Durkheim and Weber are commonly and correctly regarded as two of the foremost comparative analysts in the history of sociology. In their work they faced a number of common problems that arise in comparative analysis, and attempted to overcome them in ways that are still instructive. I shall examine the strategies of these two scholars in action in this chapter. Both these men made their contribution to sociology in the late 19th and early 20th centuries, when the field was making significant strides toward establishing itself as an academic and scientific discipline. Both of them, moreover, had occasion during the course of their careers - Durkheim in 1895 and Weber in 1904 - to produce major theoretical and methodological statements on the program for sociology. Each statement was incomplete in many ways; for example, while both theorists assigned comparative sociological analysis a central place in their programs for sociology, neither developed a detailed, explicit statement of strategies for comparative analysis. Nevertheless, their reflections, considered together, expose the major methodological dilemmas encountered in comparative analysis. Their methodological writings are further instructive in that while they began with methodological perspectives that were radically opposed to one another, each made a number of significant modifications of these starting points in the course of his argument. As a result, their practical programs for sociological investigation - to say nothing of their actual empirical research - resemble one another much more than their methodological perspectives.

In this chapter, then, I shall examine the methodological contributions of Durkheim and Weber, with an eye to identifying certain general issues in comparative sociology which recur, and which form a basis for discussing strategies for comparative analysis that have been developed more recently. More particularly, I shall compare and contrast Durkheim and Weber under the following headings:

(1) The character of scientific knowledge and its relation to other kinds of knowledge and cultural values;

(2) The appropriate range of data to be investigated by sociologists;

(3) Classification in sociological investigation;

(4) The nature of sociological explanation; and

(5) Verification in sociology.

Let me begin with Durkheim, whose program of sociological positivism was laid out clearly and forcefully. While insisting that the subject matter of sociology is distinct from that of other sciences, Durkheim also insisted that the sociologist should approach his subject matter "in the same state of mind as the physicist, chemist, or physiologist when he probes into a still unexplored region of the scientific domain." Regarding the social sciences of his day as analogous to alchemy before the rise of the natural sciences, he condemned them as having dealt "more or less exclusively with concepts and not with things" By this he meant that sociologists had approached their subject matter with some abstract notion - such as evolution in mind and attempted to ascertain how various social arrangements suit those notions. Using such preconceptions, Durkheim argued, instead of "observing, describing, and comparing things, we are content to focus our consciousness upon, to analyze, and to combine our ideas. Instead of a science concerned with realities, we produce no more than an ideological analysis."

The proper strategy for sociology, Durkheim continued, is to cast aside such preconceptions and regard social phenomena as "distinct from the consciously formed impressions of them in the mind". The most important characteristic of a "thing," moreover, is "the impossibility of its modification by a simple effort of the will." The investigator should free his mind of all preconceptions, take a more passive relationship to social reality, and deal with phenomena "in terms of their inherent properties," and their "common external characteristics" Classifications should not "depend on [the sociologist] or on the cast of his individual mind but on the nature of things".

Durkheim's positivism is understandable as an expression of his impatience with unfounded and unverified theories of his day, and as a strategic appeal for empirical observation. Yet as a general methodological program, it evidently presents serious problems. The decisive problem concerns the possibility of ridding oneself of all preconceptions and letting the real world of empirical phenomena speak for itself. Aside from the fact that this is psychologically impossible for anyone who has been socialized in a language and in a way of regarding the world, Durkheim's position would seem to involve a logical impossibility as well. Given the complexity of empirical reality, and given the innumerable ways it may present itself, how is it possible to perceive a single set of external characteristics without actively selecting from among all the possibilities? And if the necessity of selection is acknowledged, does not this imply the necessity of some preconception on the part of the investigator? Without criteria for selecting aspects of the empirical world for observation and classification, is not the investigator left in a position of methodological paralysis, unable to begin?

Considerations such as these led Weber to a contrasting formulation of the character of scientific knowledge, though he did not offer his formulation as a direct polemic against Durkheim. Weber regarded scientific knowledge of society and culture as emanating from a number of "one-sided" (that is, selective) views of different aspects of cultural life. It was by selecting, over-emphasizing, and simplifying certain aspects that bodies of scientific knowledge - like formal economics - were generated. Furthermore, he argued that "the one-sidedness is intentional". More important, selectivity is not determined by the "nature of things," as Durkheim held, but by the initiative of the investigator.

Reality, even a single object, is so complex" it presents an infinite multiplicity of successively and coexistently emerging and disappearing events, both 'within' and "outside' ourselves" - that Weber was led to a firm conclusion: "All the analysis of infinite reality which the finite human mind can conduct rests on the tacit assumption that only a finite portion of this reality constitutes the object of scientific investigation".

