or some part of it) or on the survival of an interest. Under this definition, it will soon become clear that this chapter is about functionalism and its prospects. I agree with Levy's remark that scientific concepts must have clear denotation; and I contend that nothing carries a more precise denotation than a Guttman scale. It was the major purpose of my earlier book to take the prevailing definitions of functional analysis and transform them from a gooey muddle into a Guttman scale.

One can readily imagine a study of organizational survivorship, or the survival of some interest, that would be ordered (i.e., the study would have time as a variable), multivariate, and causal; but such a study would not immediately be functionalistic. It could become so, however, if the investigator were to follow my admonition--one which Levy quotes but does not seem to grasp--that "powerful forms" of functional analysis would occur if survival were treated as an adaptation process involving feedback. In other words, when the survival of an organizational type (or interest) is threatened, what sorts of changes take place that may (or may not) lead to enhanced survival prospects? Note that I do not use tempting phrases about changes that might occur in order to enhance survivorship, leaving open a question that Levy wrongly considers closed. On Levy's question whether Lola gets what she wants, I reserve judgment: Let Mephistopheles bring her front and center and I'll be delighted to ascertain, empirically, whether she achieves her objectives.

This brings us to Levy's second major contention (1988:244): "Faia heavily emphasizes what he refers to as 'circular causation'--a term that in itself should have set off warning lights and buzzers in his mind." Believe me, it did and it does. In large part my book dealt with the costs and benefits of trying to assess two-way causal interactions through time. There are many pitfalls, but there is also a possibility of placing functional analysis--our best established and, with Marxism, our most durable theoretical tradition--on a firm methodological foundation that would force us to eliminate the traditional mumbo jumbo about prerequisites, teleology, stability, and consensus while testing falsifiable hypotheses about system interactions through time.

Incidentally, I'm a little mystified by the fact that in the same paragraph where Levy warns us about circular causation he expresses his worries about possible neglect or abuse of the "point of view of the actor." At the moment I see no connection between circular causation issues and the question of the actor's perspective. The latter is just another set of variables that we introduce into an analysis whenever it appears to be relevant. It is entirely possible to deal with circular causation in studies that give little attention to the point of view of the actor. And therefore I am not at risk, despite Levy's apprehensions, of accusing black holes of deliberately gobbling up little stars--for whatever nefarious and polymorphously perverse purposes these Sultans of Suction may have. On the other hand, natural scientists and engineers cannot live without circular causation (Faia, 1986:Chapter 5); recall Dawkins' discussion, cited earlier, of the complex causal interactions involved in the evolution of sexual ornamentation. I am therefore mystified further by the claim (Levy, 1988:244) that circular causation "... is a vice in all empirical sciences." We should realize, for instance, that when Clausewitz tells us that warfare is a continuation of diplomacy by alternative means, he implies that diplomacy is a continuation of warfare, that war and diplomacy may therefore operate as functional alternatives to one another, and that we would do well to examine the causal interaction between them.

Despite Levy's misgivings and hesitations, the functionalist model encourages us to focus on time as a variable more than do alternative social theories; circular-causation hypotheses encourage us to use methods that capture time as a variable. Until Levy demonstrates the irrelevance of my best examples--the Malthusian theory of population (which evolved into a highly sophisticated adaptation theory between 1798 and 1803), the Club of Rome world model, Zerubavel's work on the social definition and social organization of time, the sex preselection of children, sex discrimination in academic salaries, the delict-sanction model as applied to capital punishment, the interaction of military expenditures and economic stagnation at the state or nation-state level, the Clausewitzian hypothesis--I will continue to believe that each of these research traditions has achieved strength insofar as scholars have been willing to examine the reciprocal interaction of variables through time. I will continue to believe further that, even though the scholars in question may not realize it, every time they examine interactions through time in a way that emphasizes organizational survival and/or the degree to which interests (i.e., manifest functions) are served, they are practicing functional analysis. If functional analysis is nothing more than intellectual self-abuse, most of us will soon be blind.
 

ii
 

In a brief but provocative discussion of social prediction, Henshel (1976:26-32) describes "engineered systems" of the sort that regulate everything from refrigerator temperatures to airline flight paths, from carburetor fuel-air mixtures to automatic elevator doors, from household water pressure to freeway traffic flow. These systems impinge on our daily lives in ways that are so smooth and unobtrusive that we hardly ever take notice of them--until something goes wrong. Henshel explores the possibility that the social sciences might achieve comparable successes in social engineering, and might thereby enable themselves to take advantage of a benign circle in which a few modest successes--Barber's "quiet cases"--would create opportunities for us to try again, with larger stakes. I don't know whether the expanded opportunity structure envisioned by Henshel will ever come to pass, but I do know that from time to time social scientists are called upon to engage in social engineering, and that whenever we accept these challenges we must be attuned to the need for "triangulation" (Babbie, 1989:99). That is, we must learn to use a variety of conceptual and theoretical perspectives and a variety of methodological strategies and tactics, lest the sorts of opportunities held forth by Henshel be lost.

Henshel's inquiry, in essence, develops a functional theory; it is self-exemplifying. By definiton, this theory involves feedback. It is a theory about the interaction of social scientists with their constituencies, their publics: If we can manage to produce useful knowledge, says Henshel, we shall eventually be afforded many new opportunities to produce such knowledge. By useful knowledge, I refer to theories that enables us to explain and (especially) to predict social phenomena that threaten organizational survival and/or various interests. We immediately become uneasy, because we know that our historical record in predicting future social phenomena has been considerably less than perfect. However, one implication of Henshel's theory is that expanded opportunities for social engineering--and especially for designing controlled natural experiments--would enable us to improve predictions dramatically. In other words, we should try to generate a positive feedback process that would enable us to become competent social engineers, to transform ourselves. In general, we explain and predict the behavior of engineered social systems much more successfully than we explain and predict the behavior of systems that we have not designed. What Henshel implies, of course, is that experimental research is superior to non-experimental research, provided that one has opportunities for conducting experiments both within laboratories and (especially) within natural environments.
 

(2) Sex discrimination in faculty salaries: Toward projections of the middle range
 

I shall illustrate several features of social engineering by describing a participant observational study in which I was involved over a period of several years.
 

i
 

Toward the end of 1981, the Affirmative Action Advisory Committee (AAAC) of my university issued an annual report that contained the following paragraphs:
 

It was clearly the first responsibility of the committee to build support for the idea that these periodic salary assessments should be made; for many faculty members, the notion of having regular evaluations of the salary structure was totally alien, an alleged invasion of privacy and a violation of professional integrity. The committee therefore voted (for this reason and others) to make its report public, and when the university administration resisted this idea the AAAC chairman found it necessary to release the report to the local media. This action caused considerable consternation in many quarters, and the chairman found himself summarily relieved of duties as committee chair during the spring of 1982. It is my belief that the publicity surrounding the AAAC report did not retard implementation of its recommendations in any way--if anything, prospects were enhanced--and I do not believe that social engineers need to tread softly. They have to take pains, however, to avoid major blunders of a methodological or theoretical nature.
 

ii
 

As the fall semester, 1987, was getting under way, I was asked by the chair of the faculty women's caucus to prepare a memorandum that would explain to caucus members my perceptions as to how salary assessments, with a focus on gender equity, would be made. This colleague knew that I had continued gathering data on the local faculty, and she also knew that I had supervised graduate research on the university faculty and the nature of its reward system. Incidentally, I had found that because of my willingness to comply with the ethical standards against identifying individual faculty members by name (and salary)--despite the fact that Virginia law does indeed permit such identification--I had the full cooperation of the administration in all data-gathering activities and was allowed access to the university's master personnel file as a way of ensuring accuracy and speed of data recovery.

The following is my memo:
 

iii
 

Just after the end of the semester, around January 1988, it became clear that some sort of corrective action would have to be taken by the university administration. Specifically, the administration was exploring the possibility of bringing in an outside consulting firm that would evaluate the salary equity situation at William & Mary. The chair of the women's caucus and the dean of the faculty asked my opinion about the study proposed by the consulting team:
 

By 1 March 1988 matters were proceeding apace: Consultants had been hired, they were in the process of studying the faculty salary structure and formulating their recommendations, and these recommendations were likely to be implemented. In my discussions with various colleagues it became clear that the administration, for reasons on which I can only speculate, was becoming involved in a rush to judgment, and that detailed case studies of alleged instances of discrimination were not likely to be made.

Apparently, this problem arose out of bureaucratic necessity:
 

Furthermore, there was a large ethical question: Should the funds necessary to correct discrimination be taken from male faculty members (or administrators) who (1) may have had no role in perpetrating discrimination, (2) may not have benefitted in any way from the existence of discrimination, and (3) may actually have suffered an impaired collegiality arising from discrimination and the low professional morale that it may create among its victims? As a social engineer I have no qualms against saying that in this instance the system needed an infusion of new funds. As nearly as I can tell, however, this infusion did not take place.

In my opinion, the decision not to make detailed case inquiries was a mistake. As I will try to show below, it would be highly desirable for a given university to arrive at a time when its faculty would be convinced that discrimination essentially had been eliminated, so that from that moment onward procedures could be established that would ensure that the salary structure would not again be distorted.

Around 23 March 1988, word began circulating that the administration was proposing a simple regression model that would employ rank, seniority, and field--with weights for the latter derived from national surveys. Apparently a small number of women--a dozen or so--were to receive contractual adjustments based on this model. Nobody seemed especially happy about the procedure; to the administration, however, it probably looked fast and smooth.
 

Around 28 March 1988 a local committee was formed to deal with the gender-equity issue. This committee was chaired by a sociologist (not the author) and included several faculty members and administrators who had insight into the faculty salary process. The major purpose of the committee was to clarify for various faculty members, including women who were alleging discrimination, what sorts of efforts would be made to address gender-equity problems. Although the state of Virginia has a legal requirement that meetings of this kind be open to the public unless the discussion involves specific cases, it is a part of the folkways of the university that highly controversial meetings involving faculty and administrative matters are held in closed sessions, and I felt that my own position as a somewhat distant observer would have been impaired if I had insisted on attending. Consequently, my knowledge about it is indirect. Judging by the testimony of participants, I believe that the meeting clarified the nature of the corrective actions on which the institution was embarked, and there was no significant opposition expressed at the meeting against the administration's proposals.
 

iv
 

After a period of relative quiescence that lasted for nearly a year, on 6 February 1989 the Dean of the Faculty of Arts and Sciences announced "gender equity monitoring," to be conducted on an annual basis. By now we had come full circle: A key recommendation made by an affirmative action committee almost a decade earlier would finally be implemented. I do not believe that most members of the current administration or faculty remember this recommendation, and I do not believe that the affirmative action committee, per se, had much to do with the administrative decision to implement annual salary-equity assessments. I believe that these assessments have been established out of a sort of functional necessity, and that past advocates of them were merely recognizing a process that was more or less inexorable.

In the nearly twenty-five years that I have been a member of the William & Mary faculty, the sex-discrimination issue has exploded three times in a manner so compelling that, in each instance, adjustments have had to be made in the salaries of several female colleagues. I believe that the number of cases involving salary adjustments has generally been in the neighborhood of a dozen, out of a faculty of about 300. The stability of this number supports the functionalist-necessity argument: After a time lapse of perhaps five to ten years, the magnitude of the discrimination gap--both the amount of income denied and the number of victims identified--arrives at a certain threshold, and the institution must take corrective action. At least until annual equity monitoring was introduced, the process had a characteristic frequency and amplitude. Every five to ten years, a discrepancy of about five to eight per cent in female salaries has been evident and has been sufficient to activate a "critical mass," to bring about effective protest.

In developing critical mass theories of the rise of protest movements (Sullivan, 1977; Newcomb, Jesness, and Erickson, 1978; Tsai, 1978; Rahav, 1981; Wilson, 1983; Oliver, Marwell, and Teixeira, 1985; Oliver and Marwell, 1988) one would be wise to take into account both the number (or proportion) of potential protesters in an organization, their interaction, and the nature and magnitude of their grievances. When these quantities coalesce in a certain way, there is a qualitative transformation and the institution acts abruptly in an attempt to resolve its internal tensions. Or, as an alternative, it tries to introduce a qualitatively distinctive form of self-surveillance--in this instance, annual gender-equity assessments. As I write these words, I believe that this institution is about to undertake its first annual gender-equity assessment, and I am not optimistic: I have many reasons for doubting that the pattern of the last twenty-five years, involving periodic protests and precipitous fix-ups, will be interrupted. I have implied throughout this analysis that the most effective way of interrupting the established pattern would be to put together a workable combination of etic¹ modelling, emic investigation, and concurrent efforts to explain to the entire university community the nature of the salary-setting process and its outcomes.

The basic patterns and trends uncovered by regression modelling of the salary structure should be discussed in detail each year, in the various university news media, administrative memoranda, committee reports, etc. Any trendlines in sex disparities would be highly reliable if essentially the same regression model were applied each year--do not, however, lose sight of Abbott's admonitions, discussed in the preceding chapter, against neglecting the causal context. Each faculty member would be able to see his or her own relative position in the faculty salary hierarchy, and to see any changes that might occur in that position. Inquiries about perceived injustices could be addressed immediately, and misunderstandings could potentially be resolved. The sorts of tasks that we believed were to be undertaken by the university's consulting team in 1988--but which were not undertaken--should be pursued diligently each year. My prediction is that the salary-setting process would soon be demystified by these sorts of activities, and that "the functions of secrecy" would turn out to be dysfunctions.

It requires a degree of temerity to advocate social engineering at a time when the social sciences, qua sciences, are being challenged from many directions. But I continue to share Henshel's confidence that if we select opportunities carefully, we have good prospects of success. In the case at hand I exercised considerable care, even though I cannot claim to have been the chief engineer, or even a low-level foreman. I was aware, for instance, that academe is a place in which there is a relatively strong commitment to the idea that rewards should have a demonstrable relationship to productivity. There are, in short, more or less explicit rules about how rewards such as salaries should be allocated. For salaries, it is possible to explain statistically as much as 80 per cent of the variability that occurs within a given faculty. At the national level, this percentage usually drops to around 55 or 60, and my intuitions tell me that it would be far more difficult to implement salary-equity procedures at the national level than at the level of a given college or university, a given type of institution, the several institutions that comprise the higher educational system of a given state, etc. In the realm of higher education, in general, the more localized the situation, the clearer the social norms.

For the case at hand, another selection criterion was the fact that essential variables--salaries, salary discrepancies related to sex, the sense of equity resulting from a given reward structure--were readily measurable. If instead of manipulating these sorts of variables we had been dealing, say, with the development, implementation, and assessment of ways of reducing alleged sexual harassment involving faculty members and students, the tasks of social engineering would have been formidable indeed. There is little consensus about whether such an effort should be undertaken at all, it is unclear what we mean by sexual harassment, we have little knowledge about means of social control that would influence the likelihood of sexual harassment, and it would be difficult to determine whether any set of means were having the intended impact. I do not discourage social engineering in this realm; I merely warn that the prospects of success, in Henshel's sense, are far more limited.
 

v
 

Codicil 1993: My pessimism about salary-equity monitoring appears to be justified. From time to time I have spoken to the appropriate administrators and official faculty monitors. Clearly they have the data, they have done the analysis, and they have arrived at some conclusions. It is equally clear, however, that they will not release their results: They do not wish to stir things up.

As Henshel's work suggests, aspiring social engineers must first acquire sufficient prestige that we might begin the tortured process of educating our colleagues. In Comtean terms, we must become the high priests of our local administrations.
 

(3) Allocating faculty salaries with Minitab BASIC, and allowing feedback from future states
 

The preceding paragraphs apply a hypothesis that will be developed further in a later chapter, with reference to the production of beer and electricity. The hypothesis suggests that, in general, there is a tradeoff between frequency and amplitude. In Kuhn's words (1974:31), the size of an amplitude varies inversely with the speed and sensitivity of negative feedback. We might therefore expect that in the present instance frequent and sensitive "equity monitoring" of a reward process would reduce the magnitude of actual and perceived injustices.

In this section I design the second part of a two-tiered salary-determination process. In the initial phase, faculty members generate merit scores for one another; one must be assured that these scores are arrived at without prejudice and discrimination. The second phase involves the administrative contribution to the setting of faculty salaries. In my view, administrators should have hardly any role in evaluating individual faculty members, i.e., in generating merit scores. They should evaluate divisions, departments, programs, and they should participate in decisions about financial policies--such as the amount of money to allocate for "merit" rather than cost of living, or the overall difference that ought to exist between the highest and lowest salaries. My discussion assumes that these norms have been established, that administrators do not examine the merit scores of individual faculty members except when there are fairness issues such as discrimination, and that the best administrative procedures, again, are those in which etic abstractions interact with emic confrontations.

The appendix to this chapter presents a detailed design, written in Minitab BASIC, for a faculty reward process that would operate with unusual transparency, would allow various salary-distribution policies to be implemented, and would permit simulation experiments. Notice that these experiments begin with past conditions (e.g., last year's salary schedule for a faculty), enter various contemporary data (e.g., current merit scores), and project a future state (e.g., next year's salary schedule). In a moment, I shall refer to this sort of experimental model as a PPF (past-present-future) simulation.

The program is entirely deterministic: Once a given set of faculty merit scores has been inserted into it, and once an explicit set of assumptions has been made about how the program will process these scores, there is only one outcome for next year's salary schedule. Anybody who wishes to challenge this outcome will be able to identify precisely those factors that account for it. I argued in the preceding section that when the etics of the salary-determination process are made totally transparent, the emics of the situation will tend to become far less confrontational. In this section (and the associated appendix) I try to show how to develop a transparent etic model--itself a rather dialectical if not oxymoronic proposition, since we generally define etic models as being intelligible only to a small number of specialists. However, one consequence of good social engineering is that a wider public will learn to be less intimidated by, say, regression equations.

Under this program, for instance, administrators (and, one hopes, faculty) must make several interrelated policy decisions. First, they must establish (or ascertain) the size of next year's salary budget. Second, they must decide what proportion of new funds (if there are any new funds) will go to cost-of-living increases. Third, they must decide what proportion of new funds will go to "pure" merit percentage increases--i.e., a percentage increase over current salary--as opposed to a flat-dollar merit component, which provides the same monetary increase for all faculty members who have the same merit score, regardless of their current salaries. Finally, administrators must decide which departments, if any, are worthy of special over-reward (or under-reward), and which are not.

All these decisions are political, and they should all be fought out as such, openly. If there is a policy that involves "levelling" of faculty salaries, for instance, this policy will tend to increase the cost-of-living component and the flat-dollar merit component. Such a policy might affect some individual faculty members adversely, but these adverse effects can be explained entirely with reference to the policy and not with reference to alleged injustices involving personal victimization. The same can be said about the policy of rewarding and punishing specific departments, as units.

A multivariate model that makes projections into the future, and that operates deterministically or nearly so, typically involves a PPF simulation: Instead of asking, as we do in regression analysis, "What processes could have produced these results?" PPF simulations ask "What results would be produced if we were to set up such-and-such a process?" The salary-determination program asks the latter question; a program package such as the Florida Interactive Modeler, to be introduced in a moment, enables us to ask many such questions.