But by what criteria is this selection - this reduction to the finite made? According to Weber it is not made by nature, as Durkheim might argue; scientific reality is not constructed by the regular unfolding of forces dictated by abstract "laws". In particular, the social-scientific interest has its point of departure ...in the real, i.e., concrete, individually-structured configuration of our cultural life in its universal relationships which are themselves no less individually-structured, and in its development out of other social cultural conditions, which themselves are obviously likewise individually structured.

Hypothetical laws may be helpful as a set of heuristic devices in generating explanations of concrete configurations, but these configurations cannot be deduced from them". Historical configurations interest the investigator, rather, because they are culturally significant for him. This implies further that the investigator has a "value-orientation" toward historical events and situations.

Accordingly, the "presuppositionless" approach to empirical reality was, for Weber, an impossibility. Empirical description, to say nothing of explanation, is impossible without presuppositions and is pervaded by them. "A chaos of 'existential judgments' about countless individual events would be the only result of a serious attempt to analyze reality 'without presuppositions'. To attempt to be empirically exhaustive "is not only practically impossible - it is simply nonsense". It is essto bring order out of chaos by selection of aspects of events, and we select only those parts of reality that are "interesting and significant to us, because only [those parts are] ...related to the cultural values with which we approach reality". Social or cultural reality is not that which presses itself on the uncluttered mind of the investigator; it is "a finite segment of the meaningless infinity of the world process, a segment on which human beings confer meaning and significance."

Concluding this line of reasoning, Weber penned a passage that reads as if it were aimed directly at Durkheim:

What Durkheim wished to banish Weber wished to turn into a set of active guidelines to be used in description and explanation.

Thus Weber envisioned a more intimate connection between the investigator's value-preoccupations and his scientific inquiry than did Durkheim. In another connection Weber took an opposite position; I mention this here only in passing, since it is of little concern to my central purposes. Throughout his methodological writings Weber insisted on the strict separation of statements with empirical (scientific) validity and statements with normative (right and wrong) validity. In particular, he held that the latter cannot be derived from the former; consequently he was opposed to any effort to generate an empirically-based "science of ethics". Scientific investigators can make statements regarding the relative effectiveness and cost of alternative means to given ends; they can also often identify and make explicit goals or values which may not be acknowledged by the actors being studied. But they cannot go further: "empirical science cannot tell anyone what he should do." Durkheim, in contrast, envisioned arriving at assessments regarding the "normal" or "pathological" character of ethical rules on the basis of empirical examinations of the appropriateness of the rule to its social context. Regarding both the implications of cultural values for the conduct of scientific inquiry and the implications of scientific findings for value-judgments, then, Durkheim's and Weber's formulations appear to be firmly opposed to one another.

Durkheim's methodological position was clear and straightforward. He regarded the proper subject matter of sociology as "social facts." These are to be distinguished from both biological (eating, sleeping, for instance) and psychological (reasoning, for instance) facts. They include those aspects of society (for example, a society's religious system, its language, and its system of currency) which have an existence independent of the individual consciousness of society's members and exercise a constraining influence on their behavior. The existence of social facts is (1) to be defined independently of individual consciousness, (2) to be expected to manifest regularities peculiar to themselves and not expressible in psychological terms. and (3) to be expected to impose their influence on the individual's behavior "[a] social fact is to be recognized by the power of external coercion which it exercises or is capable of exercising over individuals, and the presence of this power may be recognized in its turn either by the existence of some specific sanction or by the resistance offered against every individual effort that tends to violate it.

Thus Durkheim was concerned to set the social level apart from the psychological, and to insist on their independence. In the introduction to the second edition of The Rules of Sociological Method, he asserted that social facts differ from psychological facts in quality, in substratum, and in milieu, and he reiterated that the substance of social life cannot be explained by purely psychological factors". The same insistence permeates the pages of Suicide. At one point he acknowledged that conditions affecting individual suicides are "obviously quite distinct" from those affecting the social suicide rate, and "have no social repercussions." He emphatically rejected explanations of the social suicide rates based on abnormal or normal psychological states. He acknowledged that individuals differ in their vulnerability to suicide, but the factors that make them so do not "cause a definite number to kill themselves in each society in a definite period of time". As we shall see, Durkheim's reference to psychological factors in his own comparative sociology was much more complex and subtle than his methodological observations suggest. In principle, however, he was unequivocally hostile both to the reduction of social to psychological phenomena and to the explanation of social by psychological phenomena.