Again, it should be noted that the best social simulations are essentially "ideal-type" experiments. Most of us recognize why it is that strict experimental methods are superior to our typical post-facto designs. We also know that the central problem of experimentation is that, usually, we cannot carry out experiments in "natural" situations where there is a lot at stake. Henshel asks us to contemplate what we would do if we were to achieve drastically expanded opportunities to design experiments in situations where weighty variables would be manipulated. He also challenges us to have the courage to project the results of such experimentation into the future by means of social simulations.
 

(4) From a FIRM grip on reality to the AIDS epidemic: The future of social simulations
 

i
 

The Florida InteRactive Modeler (FIRM) is a highly generalized simulation program that has a lot in common with other well known programs such as the Club of Rome world model, SLAM, Simscript, the Wharton model of the U.S. economy, and the computer simulations that accompany textbooks such as Bainbridge (1987) or Feigelman (1993). A major advantage of FIRM--one that makes this program directly comparable with SLAM or Simscript--is its generalized character that enables it to model just about any socio-psychological or socio-economic process in which there is a dependent variable that is influenced by combinations of causal agents, their interactions, constants, and random disturbances. The diskettes that accompany FIRM contain an illustration of socio-economic processes--a price-prediction model for a grocery basket filled with a standard list of "junk food" items--and several illustrations of socio-psychological processes such as a model of the famous "sleeper effect" in studies of public opinion.

FIRM simulations involve what I call forward regression models, as opposed to the usual reverse models. In creating reverse models, we gather vast amounts of information about variables that have already been processed by nature as causes and effects, and we then try to ascertain and reconstruct what nature has been doing. In forward models we make intelligent decisions about what nature could be doing or could have done under various circumstances, and then we activate the model, allow its causal factors to vary, and see what happens. Critics of this approach often make a GIGO argument, claiming that simulation models have a lot in common with Tom Lehrer's famous sewers: You get out of them exactly what you put into them. But experienced users of elaborate simulation models know that this claim is untenable. In a SLAM traffic-control simulation involving only a small number of intersections and bottlenecks, and perhaps a frantic rush hour or two, one cannot possibly know a priori how to set the timing of traffic lights so as to minimize waiting time for drivers, and one's intuitions on such matters often turn out to be misleading. Similarly, in the complicated renditions of attitude formation and change that are available in the FIRM package, one encounters many surprises.

It occurred to me recently that forward regression models have a lot in common with "counterfactual history." In her excellent book on social explanation Susan James (1984:60-62,134-36) discusses the ways in which arguments contrary to historical fact have been used to test ideas such as the "great man" theory of leadership. Fogel (1964), in a study to be discussed in the next chapter, asks what the cost-benefit consequences would have been if the United States had made the same commitment to waterways and highways that it made to the railroads--a strong commitment to the wrong transportation system could be a lot more costly than merely timing traffic lights wrongly. Counterfactual historical analysis often involves (at some level of development) the kinds of measures, weights, interactions, constants, and unknowns that comprise FIRM (and comparable) simulations.

One of the best features of FIRM as a learning tool is that it forces the model designer--and presumably the model user--to become familiar with the literature pertaining to a given process; one surely cannot write a realistic model without this familiarity. Furthermore, the FIRM approach gives us an incentive to keep up with the literature--whenever a new finding appears it is possible to incorporate it into one's models and see how large an impact it is likely to have. Finally, it would be a good training exercise to use (reverse) regression analysis in order to tease out the properties that some programmer has written into a (forward) simulation model.
 

ii
 

Taking time as a criterion, there are at least three types of social simulations: First, those that pertain entirely to past events, such as the "counterfactual history" simulation presented in the next chapter. Second, those that try to capture events that would take place entirely in the future, such as the sex selection of children (prior to conception) or mandatory AIDS testing for couples applying for marriage licenses. Finally, there are simulations that center on the present, have origins in the past, and must therefore pass muster with the contemporary and historical record, such as it is. Simulations of this type are PPF simulations. Using such a simulation, it should be possible to make projections into the future, to predict (with qualifications) events that we might otherwise miss. If one were to experiment with a large array of PPF simulations one would soon realize, with Comte, why sociology is at the apex of all academic disciplines and is therefore the most demanding of all disciplines.

The current AIDS epidemic, for instance, has been modelled as a PPF simulation by Whicker and Sigelman (1991: Chapter 6). Their program, written in BASIC, permits experiments with variables such as the transmission probabilities for various combinations of AIDS carriers and their partners (homosexuals, bisexuals, heterosexuals), the use or non-use of condoms, frequency of intercourse per month, number of new sex partners per month, and so forth. Values for these variables are allowed to vary through a certain range as the program executes, but the model will become more realistic as empirically-based estimates of actual values become more reliable. The same may be said about estimated numbers of individuals infected with AIDS at the time when the simulation begins, say 1988. For this variable also, Whicker and Sigelman experiment with several different settings.

In my own experiments with the Whicker-Sigelman program, I began by corroborating their results; I also found that when their model iterates through more than twelve months, incredibly large numbers of new AIDS infections are produced, which suggests that the model may have its greatest utility over a short term. In my runs with the model as presented (Whicker and Sigelman, 1991:106-7), the number of new infections among heterosexual females in the United States varied, in a twelve-month simulation, between 841 and 1,184,880; needless to say, these projections are not highly precise. The low estimate is based on the assumption that (1) condoms are consistently used, (2) frequency of intercourse is low, (3) there is one new relationship per month, and (4) the starting population of infected persons consists of gay males only. The high estimate resulted from a reversal of (1) through (3), while (4) remained unchanged.

The extremes of the model, of course, are unrealistic: We cannot assume, for instance, either universal condom use or consistent non-use. Again, the simulation will have utility only insofar as we build into it realistic assessments of such practices as condom use, the nature and frequency of various sexual practices, etc. If we were to make serious efforts along these lines, we would almost surely do better than Masters, Johnson, and Kolodny (1988:15), who estimated that the United States would experience 300,000 AIDS deaths by the end of 1991. Total mortality does not appear to be nearly this high; by the end of 1991, the number dead is around 120,000, well below even the 179,000 estimated a few years ago by the U.S. Public Health Service (Masters, Johnson, and Kolodny, 1988:15).

I do not wish to imply that the above estimates were "wrong" in a simplistic sense; they were both intended--as in the case of the Club of Rome world model, another PPF simulation--to act as self-destroying prophecies. Clearly Masters, Johnson, and Kolodny are social actors with a point of view: They intended their projection to serve as a warning, and they now may be delighted by the fact that this projection, whatever its basis, turned out to be wrong. The lesson for them, and for us, is that if their warnings actually served to modify behavior, then these behavior modifications need to be incorporated into our more advanced theories and simulations.

We must also incorporate the possibility that, as we learned repeatedly from the Club of Rome world model, it is hard to distinguish between a self-destroyed prophecy and a self-delayed prophecy. By the end of 1991 we did not need the celebrated case of Magic Johnson to suggest that the nation had relaxed its guard in the struggle against AIDS. One only has to spend a few moments perusing the pages of the New York Times to discover substantial evidence that, by early 1991, it was widely believed that the AIDS threat had diminished considerably (Kolata, 1991). Self-delaying (or destroying) prophecies involve exactly the form of feedback discussed throughout this chapter, and it is entirely possible that adaptive efforts in the realm of AIDS prevention have already gone through several cycles of unknown frequency and amplitude.
 

(5) Conclusions
 

i
 

In the wake of the 1990 census, many states again are finding it difficult to draw a new set of congressional and state-legislative districts. In the typical instance, simulation programs design a long series of redistricting proposals that try to satisfy simultaneously the demands made by political parties, by individual politicians intent on escaping (or taking advantage of) gerrymandering, by minority segments of the electorate who wish to be represented by their own kind, and by various constitutional principles as interpreted by the Department of Justice and the courts. Given the rapidity with which high-speed computers create new designs, it is not surprising that many simulations are evaluated and that most of them quickly become self-destroying prophecies. Eventually, however, "one true story" must emerge from these inexorable exercises in social engineering.²

The process of creating new legislative districts has a lot in common with the process of allocating faculty salaries: It must be done from time to time, at fixed intervals; there is a high level of normative regulation of the process--a high degree of consensus as to how it should be carried out--and therefore many potential solutions may be eliminated a priori; nonetheless, there remains an indefinitely large number of plausible simulations that must be examined in detail; finally, the data necessary to make such simulations are readily available and have a high quality. In Henshel's view, if social scientists seek the power to change society for the better, if we seek prestige and ways of creating new knowledge, then these are precisely the circumstances under which we need to move quickly.

ii
 

If Henshel's analysis is correct, the lives of social scientists are likely to become a series of ethical choices and ethical crises, as we decide whether it would be right to design a particular feature into an organizational process. This transformation is inescapable: The centrality of ethical issues in the lives of natural scientists is clearly shown by every issue of the Bulletin of the Atomic Scientists, by every meeting of the "committee on human subjects research" of one's university, and by every instance in which a medical doctor must make an appearance before an abortion committee. It may seem ironic that social scientists, currently, are far less likely than their natural science colleagues to face agonizing ethical choices: I assume, however, a very high correlation among the prestige of an academic discipline, its competency, the extent to which it has opportunities to conduct controlled natural experiments, and the frequency with which its practitioners must face ethical dilemmas.

In brief, we need a code of ethics for social engineers. This code must be based on adaptation theories of high credibility, so that what appear to be successful ventures will not turn out to be a series of disasters. It is arguable that, up to now, our best opportunity for social engineering was provided by the American military establishment (Bowers, 1967); this venture did not have a happy outcome.
 