Weber differed significantly from Durkheim in his starting point for sociological analysis. He incorporated a distinctively psychological level into his definition of the basic substance of sociology and social action. Action is defined as such "insofar as the acting individual attaches a subjective meaning to his behavior - be it overt or covert, omission or acquiescence." Action is "social" insofar as "its subjective meaning takes account of the behavior of others and is thereby oriented in its course".

Weber's concern with subjective meaning implies that he regarded the individual as motivated, assessing his environment in terms of its significance for him, and organizing his behavior accordingly; furthermore, social action cannot be understood, described, or analyzed without reference to this subjective meaning. Durkheim may have agreed, but would have insisted that such meaning is relevant for psychology but not for sociology.

Weber qualified his emphasis on subjective meaning in a variety of ways. First, not all behavior is subjectively meaningful; some is merely reactive, though the distinction between the two is subtle. The meaning of action as rendered by the actor must be taken into account in explaining his behavior, but it is not to be regarded as constituting a valid scientific explanation of his behavior. Not all of his motives and perceptions may be conscious; and, in any event, various "non-meaningful" stimuli (for example, environmental or biological events) condition a person's behavior by acting to favor or hinder circumstances. He asserted, furthermore - in reasoning akin to Durkheim's - that the "subjective meaning-complex of action" had to stand on its own level, and could not be decomposed without loss into bio-chemical reactions. And he was careful to avoid confusing the analysis of subjectively meaningful experience with psychology in general, thus giving the former a distinctive conceptual independence."

These divergent starting points propelled Durkheim and Weber in different directions with respect to their approach to empirical data. Durkheim focused upon the observable and the measurable. A social fact such as social solidarity, he noted, "is a completely moral phenomenon which, taken by itself, does not lend itself to exact observation nor indeed to measurement." Therefore, he added, "we must substitute for this internal fact which escapes us an external index which symbolizes it and study the former in the light of the latter." He was drawn to study variobservable kinds of statistics, which record "the currents of daily life" (for example, market statistics); costumes, which record fashions; and works of art, which record taste. Psychology suffered on this count, Durkheim added, because psychological facts are "internal by definition," and therefore inaccessible; "it seems that they can be treated as external only by doing violence to their nature. Weber, because he focused on subjective meaning, was less prepared to treat socio-cultural phenomena as "things." The phenomena of the social sciences involve "a problem of a specifically different type from those which the schemes of the exact natural sciences in general can or seek to solve. " These phenomena are "psychological and intellectual" and call for "empathetic understanding."

Accordingly, Weber devoted a considerable proportion of his methodological statement in Economy and Society to a discussion of different types of understanding, the ways in which meaning can be sensitively and accurately grasped. Weber was also interested in statistical uniformities, but only in so far as they "can be regarded as manifestations of the understandable subjective meaning of a course of social action," as in the case of crime rates or occupational distributions. Furthermore, the two scholars leaned toward a different approach to their data. For Durkheim, the preference would be to regard statistical series as standardized expressions of definite "things" distinct from any meaning that individuals attached to them; for Weber a statistic would be reflective of the subjective-meaning complex of an actor, and would derive its significance from that

complex rather than from any "external" or "superficial" characteristic".