NOTES (Chapter 5)
 

(1) For an excellent discussion of the uses of the terms "etics" and "emics," especially among cultural anthropologists and linguists, see Harris (1979:32-45). In emic analysis, the "native informant" is the "... ultimate judge of the adequacy of the observer's descriptions and analyses." In etic procedures, the observer is the ultimate judge (Harris, 1979:32). Regression analysis is essentially an etic operation; depth interviews that probe perceptions, explanations, and judgments of various aspects of, say, salary distributions would be an emic operation.
 

(2) For a discussion of Virginia's situation, see Baker (1991) and Harris (1991).
 

REFERENCES (Chapter 5)
 

Babbie, Earl. 1989. The Practice of Social Research. Belmont, CA: Wadsworth.
 

Bainbridge, William S. 1987. Sociology Laboratory: Computer Simulations for Learning Sociology. Belmont, CA: Wadsworth.
 

Baker, Peter. September 24, 1991. "Behind enemy lines, candidates struggle for political survival." Washington Post 114:B1.
 

Barber, Bernard. 1987. Effective Social Science: Eight Cases in Economics, Political Science, and Sociology. New York: Russell Sage Foundation.
 

Bowers, Raymond V. 1967. "The military establishment." Pp. 234-74 in Paul F. Lazarsfeld, William H. Sewell, and Harold L. Wilensky, editors. The Uses of Sociology. New York: Basic.
 

Elliott, Jeffrey D., Thomas K. Srull, Marc W. Durnin, and C. Michael Levy. 1983. FIRM: Florida InteRactive Modeler: Volume I: The Nature of Attitudes and Attitude Change. Iowa City: CONDUIT.
 

Faia, Michael A. 1986. Dynamic Functionalism: Strategy and Tactics. New York: Cambridge University Press.
 

Feigelman, William. 1993. Hands On Sociology: Using Computers to Learn about Society. New York: McGraw-Hill.
 

Fogel, Robert. 1964. Railroads and American Economic Growth. Baltimore: Johns Hopkins University Press.
 

Harris, John F. July 24, 1991. "U.S. backs revamped Va. districts." Washington Post 114:C1.

Harris, Marvin. 1979. Cultural Materialism: The Struggle for a Science of Culture. New York: Vintage Books.
 

Henshel, Richard L. 1976. On the Future of Social Prediction. Indianapolis: Bobbs-Merrill.
 

James, Susan. 1984. The Content of Social Explanation. New York: Cambridge University Press.
 

Kolata, G. June 18, 1991. "Experts debate if AIDS epidemic has at last crested in U.S." New York Times 140:C1.
 

Kuhn, Alfred. 1974. The Logic of Social Systems. San Francisco: Jossey-Bass.
 

Levy, Marion J., Jr. 1988. "Dynamic Functionalism: An Exchange." Social Forces 67:247-49.
 

Levy, C. Michael and Marc W. Durnin. 1981. FIRM: Florida InteRactive Modeler: Author and Programmer Manual. Iowa City: CONDUIT.
 

Marwell, Gerald, Pamela E. Oliver, and Ralph Prahl. 1988. "Social Networks and Collective Action: A Theory of the Critical Mass. III." American Journal of Sociology 94:502-534.
 

Masters, Wm. H., Virginia E. Johnson, and Robt. C. Kolodny. 1988. Crisis: Heterosexual Behavior in the Age of AIDS. New York: Grove Press.
 

Newcomb, Theodore M., Carl F. Jesness, and Maynard L. Erickson. 1978. "Characteristics of Youths in a Sample of Correctional Programs: Differences Associated with Programs and Time Spent in Them." Journal of Research in Crime and Delinquency 15:3-24.
 

Oliver, Pamela, Gerald Marwell, and Ruy Teixeira. 1985. "A Theory of the Critical Mass. I. Interdependence, Group Heterogeneity, and the Production of Collective Action." American Journal of Sociology 91:522-556.
 

Rahav, Giora. 1981. "Culture Conflict, Urbanism, and Delinquency." Criminology 18:523-530.
 

Sullivan, Thomas J. 1977. "The 'Critical Mass' in Crowd Behavior: Crowd Size, Contagion and the Evolution of Riots." Humboldt Journal of Social Relations 4:46-59.
 

Tsai, Yung Mei. 1978. "Urbanism and Restrictions on Women's Socioeconomic Roles: An Analysis of Communities in Taiwan." Journal of Asian and African Studies 13:256-264.
 

Whicker, Marcia Lynn and Lee Sigelman. 1991. Computer Simulation Applications: An Introduction. Newbury Park: Sage.
 

Wilson, William J. 1983. "Inner-City Dislocations." Society 21:80-86.
 

APPENDIX (Chapter 5)
 

This program is intended as an efficient, philosophically defensible way of allocating faculty salary increases, where these increases have three components: a cost-of-living increase received by everybody, the usual percentage-based merit increase, and an unusual merit increase that provides the same amount of money for all faculty members who have a given merit score, regardless of their initial salaries. The last component will be called a flat-dollar merit increase.

The program is very simple. Administrators of the salary budget must enter the following items:
 

The last three items involve "political" decisions that are usually the prerogative of local administrators and faculty, and under this program these decisions must be made clearly, explicitly, and understandably. Administrators are able to experiment with any number of combinations of values for the last three "inputs" listed above. The program, in other words, is a simulation; as such, it works as a forward regression process in the sense defined earlier. It is interesting to note that the usual reverse regression analysis would not be able to solve this simulation program correctly--to figure out, in other words, how it was put together--unless the regression model were written by somebody who understood the unusual way of rewarding "merit": One way involves a percentage increase in salary, while the other involves a flat-dollar amount for anybody with a given merit score.

The listing below is output from a Minitab program that simulates the process of setting next-year salaries for all members of five departments. Any number of departments can be written into the program, and a given department can have any number of members. Departments are allowed to use any merit evaluation scheme they wish, with the understanding that they must provide a single numerical score for each member, and that all these scores must be positive numbers. Departments should be made to understand, further, that merit scores will be treated as a "ratio" scale, so that a score, say, of 6, ceteris paribus, will have twice the impact on one's salary increase as a score of 3.

This program uses the "proportions-of-pies" method: All merit scores are converted to proportions of all points awarded by a given department, and these proportions are then multiplied against three "pies": the cost-of-living increase pie (PIE(1)), the percentage-increase pie (PIE(2)), and the flat-dollar pie (PIE(3)).

The Minitab statistical package has been used to simulate scores for merit (lines 0110-0510), and these scores are printed out. Note that proportional merit scores all sum to 1.0 within a given department, so that any pie made available to a given department will be allocated completely according to these proportional merit scores. In actual implementations of the program, of course, the administrator must enter the faculty merit scores into the program. I assume that administrators regard the faculty merit scores as sacrosanct, and that they (the scores) cannot be changed except in cases involving provable injustices such as discrimination based on age, sex, race, etc.

Minitab next simulates a salary schedule for the current year, with departmental means and standard deviations varying considerably (lines 1420-1680). In actual implementations of the program, again, administrators must enter the current salary schedule.

In lines 1720-1860, Minitab calculates the correlations of merit and current salaries, by department. If these correlations tend to be above 0.0, it is likely that the percentage-based merit increase part of the budget will rise above the amount targeted (e.g., .5). If correlations tend to be below 0.0, the opposite effect is likely. My own simulations indicate that this proportion remains stable, despite the fact that the percentage-based merit increases are allocated without taking into account the distribution of base salaries.

The program calculates the total size of the current budget (lines 1900-1910), the size of each department (lines 1940-1990), and the size of the total faculty (lines 2020-2030). All these values should be used as checks.

After the administrator has entered the size of next year's total salary budget (lines 2060-2070), the program obtains the amount of the budget increase (lines 2100-2110). The proportion of this amount that is intended for pure merit increases should be entered at line 2160, with the understanding that if percentage increases, when applied to current salaries, do not precisely equal the proportion of the budget add-on that is intended for this purpose, the discrepancy is allowed to stand. Percentage-based merit increases are taken as inviolable, so that if (due to base salaries) they take a larger part of the budget add-on than intended, they are not reduced. The excess is taken out of the third pie, the flat-dollar pie. (In this tactic, the loss of a degree of freedom, usually imposed on percentage increases, is imposed on the size of the third pie. This pie must therefore be a residual of the first two pies, assuming that the overall budget increase is a fixed amount.) The rationale for this feature is that those faculty members who have high merit should not, for purposes of the percentage-based merit increase, be penalized solely because they have relatively high base salaries. (It should also be noted that the second pie could be smaller than "intended," if high merit scores tend to go with low base salaries.)

The cost-of-living increase pie (PIE(1)) is fixed at a given value, and it cannot be changed (line 2220).

The various pies are calculated in lines 2190-3240. Note, again, that the third pie is a residual of the first two. The method used to calculate the second pie is to obtain the average salary increase that will be based on pure merit, and then to multiply this amount by the size of each department, so that we have a "percentage point pie" for each department that is allocated on the basis of merit scores. The second component of one's salary increase, then, is the merit score times the percentage point pie times the current salary. The second component of all salary increases is listed separately.

The residual flat-dollar pie, PIE(3), is allocated as follows: Proportional merit scores are multiplied by the proportional size of a given department, and this product is multiplied by PIE(3). This third component of all salaries is listed separately.

Salaries for the coming year are then calculated (lines 3410-3660) by adding, within each department, the current salary of a given faculty member, the cost-of-living increase, the percentage-based merit increase, and the flat-dollar increase. These salaries, followed by overall percentage increases in salaries, are printed out separately. Percentage-based merit increases and flat-dollar increases are also printed out separately. The salary-modification process for any individual can be traced by following that individual from one data matrix to the next. For instance, the upper-left entry for each matrix refers to the first individual listed for department 1, etc.

In any instance where the administrator wishes to over-reward a given department, this should be done by increasing the department's share of the residual flat-dollar pie (PIE(3)), primarily because the other two pies are considered sacrosanct. (It is helpful that the third pie is a relatively new invention.) At the point where the third pie is allocated among departments (lines 2930-2990), a given department may be over-rewarded by arbitrarily "adding" to its size. A 7-person department, for instance, deemed to have the productivity of an 8-person department, could be rewarded by merely raising its size constant, called k82, by one unit. (Fractional values may also be used.) This method has the effect of rewarding that department by giving it a larger part of the third pie, while allowing all other departments to absorb equally this instance of over-reward. The method has the virtue of making the administrator's action in rewarding or punishing given departments completely clear and understandable. Further, it can be used in a way that emphasizes the "carrot" rather than the "stick." That is, over-reward may be the only technique explicitly used, so that the most under-rewarded departments would be those that rarely receive an over-reward and are thereby constrained to help absorb a series of over-rewards going elsewhere.

The output listing for this program requires more than 500 lines; only a few excerpts will be shown here. I will provide the complete listing for anybody who wishes to see it. Notice that I have selected a year that reflects prosperous times.
 

.

.

.
 

MTB LET C14 = C4/K4 0920

MTB LET C15 = C5/K5 0930

MTB NAME C11 'MERIT1' 0940

MTB NAME C12 'MERIT2' 0950

MTB NAME C13 'MERIT3' 0960

MTB NAME C14 'MERIT4' 0970

MTB NAME C15 'MERIT5' 0980

MTB PRINT C11-C15 0990

1000

ROW MERIT1 MERIT2 MERIT3 MERIT4 MERIT5 1010

1020

1 0.166667 0.054422 0.0751879 0.0540541 0.0476190 1030

2 0.119048 0.088435 0.0075188 0.0720720 0.0714285 1040

3 0.190476 0.095238 0.0375940 0.0720720 0.0595238 1050

4 0.119048 0.068027 0.0375940 0.0540541 0.0714285 1060

5 0.095238 0.088435 0.0451128 0.0720720 0.0714285 1070

1080

6 0.095238 0.102041 0.0676692 0.0540541 0.0714285 1090

7 0.214286 0.061224 0.0375940 0.0450450 0.0714285 1100

8 0.061224 0.0601504 0.0540541 0.0595238 1110

9 0.102041 0.0300752 0.0450450 0.0476190 1120

10 0.054422 0.0526316 0.0720720 0.0714285 1130

11 0.068027 0.0751879 0.0450450 0.0952381 1140

12 0.068027 0.0676692 0.0540541 0.0833333 1150

13 0.088435 0.0300752 0.0720720 0.0952381 1160

14 0.0375940 0.0630630 0.0833333 1170

15 0.0526316 0.0540541 1180

16 0.0300752 0.0540541 1190

17 0.0676692 0.0630630 1200

18 0.0150376 1210

19 0.0075188 1220

20 0.0751879 1230

21 0.0526316 1240

22 0.0375940 1250

1260

MTB note 1270

MTB note 1280

MTB note : Check sums of merit scores, all equalling 1.0. 1290

MTB SUM 'MERIT1' 1300

SUM = 1.0000 1310

MTB SUM 'MERIT2' 1320

SUM = 1.0000 1330

MTB SUM 'MERIT3' 1340

SUM = 1.0000 1350

MTB SUM 'MERIT4' 1360

SUM = 1.0000 1370

MTB SUM 'MERIT5' 1380

SUM = 1.0000 1390

MTB note 1400

MTB note 1410

MTB note : Enter current salaries within each department. 1420

1430

ROW SALARY1 SALARY2 SALARY3 SALARY4 SALARY5 1440

1450

1 62929.9 84922.9 36398.4 66071.7 46698.4 1460

2 51338.3 46961.7 50337.2 64037.5 48377.1 1470

3 61650.1 66293.2 41396.6 37686.8 34901.7 1480

4 41456.6 58758.2 39825.8 58630.4 53902.1 1490

5 64132.4 84312.6 40422.4 44990.5 49313.2 1500

6 35187.0 45240.2 34364.0 45043.5 39992.6 1510

7 42021.3 54404.1 52475.2 56859.3 41594.7 1520

8 54625.3 52219.6 64107.7 40664.6 1530

.

.

.
 

: List salaries for next year. 3410

3420

ROW SALARY11 SALARY12 SALARY13 SALARY14 SALARY15 3430

3440

1 73964.2 92842.0 47438.5 74735.9 52185.8 3450

2 58689.2 55850.5 52367.4 75116.7 56328.6 3460

3 74007.3 77701.8 47657.6 46203.7 40774.3 3470

4 48153.9 66614.2 45983.7 66751.9 62292.1 3480

5 70855.2 96609.4 47693.7 54217.6 57339.0 3490

6 40378.3 55188.3 44142.3 52174.1 47278.6 3500

7 53483.2 61282.1 59463.4 63657.2 49007.9 3510

8 61517.2 62878.5 72628.6 46918.4 3520

9 61977.0 48125.4 66570.5 66218.8 3530

10 77694.9 52594.6 67568.5 60594.5 3540

11 73148.0 67485.2 67882.7 64605.0 3550

12 86739.6 50916.6 77064.2 39546.3 3560

13 60075.9 59623.8 73797.7 51764.8 3570

14 56883.2 53412.9 82666.0 3580

15 44460.9 73848.0 3590

16 44016.4 72838.7 3600

17 61912.1 70901.3 3610

18 52270.2 3620

19 45753.1 3630

20 66946.7 3640

21 43249.6 3650

22 44476.3 3660

3670

3680

K91 4399996 3690

3700

3710

: Check that k91 equals k31, that is, total of all 3720

: salaries equals the total budget. 3730

3740

3750

: Calculate total percentage increases for each dept. 3760

3770

ROW C71 C72 C73 C74 C75 3780

3790

1 17.5343 9.3251 30.3313 13.1133 11.7507 3800

2 14.3186 18.9277 4.0330 17.3011 16.4365 3810

3 20.0441 17.2092 15.1244 22.5990 16.8261 3820

4 16.1551 13.3701 15.4620 13.8520 15.5653 3830

5 10.4826 14.5848 17.9883 20.5090 16.2751 3840

6 14.7534 21.9893 28.4550 15.8305 18.2185 3850

7 27.2764 12.6424 13.3172 11.9557 17.8224 3860

8 12.6168 20.4117 13.2916 15.3789 3870

9 20.6193 12.2488 11.6848 10.3165 3880

10 10.0789 19.7320 18.2130 15.7945 3890

11 12.7723 24.6984 11.5708 20.2770 3900

12 11.8302 25.9612 12.9282 23.8527 3910

13 18.1810 10.8278 17.4460 22.9507 3920

14 13.6439 18.1264 15.7859 3930

15 21.8848 13.1870 3940

16 12.9497 13.2733 3950

17 23.1933 15.5990 3960

18 6.5240 3970

19 4.4384 3980

20 24.8010 3990

21 22.2829 4000

22 15.7897 4010

.

.

.

Chapter 6

CULTURAL MATERIALISM IN THE SYSTEMIC MODE
 
 
 
 
 
 
 
 
 
 
 

The first part of this chapter suggests that functional analysis, also known as general systems theory (Turner, 1991:93-130), remains one of the more vigorous theoretical traditions of the contemporary social sciences. "Cultural" functionalism, as understood by theorists such as Thompson, Ellis, and Wildavsky (1990), has many features in common with recent versions of functionalist theory, and it is possible to combine these approaches in a way that resolves many of the alleged weaknesses of the functionalist paradigm. The second and third sections develop the thesis that the systemic emphasis of demography and human ecology, which involves interaction among the POET variables (population, organization, environment, technology), provides a natural starting point for an integrated functionalist/systemic mode of analysis. A partly counterfactual study of the history of the American railroads provides an extended illustration of the utility of the neofunctionalist point of view.
 

(1) An excursion into "cultural" functionalism
 

--Paul M. Lloyd (1987:39)
 

Toward the end of 1988 I received a manuscript by Thompson, Ellis, and Wildavsky--hereinafter referred to as TEW--on the history of functional analysis and its prospects for the future.¹ I found this work to be most provocative for several reasons: First, what these authors say regarding the strengths and weaknesses of functional analysis closely parallels my own arguments in Dynamic Functionalism (Faia, 1986)--referred to below as DF. I believe that TEW and I have arrived independently at similar conclusions about how to resolve difficult questions of teleology, requisite analysis, social change, etc., within a contemporary functionalist framework. Second, we have arrived at a common understanding of functional analysis despite the fact that we have sharp disagreements about the proper central orientations of the social science disciplines. In Wildavsky's work, especially that done in collaboration with Mary Douglas, I discern a large emphasis on social constructionism, a highly ideational and mentalistic approach to topics such as the ways in which environmental hazards threaten human health and tranquillity (Wildavsky, 1988). My own preferences tend toward the "cultural materialist" perspectives of scholars such as Marvin Harris (1979). As in the case of Harris, I strive to reconcile my work with the Marxian traditions, and this effort is apparent throughout DF.

In the rest of Section (1) I shall make a series of comments on various parts of TEW's work on functionalism. I believe that most of these comments will elaborate and elucidate the claims made above.
 