At this moment let me pause and attempt to represent formally the different paradigms for the generation of knowledge emerging from Durkheim's and Weber's methodology. Most of the significant contrasts between the two theorists, as reviewed up to this point, may be understood in terms of how each conceptualized the role of the investigator (observer) and the role of the actor (observed) in the generation of knowledge. Durkheim assigned a passive role to both. In his insistence that facts are "things" he held that they cannot be modified by a "simple act of the [observer's] will"; in his insistence that the observer free himself of all previous preoccupations, he called on him not to attempt to influence empirical facts, but to let them impress themselves upon his mind according to their inherent properties. In these ways the observer is regarded as passive. And because facts are "social," they enjoy an existence independent from the individual, work their influence upon him despite his efforts to resist, and are governed by laws specific to the social level. In these senses, actors as individuals contribute little to sociological knowledge. Such are the key ingredients of sociological positivism as presented by Durkheim. Weber contrasts with Durkheim on both counts. By insisting on the impossibility of a "presuppositionless" sociology, he afforded the observer a more active role in the generation of scientific knowledge. The precise role of the observer, moreover, is guided toward empirical data and problems, which are significant to him from a cultural point of view. And by insisting on the centrality of subjective meaning as the basic ingredient of action, including social action, Weber gave both the actor and the investigator a more active role. In regarding the actor as meaningfully oriented to his environment, Weber insisted that a significant portion of the variables that "explain" human behavior had to be found in the pattern of meanings given to his environment and behaving in accord with these meanings. The actor's own definition of the situation, in short, contributes to explaining his behavior. (Durkheim might have acknowledged the importance of these meaning-complexes, but would have argued that they have no import for sociological knowledge). Also, by insisting on the importance of subjective meaning, Weber saw the task of the observer - that of empathetic understanding - as calling for a more active effort than observing and recording the phenomena that nature produces. Figure 3-1 locates the positions of Durkheim and Weber according to the dimensions of activity and passivity of the observer and the actor, and indicates their general opposition to one another. The differences among them, moreover, are philosophical - or paradigmatic, if you will - in that they deal with first assumptions about the nature and source of empirical knowledge, assumptions that are rooted in articles of conviction or faith and are not easily settled on empirical grounds. Furthermore, the other two "cells" of figure 3-1 indicate two other variants of paradigms for the generation of scientific sociological knowledge. Those who regard the observer as passive in this process, and who insist upon the importance (and the integrity) of the actor as a source of knowledge have assumed many guises, among which phenomenology, relativism, and historicism are the most conspicuous; each regards valid knowledge as emanating from the experience, world-view, or way of life of the people under study. The fourth cell might be termed "sociological nominalism" (for lack of a better term) which acknowledges the independent role of the investigator in formulating categories, constructs, and theories, but does not take into account the meaning-states of the actors being studied. The approach of the self-conscious "model-building" of many contemporary comparative analysts fits into this category.

Because each of these four paradigms rests on different epistemological assumptions, each produces a distinctive kind of knowledge. Theorists of each persuasion create distinctive strengths and vulnerabilities for themselves in generating knowledge. For example positivistic approaches to the analysis of dissimilar systems encourage quantitative comparisons (a strength) but tend to become involved in comparing incomparable data. Relativistic or phenomenological approaches on the other hand are more likely to remain faithful to the perspectives of those who have produced the data but have difficulty in generating categories by which those data can be compared.

Both Durkheim and Weber were committed to the principle that sociology should be a generalizing science, as contrasted, for example, with history. Because of this commitment, they were concerned with generating statements that could deal with many cases (many individuals, many societies). A necessary preliminary to such statements, moreover, is to develop concepts which apply to more than one case. How did Weber and Durkheim come to terms with the necessity of generating general descriptive and classificatory categories, and how did their efforts square with the paradigmatic assumptions each embraced?

One way Durkheim assessed the general significance of social facts was to relate them to a conception of "normal" or "pathological." In characteristic manner, however, he rejected any "premature attempt to grasp the essence" of normality and abnormality, and attempted instead to "seek some external and perceptible characteristic which will enable us merely to distinguish these two orders of facts. In this enterprise he closely followed the reasoning of biology:

Traits as such cannot be generally assumed to be normal or pathological; they must be "defined . . . only in relation to a - given species" and "only in relation to a given phase of its development." What is normal for a simple, preliterate society is certainly not normal for an advanced, complex society. For any given species it is the statistical generality of a social fact that gives it its normality. But in addition, Durkheim wished to explain normality in another way - not only a "normality of fact" but also a "normality of logical necessity." He proposed to do this by referring to what we now might term the "functional significance" of the social fact for the species:

If a phenomenon persists, for example, throughout a long period of social change when, in fact, the conditions for its existence are no longer present, it may be regarded as "pathological."

The conception of the normal and abnormal in statistical terms, however qualified, is open to severe criticisms. Rather than dwell on these, however, I should like to point out how Durkheim's effort to assess the functional significance of social facts led him directly into the comparative analysis of social systems. The significance of a social fact - that is, whether it is normal or pathological - is to be assessed not by some intrinsic feature of the fact but by the societal context of the fact, viz., the requirements of the species at its level of development. Such a formulation calls immediately for a classification of species and of levels of development, since without it the investigator could not make the necessary assessments.

Durkheim was aware of this pressure to classify that arose from his formulation. And in proposing to classify, he tried, much like Weber, to steer a course between "the nominalism of the historian and the extreme realism of the philosopher":

To avoid these two extremes, Durkheim proposed to approach social reality by classifying it into social species. Such a procedure seemed to incorporate "both the unity that all truly scientific research demands and the diversity that is given in the facts, since the species is the same for all the individual units that make it up, and since, on the other hand, the species differ among themselves."

In proceeding, however, Durkheim departed from his general methodological position to a certain degree. On the one hand, he argued that "science can . . . establish classes only after having described, in their entirety, the individuals they comprise. But rather than proceed inductively, he felt it possible to identify certain decisive or crucial facts - the most essential characteristics of social types" - without entering . . . too far into the study of the facts. In particular, he focused on the fact "that societies are composed in various parts in combination" and that they may be arrayed according to their complexity of parts. Accordingly, he defined a simple society as "[one] which does not include others more simple than itself, and which not only at present contains but a single segment but also presents no trace of previous segmentation," and applied this definition to a "horde" When the horde becomes a segment of a society, rather than a society itself, however, a more complex type of society presents itself, a "clan society." And, as societies combine, give birth to new societies, which themselves combine, a typology of societies based on the degree of differentiation of their parts emerges.

At one point Durkheim noted, almost in passing, that "[it] is true, perhaps, that no historical society corresponds exactly to [the] description [of a horde]." Shortly thereafter, he noted that the notion of the horde may be "conceived as a historic reality or as a hypothesis of science". By venturing these observations - and, indeed, by proceeding as he did in proposing his scheme for classification - Durkheim made several crucial departures from his paradigmatic insistence on the passivity of the investigator who could not produce changes in positive facts by an act of will. Not only did he acknowledge the necessity - even desirability - of proceeding to some degree in an a priori way in classifying societies. He also stressed the need to select decisive or crucial facts as the basis of classification. Furthermore, he acknowledged the legitimacy of the investigator's distorting empirical reality by creating a hypothetically pure type not found in empirical reality for purposes of building a classificatory scheme. In making these departures, moreover, Durkheim moved significantly in the direction of Weber's formulation of the role of the investigator, and - as we shall see presently - in the direction of Weber's view of the nature and purposes of classification.

Weber, too, was aware of the tension between nominalism and realism, though it took a somewhat different form than it did for Durkheim. On the one hand, he rejected any position that social reality could be regarded as a manifestation of general social laws; "the reality of which . . . laws apply always remains equally individual, equally undeducible from laws. ". Because he focused on the concrete individual, and, perhaps more important, because he gave the individual's subjective meaning such salience in the definition of social action, he seemed to be flirting with an extreme historical nominalist position. How could individuals, who vary so greatly in their subjective assessment of social reality, be compared with one another? How is it possible to move to the level of social institutions and social structures, which are presumably among the main foci of interest of the sociologist? Weber rejected the idea that the sociologist could "deduce.. institutions from psychological laws or explain them by elementary psychological phenomena". How then to proceed, simultaneously holding both his view of social reality and his commitment to sociology as a generalizing science?

It is in the context of this tension that Weber's famous "ideal type" becomes significant. An ideal type is a device employed by an investigator to facilitate empirical analysis. It is not a description of reality; it is not an hypothesis. Rather, according to Weber's somewhat cumbersome definit, it is "formed by the one-sided accentuation of one or more points of view and by the synthesis of a great many diffuse, discrete, more or less present and occasionally absent concrete individual phenomena, which are arranged according to those one-sidedly emphasized viewpoints into a unified analytical construct." The ideal type is not derived from empirical reality; rather, it is the selection of the essential - indeed, one might say "decisive," as did Durkheim - features of a complex historical situation and molding them into a simplified picture. By drawing out type-elements from the myriad unique historical experiences of concrete acting individuals, the investigator makes them comparable with one another. By constructing an ideal-typical capitalistic system of pricing and marketing, the investigator characterizes in general terms the orientation of numerous actors, who may differ in detail in their concrete subjective orientations to the market. Description in terms of the ideal type selects from those orientations and makes them similar.

It is apparent from Weber's illustrations of ideal-type concepts that they differ in level of generality. In his commentary he mentioned such varying possibilities as "church" and "sect,". "capitalistic culture," "city-economy," "handicraft," "liberalism," "Methodism," "socialism," and "Christianity." In addition to these "abstract concepts of relationships which are conceived by us as stable in the flux of events, he noted that developmental sequences could also be represented in ideal-typical form for example, the "typical" shift from handicraft to capitalistic economic organization, or, more generally, the historical sweeps envisioned in

Marxian theory."

In Economy and Society Weber discussed and illustrated the notion of the ideal type in terms more similar to the way we would currently describe a scientific "model" or "theoretical framework." Constructing a type involves first hypothesizing what course action would take if the actors in a situation were motivated consistently by a single orientation:

Weber regarded the laws of economic theory (including the postulate of maximization) as a representation of "the meaning appropriate to a scientifically formulated pure type (an ideal type) of a common phenomenon. Most sociological laws involve building up hypothetical constructs "on the basis of such rational assumptions".