(1.1) An emergent consensus
 

There are at least four major points on which TEW and I agree:

First, TEW say (1990:105) that in recent years it has become fashionable to dismiss functionalism as static, conservative, tautological, irrelevant, and so forth, with little or no justification. Some critics, they maintain, have tried to account for the dominance of the functionalist perspective by pointing to its alleged utility as an ideology of rule, to its role in maintaining the "colonial commitments" of various Western nations, or to its role in maintaining the "social legitimacy" of the middle class (1990:105). These critics, in other words, have offered a functionalist explanation for functionalism itself--an early instance of reflexive sociology that was surely not committed to maintaining the intellectual status quo. I mention a similar irony in DF (1986:21), where I review Kingsley Davis' Presidential address before the American Sociological Association: In the process of criticizing functionalism, Davis (1959:768) offers a functionalist theory to explain the rise of functionalism as a highly successful academic pursuit. TEW find the irony delightful, I suspect, because they believe, as I do, that most critics of functional analysis end by conducting precisely the sort of analysis that they claim to eschew. In effect they are closet functionalists, and they themselves may not comprehend the truth about what they have become.

Second, TEW (1990:107) and I share the belief that the old lists of functional "requisites" were trite, and surely did not further the interests of those who advocated functionalist theory. In the opinion of TEW,
 

I argue in DF (Chapter 1) that traditional requisite analysis should be abandoned altogether, regardless whether we analyze whole societies, ecosystems, organizations, cultural traits, personalities, or whatever. In working toward this conclusion I found the banality argument to be highly compelling, but I was far more firmly convinced by Fogel (1964), who provides an excellent illustration of how one might go about testing the hypothesis--more accurately, the conventional wisdom--that in nineteenth-century America the railroads were "indispensable." As it turns out, Fogel does not really indulge in requisite analysis. His work has immense value, however, for functionalist/systemic sociology, and it will be discussed in detail in a later section of this chapter. The general utility of Fogel's style of counterfactual history, especially for social scientists interested in testing functionalist hypotheses by means of computer simulations, is discussed in Faia (1989a).

Third, TEW's section on Comte argues that a major difficulty of his work was his insistence that whole societies be taken as the dependent variable, for most social science analyses. But it is precisely those forms of functional analysis that focus solely on whole societies, contend TEW (1990:114), that produce banalities (the manuscript version says "tautologies") such as the lists of requisites mentioned earlier. It must be pointed out in his favor, however, that Comte advocated a nomothetic approach to social analysis, as contrasted with Montesquieu's drift toward idiographies (TEW, 1990:113), and that Comte also advocated careful attention to taxonomies despite his interest in global social entities. It appears, then, that Comte maximizes our opportunities to pursue either the general systems form of functionalism or TEW's brand of culture-functionalism: Both structures and cultural traits will be found at various locations in a good taxonomy,² and it is encouraging that TEW recognize (1990:103) that sociology gets into lots of trouble when it fails to clarify its typologies, i.e., its taxonomies.

Fourth, TEW include both Montesquieu and Weber as major exemplars of the functionalist traditions. For the early years I tend to cite Smith and Malthus rather than Montesquieu, but I never had the courage to claim Weber explicitly--despite the fact that my intuitions have always told me to do so. I don't know Montesquieu well, but I once found in The Spirit of the Laws what I regard as an excellent functional theory about why parliaments tend to be heavily populated by lawyers. Weber's significance in the history of functionalism seems to tie in with my earlier remarks about ironies: There are major Weberian scholars abroad at this moment who pride themselves on having moved beyond the outmoded traditions of functionalist theory, and who yet do not understand, as TEW do, that Verstandnis may be pursued effectively within a functionalist framework. Sica (1988:210), for example, discusses Weber's insights into the persistence of narrow managerial control of firms, but does not recognize these insights as being essentially a functionalist theory.
 

(1.2) Sub-cranial worlds and other pitfalls
 

I turn now to the major issue on which TEW and I do not agree--their overemphasis on Mind, and on socio-psychological phenomena.

As I suggest above in reference to Weber, I am bothered by TEW's strong focus on mental events, e.g., their ubiquitous arguments about "consensus" and about its alleged effects in keeping society together. These arguments start with the discussion of Comte and continue in the Parsons section. In the latter the authors assert that the United States coheres well despite the fact that there are major forms of dissensus. But one could argue, for instance, that even the most vehement pro-life or pro-choice advocates in the United States have far more in common than is implied by the fact that they may be assaulting each other at the entrances of abortion clinics. They have a common culture. They may cooperate through various forms of exchange such as impersonal market transactions, by forcing legislatures and courts to seek a popular consensus about means of fertility regulation, by encouraging researchers to seek more effective means of contraception, by signalling their own intention--conscious or otherwise--to use a confrontational style that will avoid violence, etc. TEW do not address this sort of question adequately; their analysis implies that the issues are so complex as to be beyond solution.

After reading the TEW section on Marx (1990:147-60) I am convinced once again that an etically-oriented cultural materialism, shorn of Hegelian notions about inexorable dialectical forces and allowing an important (though not predominant) role for Verstandnis, provides a grist of high quality for the functionalist mills. It is encouraging that Harris usually describes himself as a non-dialectical materialist, and it is also encouraging to realize that TEW (1990:154) overgeneralize when they say that Marx and Marxism posit a universal "world-historical" triumph of capitalism: It is entirely possible, for instance, that capitalism will not triumph, despite a strong attempt, as a feature of Mexican agriculture (Dumont and Mottin, 1982:17-78). And my volume on functionalism contains many examples that raise doubts about the TEW assertion (1990:148)--really a postulate--that "... assigning primacy to technological-material forces is misleading. ... The human controls the nonhuman, not the reverse. Technologies come and go, but the viable ways of life are always with us." One of my favorite contrary examples occurred to me recently: There are interactions now taking place among the earth's atmosphere, biosphere, and hydrosphere that humans clearly do not control despite the fact that they involve technology, and these interactions, perhaps creating "greenhouse effect," could ultimately thrust us into grave difficulties. Interactions of a global magnitude can be understood properly from a functionalist point of view, and they may involve both equilibria and disequilibria.

I would not have mentioned this example, however, if I had not discovered that Douglas and Wildavsky (1982:63) cite with approval the work of James Lovelock, an atmospheric scientist. Lovelock may think that the goddess Gaia will ultimately protect us, but his theories and those of other scholars such as Stephen H. Schneider and Randi Londer (1984) provide many examples in which the human psyche clearly is not in the driver's seat, in which the Parsonian cybernetic hierarchy is inoperative, and in which there are objective risks that may result in massive eco-catastrophes whether or not we distort these risks through social construction. I fear that Wildavsky, Douglas, Parsons, and Lovelock all have been inspired indirectly by Bishop Berkeley, and that at the apex of their cybernetic hierarchies we find the Deity; we find, as TEW say in the manuscript version of their work (1988, Parsons section, p. 2), that in the Parsonian scheme as in that of Parson Malthus our standards seem to "... descend from on high ..." Incidentally, Schneider and Londer (1984:235) say that Lovelock's work has a "... philosophical--perhaps theological--significance ...," and one can only hope that the Gaia hypothesis will evolve into a formulation of greater scientific utility. A fine example of this sort of evolution is provided by James (1987a; 1987b), who points out that "... wartime rises in sex ratio ... have hitherto defied explanation: indeed it has been acknowledged that divine intercession is as good an explanation as any ..." (1987a:734), but who then suggests (1987b:876) that high sex ratios during wartime may be associated with "the coital excesses of returning servicemen ..." The latter hypothesis strikes me as being far more down to earth.

There is a similar issue, raised in TEW's Marx section. Under socialist theory, contrary to TEW's view (1990:150), it is not anticipated that we will witness "... the replacement of latent functions with manifest functions." Cubans, for instance, must study sugar production functionalistically; if they do not do so, the resulting ignorance will invariably get them into trouble. I anticipate that socialist revolutions (and "free-market" revolutions), wherever they occur, eventually will create many job opportunities for functionalist/systemic/marxist sociologists; there will surely be enough false consciousness to go around, and substantial numbers of latent functions will be discovered. In B. F. Skinner's Walden Two by contrast, very few functions are ever made manifest: Only Skinner's alter-ego, sitting on a hillside above the idealized community, understands how things work--and he (the character, if not Skinner) keeps all insights to himself. We Marxian functionalists intend to be a part of the community, and we hope to be teachers. Given their sophistication, TEW will surely wish to join us.

It is clear, then, that we are dealing with differences of degree. Those who are exploited by the ruling class have only a partial understanding of class-based ideologies of rule: They do not understand how these ideologies result in that very exploitation of which they are the victims. This misunderstanding is called false consciousness, and around page 7 of DF I provide many examples of functions manifest primarily to elitists. Marxists have no qualms against saying that "wants" may not be the same as "needs," and that we may have a sense of fulfilling wants while at the same time not realizing the extent of misery due to the frustration of basic human needs. These frustrations are not unlike risks: They can threaten us even when we do not recognize their existence. Presser (1974) and Zelnik (1979), for instance, make it clear that many teenagers are highly at risk of unwanted pregnancies without knowing it and without knowing the degree of misery that may be theirs as a consequence of these pregnancies.

In TEW's discussion of Merton (1990:199-200), the question of manifest versus latent functions is confused even further:
 

Why should feedback loops provide a special problem? Feedback loops involve reciprocal causation, and reciprocal causation has been reasonably well understood by statisticians since the days of the geneticist Sewall Wright. We should all learn the statistical theory a lot better, but the theory is definitely in place. Beyond that, DF contains any number of instances in which circular causation operates independently of actors' intentions and perceptions, in which functions remain essentially latent. TEW's argument would make no sense to the linguist who provides the epigram at the head of this chapter. These authors themselves--irony again--provide an excellent illustration: Distorted comparisons of the United States and, say, the Soviet Union may serve various political interests (i-functions) and may also have something to do with maintaining political stability and the survivorship of various social arrangements (s-functions). And scholars such as TEW, producing and perpetuating these distorted comparisons, may receive positive feedback, positive reinforcement, from several sources including the political interests served. Stryker (1990:694-95) provides an excellent example, pointing out that
 

Thus, we have a feedback loop that is not generally recognized, and it is therefore an instance of a latent function.