Weber insisted that the ideal type was a generalizing device, to be applied to a variety of instances of action. In this way he was moving away from the historicism which he felt could not produce general statements. In fact, he regarded the use of general types as inevitable in any kind of historical analysis. If the historian rejects such general theoretical constructs, "the inevitable consequence is either that he consciously or unconsciously uses other similar concepts without formulating them verbally and elaborating them logically or that he remains stuck in the realm of the vaguely 'felt.'" Nevertheless, because Weber laid down no rules for how general

an ideal type should be - ideal types could include anything from general economic models to historical phenomena such as "Methodism" and presumably even sub-varieties of "Methodism" - his own conception ran the risks of manifesting a kind of theoretical indeterminacy, an endless creation of types depending on the historical research at hand, and, indeed, a reversion to historical particularism on a slightly higher level of abstraction."

What are the functions of ideal-type concepts from the standpoint of scientific inquiry? Weber stressed several heuristic uses. They "are of great value for research and of high systematic value for expository purposes when they are used as conceptual instruments for comparison with and the measurement of reality." This would include, as I have mentioned, comparing otherwise different individuals with unique subjective-meaning complexes. They permit the transition from focusing on the individual, concrete actor to the analysis of institutional action - including the influence of one institutional complex on another - by treating it as having a common ideal-type meaning for actors. And they are of explanatory value, particularly those that deal with "laws" or "developmental sequences"; they are themselves compared with some empirical course of events to see to what extent the factors constructed into the ideal type actually account for its regularities. We shall examine Weber's conception of sociological explanation in more detail in the remaining sections.

What are the criteria that guide the selection of those facets of concrete reality that make up any given ideal type? Weber never developed a systematic methodology for this operation - and for this he can be legitimately criticized, since vagueness on this score means that different investigators would undoubtedly come up with different versions of the ideal type for any given historical situation. Nevertheless, Roth has identified a number of "rules of experience" which informed Weber's own construction of types. These rules mainly involved empirical claims that the chosen aspect - for example, legitimation of authority - is a regularly recurring and an important feature of social action.

Whatever the difficulties in Weber's conceptualization of ideal types, it is clear that by employing this device he was attempting an escape from the "nominalism of the historians," as Durkheim called it, and was creating an order of concepts that enabled him to analyze phenomena at a level closer to that which Durkheim called "social facts." In that respect Durkheim's recognition of the necessity to construct a typology of social species and Weber's strategy of generating ideal types moved the two scholars closer to one another than they were in their original paradigmatic statements. Certain differences between the two remained, however. Durkheim consistently adhered to the primary reality of the social level independent of individual actors, whereas Weber insisted that his ideal-type constructions were inferences rooted ultimately in the substratum of the subjectively meaningful experience of individual actors. Furthermore, because of their different starting points, Durkheim and Weber chose to classify different orders of phenomena. Preoccupied with assessing normality and pathology as a product of societal context, Durkheim was driven to classify types of societies. Preoccupied with attaining modest generalizations about typical historical constellations and processes, Weber remained at the level of identifying typical historical clusters of meaningful action. Their mode of classification, then, like all investigative strategies, depended in part on their more general theoretical preoccupations.

Durkheim saw the classification of social species as "a means of grouping facts to facilitate their interpretation". It is "only an introduction to the truly explanatory part of the science". Of what does the latter consist?. Durkheim distinguished between the function fulfilled by a phenomenon that is, its effects, its usefulness and "the efficient cause which produces it." He argued further that the former does not constitute an explanation. Causal analysis, instead, involves the search for "a correspondence between the fact under consideration and the general needs of the social organism, in what this correspondence consists, without occupying ourselves with whether it has been intentional [i.e., directed toward a given end] or not". While stressing this priority, Durkheim acknowledged that knowledge of the function was "necessary for the complete explanation of the phenomena," because "it is generally necessary that [a fact] be useful in order that it may maintain itself".

Durkheim also rejected explanations of social phenomena that called on psychological factors. His position followed from his original definition of a social fact; "[since] their essential characteristic is their power of exerting pressure on individual consciousness, it follows that they are not derived from the latter and, consequently, that sociology is not a corollary of individual psychology." The fact that individuals are the ultimate elements of society is not a compelling reason to seek psychological explanations of social phenomena, any more than it is appropriate to seek inorganic explanations for phenomena constituted at an organic level. The whole is more than the sum of its parts, and explanations appropriate to the whole must be sought: "The group thinks, feels, and acts quite differently from the way in which its members would were they isolated." On the basis of such reasoning, Durkheim concluded that "every time a social phenomenon is directly explained by a psychological phenomenon, we may be sure that the explanation is false. Specifically, the falsity lies in mistaking effect for cause, assuming that the psychological effects produced by collective life are the determinants of that life.