It is therefore inappropriate to claim, as do TEW (1990:169), that The Protestant Ethic is non-functionalist because it is concerned primarily with unintended consequences. Stryker's study deals with unintended consequences, and it is explicitly functionalistic. Functionalist explanations do not presuppose intentions, and they do not presuppose that anybody (beyond a few social scientists) knows about a given function. Intentions and understandings may not exist at all, they may exist in the unconscious minds of some members of society in a form that psychologists would call unconscious affectivity or unconscious cognition, or they may be largely manifest. Both manifest and latent functions, of course, are ideal types: As for the former, no knowledge is perfect or shared universally; as for the latter, the moment we discover a latent function and begin to talk and write about it, it is no longer latent by definition. But for long periods of time the etic knowledge of the social sciences may remain known only to a few--those who are lucky enough to be social scientists.

I allude to Freud and the unconscious because a large part of this discussion reminds me of that ancient staple of graduate seminars, the Freud versus Malinowski debate. My recollection is that whenever Malinowski demanded that competition between fathers and sons be explained in sociological terms, Ernest Jones (representing Freudian theory) would insist that the oedipus complex, instinctively based, is an adequate explanation. When Malinowski then demanded evidence for the existence of this instinct, Jones invoked a sliding infinity of regressions, maintaining that the oedipal instinct may be buried so deeply in the unconscious that it should be regarded as an unobservable. Malinowski, of course, called it a metaphysical construct. We fledgling social scientists, while realizing the manifest and latent functions of siding with Malinowski, may actually have found his explanation to be the more plausible. In any case the debate made it clear that social structures, although they must be demonstrable in some way, may not be widely recognized and need not be motivated by conscious or unconscious human volition. As Harris and Ross show (1987), we learn next to nothing about female infanticide as a potential fertility regulating device when we ask people about the motives and understandings that underlie this practice.

In their Stinchcombe-Elster section (1990:200-209), TEW surprise their readers by abandoning the notion that all functions must have elements of intention and/or understanding. Elster seems to support their position, but Stinchcombe completely undermines it with the concept of social selection modelled on natural selection. Natural selection works entirely without intention or design, and social selection is alleged to have the same characteristic. TEW have trouble with this idea, and their position becomes highly ambiguous: They discuss (1990:202-03) a hypothesis about the "survival" of members of society who either do or do not engage in what Veblen called conspicuous consumption, and it is implied that people in this situation do not always realize that survival--whatever the term means in this context--is at stake. In DF, as I've said, there are many examples of functionalist explanations that do not involve affect or cognition. And in a discussion to follow, I argue that Fogel's (1964) study of the railroad and its alternatives, in nineteenth-century America, arrives at a functionalist explanation that cites intentions and understandings primarily to show that they had little to do with the realities of the situation.

Although the TEW history does not include Pareto, whom I find to be of immense value as a source of functionalist ideas and hypotheses, it does present a discussion of Spencer. As I suggest in an appendix to DF, Spencer remains excellent reading for those who wish to get a handle on early modern functionalist theory. Again, since I consider the search for "feedback loops" to be a crucial aspect of functionalism, I consider Spencer's formulations to be a crucial part of the history of this form of theory. As an example of Spencerian functionalism, TEW (1990:118) cite the hypothesis that societies with high levels of sexual promiscuity have tended to produce fewer and weaker offspring, and have tended as a result to disappear. Whether or not this hypothesis is correct, we encounter here the idea of dysfunctions that remain latent, unrecognized, unmotivated, and we also encounter the idea of social selection as a form of feedback. It should be remembered that Spencer regarded natural selection ("survival of the fittest") and social selection as more or less identical processes, and he therefore implies that we need not worry unduly over questions of teleology and/or motivation. Social selection can be latent in the same way that natural selection is almost invariably latent.
 

(2) The POET paradigm: A transcendence of hierarchy³
 

Although Namboodiri's (1988) defense of classic ecological demography is highly compelling, I believe that he misses several opportunities to demonstrate even more strongly the power and utility of this neglected perspective. My own position on this question has been worked out in some detail in a separate paper (Faia, 1989b), and holds briefly that the strength of ecological demography lies in its having developed conventional definitions--what Levine (1985) would call univocal definitions--of many of its key concepts; in its ability to exploit deterministic models that explain essentially all the variability of certain social phenomena, as in the explanation of labor force composition (and its changes) with reference to fertility, mortality, migration, and social mobility; in its sophisticated level of theoretical development, involving Malthusian and neo-Malthusian perspectives, Marxism, optimal population theories, demographic transition theory, the stable population model, various socio-psychological theories of vital phenomena, decision theory, functionalist/systemic analysis, and so forth.

The POET acronym refers to the interaction of population, social organization, environment, and technology. In cause-effect terms, interaction may run both ways: Climate, for instance, may affect social organization, and social organization may have an impact on climate (see, for instance, Schneider and Londer, 1984). And these interactions, of course, are not limited to two elements. Phenomena such as greenhouse effect involve complex interactions among several population variables, several organizational variables, and many environmental and technological factors. If we assume that ecological variables may be selected from any combination of the four POET categories, we discover that there are 25 areas in which we might develop ideas about multivariate interactions. Human population, for instance, might affect O, E, or T; or human population might affect an interaction of O and E variables, or O and T variables, or E and T variables; or it might affect an interaction involving variables drawn from all three of the other sectors. When all such interaction prospects are added up, they total 25.⁴

The two-way character of social and ecological interaction is a matter that cannot be ignored.⁵ In general, it implies that no variable has a predominant role as a cause, and none acts consistently as an effect. In fact, the reference above to the many multivariate interactions among the POET variables implies that, in many cases, we are not explaining things from causes to single effects: We might, for instance, try to determine whether an alleged causal agent (or several causal agents) has an effect on the interaction of two other variables. If we are interested in the interaction of the human population with computers, for instance, the idea of two-way causation reminds us that we must not assume a priori, with TEW, that "... the human controls the non-human ..." From an ecological standpoint it is true that families identified by the Bureau of the Census (1988) as having personal computers in their homes may have acquired these machines because "they wanted to do so," but members of these families will probably be changed by computers in ways that they cannot anticipate. If we ask, for instance, how the nature and types of computer use in a community's schools and offices influence the relationship between sex and computer use among the children of these families, we are raising a question that renders the TEW claim unintelligible.
 

(3) Reciprocal interaction: Organization and technology
 

Fogel's study (1964) of the American railroads is a prime illustration of the preceding arguments: It shows us (1) how functionalist/systemic analyses are conducted when they avoid the vagaries of requisite analysis; (2) what the POET paradigm is able to accomplish when it is applied to a large ecosystem; (3) how simulation models help us to understand social development, i.e., the ways in which ecological systems select among structural alternatives.
 

(3.1) Wagons, waterways, and William F. Ogburn
 

Tables 1 and 2, based on data available in Fogel (1964, Tables 3.7, 3.12, 3.13, and 3.14), provide a summary of the study's main arguments. Fogel's intention was to measure the economic impact of the railroads as compared with hypothetical (i.e., counterfactual) canals and wagons, which could have been constructed instead of railroads. Counterfactual history, incidentally, makes heavy use of phrases such as "could have been" and "would have been," and these expressions capture clearly the essence of this approach to historical analysis. As Fogel (1964:10) says, "evaluation of the axiom of indispensability ... requires not only an examination of what the railroad did but also an examination of what substitutes for the railroad could have done." TEW would be pleased by the idea of substitutability, which necessarily raises questions about the existence of prerequisites: If tea makes us just as happy as coffee, then coffee is not a prerequisite.
 

Table 1. Replication of Fogel's Estimate of the Loss in National Product due to Substitution of Canals for Railroads (thousands of dollars)
 

========================================================================
 

C1 C2 C3 C4 C5 C6 C7
 

State Value of Value of C2 Value of C4 Loss in Loss in

or farmland farmland as % farmland as % national national

Division 1890 beyond C1 beyond C1 product, product,

feasible feasible before after

region, region, canals canals,

before after (C4/C2)*C6

canals canals
 
 
 

N-At-Div 1092281 5637 0.52 5637 0.52 331 331.00

S-At-Div 557399 117866 21.15 117866 21.15 8452 8452.00

Ohio 666118 20205 3.03 20205 3.03 1332 1332.00

Indiana 481684 38783 8.05 18150 3.77 2587 1210.69

Illinois 964612 174567 18.10 0 0.00 11696 0.00

Michigan 301739 106110 35.17 61 0.02 7449 4.28

Wisconsin 264049 53483 20.25 16636 6.30 3610 1122.90

Mnnesota 254261 81649 32.11 12023 4.73 6499 956.99

Iowa 678364 372339 54.89 0 0.00 27590 0.00

Missouri 443611 68891 15.53 1260 0.28 5539 101.31

N. Dakota 51257 19483 38.01 19483 38.01 1808 1808.00

S. Dakota 79676 43114 54.11 3267 4.10 3898 295.37

Nebraska 310302 187149 60.31 21373 6.89 15328 1750.51

Kansas 435934 276179 63.35 29798 6.84 23143 2496.99

Kentucky 178670 1893 1.06 1893 1.06 119 119.00

Tennessee 123633 7936 6.42 7936 6.42 475 475.00

Alabama 32795 1223 3.73 1223 3.73 102 102.00

Mississipp 51166 342 0.67 342 0.67 33 33.00

Louisiana 42901 0 0.00 0 0.00 0 0.00

Texas 265841 145248 54.64 73854 27.78 13261 6742.80

Arkansas 43327 2224 5.13 2224 5.13 201 201.00

West-Div 800952 218216 27.24 218216 27.24 19919 19919.00
 

U.S. Total 8120572 1942537 23.92 571447 7.04 153372 47453.83
 

Tabl-3.13 x 000
 

Prelim 248100

Chg in L -105918

Chg a-bar 16100

Chg ind-c 16600
 

Total 174882 Pct GNP 1.46
 

========================================================================
 

Source: Fogel (1964, Tables 3.7, 3.12, and 3.13)
 

The part of Fogel's analysis that I wish to discuss deals with the intraregional distribution of agricultural products, i.e., how farmers got their livestock and crops to the primary markets in cities such as Chicago or St. Louis. Interregional distribution systems--from, let us say, a primary market in Chicago to a secondary market in Boston--and non-agricultural products are treated elsewhere in Fogel's book. Each of my tables provides data for the several states and divisions indicated in the leftward column.
 