From this follows Durkheim's principle that "[the] determining cause of a social fact should be sought among the social facts preceding it and not among the states of the individual consciousness. The sociologist's main task is to discover features of the social milieu that contribute to the character of social life; Durkheim himself sought to explain the social division of labor by reference to social facts such as the size of society and its dynamic density, and to explain variations in the social suicide rate by reference to the ways in which groups are integrated and regulated. Hostile as his polemic toward psychology was, he nonetheless showed some ambivalence:

Psychology can thus provide "useful suggestions" as to the likely effects of the social milieu on individuals, but in this formulation the individual remains a relatively passive vessel. Psychological phenomena have social consequences only when "the action of the psychological and of the social phenomena is necessarily fused," as in the case of the decision of a public official whose behavior can exert a social influence while being determined by psychological causes. Yet Durkheim regarded such cases as being "due to individual accidents and, consequently, cannot affect the constitutive traits of the social species which, alone, is the object of science." In the end, sociological explanations and sociological knowledge can be generated only if the sociologist "[establishes] himself in the very heart of social facts, in order to observe them directly." In chapter four I shall resume discussion of this formulation, both in connection with observations of how closely Durkheim lived up to his admonitions in his own empirical research and in connection with my own re-formulations on the role of social-structural and psychological variables in comparative analysis.

Weber's conception of sociological explanation is rooted in his notions of interpretation and the ideal type. In his general discussion of how subjective meaning can be understood, he spoke of the importance of "explanatory understanding." In particular this involves grasping the motive of an individual actor, or understanding "what makes him do [something] at precisely this moment and in these circumstances. " The understanding of the act of a man hitting a log with an ax is understood if it is understood that he is working for a wage. "Similarly we understand the motive of a person aiming a gun if we know that he has been commanded to shoot as a member of a firing squad, that he is fighting against an enemy, or that he is doing it for revenge." Motives are highly diverse, and Weber did not conceive of them in a narrow psychological sense. They might include, for example, an individual's self-interest in a given situation, his inclination to adhere to normative standards, or his belief in the legitimacy of a given set of social relationships. In any historical situation the investigator should expect to find not single or pure motives, but a number in complex combination.

Weber identified two types of explanatory understanding. The first involves the interpretative grasp of the "actually intended meaning" - that is to say, meaning for concrete individual action, represented in its historical complexity. (Weber added a sub-type to this, the understanding of the actually intended meaning of numbers of individuals in sociological mass phenomena.) The second, involving a process of abstraction on the part of the observer, is the grasp of "the meaning appropriate to a scientifically formulated pure type (an ideal type) of a common phenomenon." Understanding action from the standpoint of a model of economic rationality would be a case in point; another would be understanding action from the standpoint of being a member of a patrimonial or a bureaucratic staff. In short, it is the identification of a typical complex of motives in a more or less common historical situation.

Even though an ideal-type understanding of motives is an "explanation" of behavior in some sense, Weber insisted that this operation alone, no matter how clear and certain the interpretation, "cannot on this account claim to be the causally valid interpretation." A further operation is necessary; in particular, "verification of subjective interpretation by comparison with the concrete course of events is, as in the case of all hypotheses, indispensable." What is involved in the operation of verification will be taken up in the final section of this chapter. At the moment I stress only Weber's insistence on the distinction between explanatory interpretation and causal verification.

Weber advanced another distinction that made the same point. He referred on the one hand to interpretation of a course of conduct that is "adequate on the level of meaning "that is to say, a satisfactory account of the motives for the conduct from the subjective standpoint of the actor (or that standpoint as assessed by the observer). A causally adequate interpretation, however, involves a statement of the way in which a sequence of events will unfold, and an effort to confirm that statement empirically. But in his concluding statement on causal adequacy, he insisted that a correct causal interpretation required "that the process which is claimed to be typical is shown to be both adequately grasped on the level of meaning and at the same time the interpretation is to some degree causally adequate." If the meaningful - that is, motivational - connection between events is not apprehended, then no matter how close their association, it can remain only "an incomprehensible statistical probability." On the other hand, even if the subjective behavior has been fully grasped on the level of meaning, there can be causal significance "only insofar as there is some kind of proof [verification] that action in fact normally takes the course which been held to be meaningful" .