Table 2. Replication of Fogel's Estimate of the Effects of New Canal Construction and Reduced Wagon Rates on Loss in National Product (thousands of dollars)
 

==================================================================
 

CR4 CR5 C6 CR7
 

Reduced CR4 Loss in Reduced

C4, as % national loss in

due to C1 product, national

improved before product,

wagon canals canals and

rates, improved

.5*C4 road tran.,

(CR4/C2)*C6
 

2819 0.26 331 165.50

58933 10.57 8452 4226.00

10103 1.52 1332 666.00

9075 1.88 2587 605.34

0 0.00 11696 0.00

31 0.01 7449 2.14

8318 3.15 3610 561.45

6012 2.36 6499 478.50

0 0.00 27590 0.00

630 0.14 5539 50.65

9742 19.01 1808 904.00

1634 2.05 3898 147.69

10687 3.44 15328 875.25

14899 3.42 23143 1248.49

947 0.53 119 59.50

3968 3.21 475 237.50

612 1.86 102 51.00

171 0.33 33 16.50

0 0.00 0 0.00

36927 13.89 13261 3371.40

1112 2.57 201 100.50

109108 13.62 19919 9959.50
 

285724 3.52 153372 23727
 

Tabl-3.14 x 000
 

Prelim 248100

Chg in L -129645

Chg a-bar -18600

Chg ind-c 17900
 

117755 Pct GNP 0.98
 

==================================================================
 

Source: Table 1 and Fogel (1964, Table 3.14)
 

I shall define and discuss the columns of Tables 1 and 2 for the state of Indiana (fourth data row) because it appears that new canals and better wagon transport would have had a relatively large impact on this state. For Table 1:
 

For Table 2:
 

The same data are provided for all other states and regions listed in the tables. At the foot of Table 1 is Fogel's Table 3.13, which calculates the proportion of national product that would have been lost if canals had replaced railroads. At the foot of Table 2 we find calculations for Fogel's Table 3.14, which shows the proportion of national product that would have been lost if canals and better road transportation had replaced railroads.

Let me anticipate my explanation of these calculations by saying that Fogel regards a .98 per cent reduction of national product--the bottom line of Table 2--as trivial, since national product was growing by something like 3-5 per cent per year. His general conclusion, then, is that the United States would have had to absorb a lag in economic growth of only a few months' duration if the railroads had been replaced (or displaced) by wagons and canals (Fogel, 1964:47n). Let me also anticipate another point: In Tables 1 and 2 most of the columns contain raw data, the only exceptions being the columns with percentages and columns C7, CR4, and CR7. This means that, using a spreadsheet, we might experiment with changes in columns C1, C2, C4, and C6. And since CR4 involves a formula with a multiplier (.5) that represents Fogel's rough estimate of the effects of cheaper wagon transportation, this formula also would allow simulation experiments. In a later section, I'll present an experiment involving CR4 of Table 2.

Fogel's (1964:100,110) calculations for Tables 3.13 and 3.14 assume that substituting the existing waterways, as of 1890, for railroads would have cost the nation (according to his "beta" estimate) about 221 million dollars, plus another 27 million dollars due to additional cargo losses, slowness, and the seasonality of water transportation. (In my table, the total comes out to 248.1 million dollars.) Some of these losses, however, could have been avoided by expanding the canal system. In Fogel's Table 3.13, losses are thereby reduced by 106.1 million dollars; due to rounding error, my figure is 105.9 million dollars. After we add small adjustments for changes in the size of the feasible farming region and for additional indirect costs, the substitution of waterways for railroads costs about 175 million dollars per year, or about 1.46 per cent of the national product of around 12 billion dollars. Similar calculations for Table 3.14, assuming new canals and cheaper road transportation, place the loss in national product at 118 million dollars, around 1 per cent of the national product.

But there is a big question here: How cheap could highway transportation have become? William F. Ogburn (1964) showed that when technological innovations diffuse through a society, they are not used in exactly the same way by all adopters. For instance, once the internal combustion engine became readily available and highly efficient, an infinitude of unanticipated uses for it were discovered. Similarly, an innovation such as expanded inland and intercoastal waterways would have had ramifying effects on other parts of the North American ecosystem. Fogel (1964) alludes to this possibility in several parts of his book: On page 15: "The axiom of indispensability proceeds on the implicit and unverified assumption that the success of the railroads did not choke off the search for other solutions to the problem of overland transportation." On page 20, in purely Ogburnian phrases: "Forcing the pattern of shipments in the non-rail situation to conform to the pattern that actually existed is equivalent to the imposition of a restraint on society's freedom to adjust to an alternative technological situation." Page 91: "The discussion thus far has been based on the assumption that in the absence of the railroad all other aspects of technology would have remained fixed in the 1890 pattern. This is a severe assumption." And the following twenty pages or so discuss the prospect of building better waterways and roads, a more rapid development of the internal combustion engine, etc.

Often designed to be counterfactual, simulations allow us to explore these hypothetical prospects. In one series of experiments I modified the data for four areas listed in Table 1 where there seemed to be little chance of improved waterways (column 5). Each of these states or regions (South Atlantic Division, North Dakota, Texas, and the Western Division) had large areas of farmland remaining without access to waterways even after a substantial (hypothetical) improvement of the waterways. I found that under highly optimistic assumptions about the possibility of new canals in these four areas, Fogel's estimate of a 1.46 per cent loss in national product would have been reduced to about 1.19 per cent--not an impressive degree of elasticity. In another experiment, summarized in Table 3, I show what would have happened to the national product if the multiplier used in column CR4 were changed from .5 to .1 as a result of a larger reduction in the cost of wagon transportation than that contemplated by Fogel--again, this is a highly Ogburnian experiment. Such a large reduction would have allowed farmers to get their products to the nearest waterways at a relatively low cost. Examining Table 3, we see that this change would have reduced losses in national income from .98 per cent to .82 per cent. This is a surprisingly small change, and it suggests that extraordinary improvements in roads and in the efficiency of internal combustion engines would have had a small impact on intraregional agriculture as a component of national income. To be realistic, however, cost-benefit calculations for intraregional agriculture would have to be combined with similar calculations for other segments of the economy dependent on efficient short-distance hauling.
 

Table 3. Replication of Fogel's Estimate of the Effects of New Canal Construction and Reduced Wagon Rates on Loss in National Product (thousands of dollars), With Further Reductions of Wagon Rates
 

==================================================================
 

CR4 CR5 C6 CR7
 

Reduced CR4 Loss in Reduced

C4, as % national loss in

due to C1 product, national

improved before product,

wagon canals canals and

rates, improved

.1*C4 road tran.,

(CR4/C2)*C6
 

564 0.05 331 33.10

11787 2.11 8452 845.20

2021 0.30 1332 133.20

1815 0.38 2587 121.07

0 0.00 11696 0.00

6 0.00 7449 0.43

1664 0.63 3610 112.29

1202 0.47 6499 95.70

0 0.00 27590 0.00

126 0.03 5539 10.13

1948 3.80 1808 180.80

327 0.41 3898 29.54

2137 0.69 15328 175.05

2980 0.68 23143 249.70

189 0.11 119 11.90

794 0.64 475 47.50

122 0.37 102 10.20

34 0.07 33 3.30

0 0.00 0 0.00

7385 2.78 13261 674.28

222 0.51 201 20.10

21822 2.72 19919 1991.90
 

57145 0.70 153372 4745
 

Tabl-3.14 x 000
 

Prelim 248100

Chg in L -148627

Chg a-bar -18600

Chg ind-c 17900
 

98773 Pct GNP 0.82
 

==================================================================
 

Source: Spreadsheet experiment based on Table 2
 

(3.2) A functionalist overview of an ecosystem
 

Fogel's work provides strong support for TEW's (1990:114) suggestion, in reference to Comte, that functionalists are not likely to carry out their most effective analyses when they focus on whole societies (cf. Tilly, 1984:11-12) and the conditions under which these societies survive. It is arguable, of course, that the United States could not have achieved the degree of social integration necessary for nationhood without an adequate interstate and intrastate transportation system, but this insight has little value. The real question is: What sort of system, in what part of the country, why, when, and under whose ownership and control? As Fogel works at answering this question, he seems to be saying that while the railroads served the national interest well, there were no highly compelling reasons for selecting them rather than an alternative transportation system emphasizing some combination of waterways and roadways. Not only was the nation as an ecological entity not at stake as a result of this selection, but the economic consequences of the selection, while not at all clear, do not seem to have been critical: One can hardly argue that railroads were selected because of powerful economic advantages to the nation as a whole. In reference to specific interest groups, as we shall see, the story may be very different.

As we try to explain this selection--rather, the many selections involved in this historical process--it would be appropriate to examine public opinion, what I have called elsewhere the POETry of