While Weber did not develop his point about the necessity for a meaningful connection further, it seems to be worth elaborating, because it leads to a contrast with Durkheim and points the way toward a reformulation we shall undertake in chapter seven. What Weber seemed to be saying is that statistical regularities between, say, aggregated rates of behavior are meaningless unless reference is made to some kind of subjective or psychological link between them. For example, the facts that various classes in French society were making irregular forward progress (a statistical regularity) and that numerous members of this class showed evidence of dissatisfaction with the French social order (a statistical regularity) bear no intelligible connection with one another until some typical meaningful connection is made (in Tocqueville's case, a postulate of the principle of relative deprivation). Even further, Weber appeared to suggest that the theoretical significance of regularities is to be found in the realm of subjective meaning:

This statement highlights the difference between Durkheim and Weber on the issue of explanation. Durkheim, embracing a "natural science" model for sociology - at least in his manifesto - envisioned the possibility in sociology of discovering causal uniformities and explaining individual facts by applying them. Sociological theory should emerge at its own level on the basis of observation of social facts. No recourse need be made to that separate realm of psychology except for "useful suggestions". Weber, however, stressing the differences between sociology and what he understood to be the natural sciences of his day, found it essential to construct idealized psychological accounts to give theoretical meaning to social regularities". As we shall see presently, Durkheim turned out to be more Weberian than Durkheimian in his own comparative empirical research.

Both Durkheim and Weber addressed themselves to the empirical procedures available in order to lend empirical support to sociological propositions, and both defined this task in terms of linking causes and effects. Durkheim spoke of "establishing relations of causality," whereas Weber spoke of "causal significance" and "causally adequate interpretation." Durkheim regarded the experiment - when causes and effects "can be artificially produced at the will of the observer" as a potent device for investigation, but found its use limited in sociology, in which "social phenomena evidently escape the control of the experimenter. Weber found experimentation applicable only in "a few very special " Accordingly, each addressed the issue of attaining reliable empirical knowledge in the absence of experimentation.

Durkheim's general answer was simple: when the experiment is not available, the only recourse is indirect comparison, or the comparative method. Before characterizing the particular ways in which he suggested employing it, however, he launched a brief polemic against John Stuart Mill's observation that a given event may have different causes under different circumstances, and enunciated the principle that "a given effect has always a single corresponding cause," adding that, for example, "if suicide depends on more than one cause, it is because, in reality, there are several kinds of suicide."

Durkheim was also skeptical about the applicability of the several strategies that Mill had enunciated in his classic systematic exposition of methods of experimental inquiry. He rejected Mill's "method of residues" establishing cause by removing all known causes with the remainder constituting the cause-- noting that it is inappropriate for sociology because it presupposes the existence of known laws already and is, in any case, impractical. He also found the methods of agreement and difference wanting for similar reasons. (The method of agreement establishes cause by grouping cases which agree in one circumstance and differ in all others; the method of difference establishes cause by grouping cases which differ in one circumstance and agree in all others) . The conditions for such methods can never be absolutely met even in the experimental sciences, Durkheim argued, and in sociology their application is impossible; no conceivable inventory of facts could permit an investigator to be certain, for example, "that two societies agree or differ in all respects save one". Durkheim's chosen method to establish cause and effect was the method of concomitant variation or correlation.

Such reasoning shows the necessity for Durkheim's postulate that a given effect has always a single corresponding cause, which, if correct, permits stronger inference from the correlation than might otherwise be the case. He regarded constant concomitance of cause and effect as "a law in itself, whatever may be the condition of the phenomena excluded from the comparison." It is a very powerful method of proof, even though in any given case covariation might be weakened by the action of other causes that produce "exceptions" to the laws established by covariation. Durkheim felt such exceptions should not lead the investigator to "abandon hastily the results of a methodically conducted demonstration".

But how is the investigator to know, for any given correlation, which is cause and which is effect? Durkheim acknowledged that this is not readily apparent from a correlation - the covariation may result from the fact that both are results of the same cause, or that "there exists between them a third phenomenon, interposed but unperceived, which is the effect of the first and the cause of the second." Durkheim proposed to meet this challenge by a combination of "deduction," or inquiry into "how one of the two terms has produced the other," and new comparisons. Durkheim gave an illustration of a finding to appear subsequently in Suicide:

The example is revealing. In attempting to render intelligible the connection among the three "social facts" - education, suicide, and religious traditionalism - was Durkheim not engaging in precisely the operation that Weber described as the interpretative grasp of ideal-typical meanings? Is it not an assessment of the different meanings of the drive for self-preservation, increasing knowledge, and decline in religious traditions for the typical actor exposed to these phenomena? Was not Durkheim turning to the psychological level - much as Weber did - to seek abstracted statements of the "laws" which dictate the direction of causality in social life?

The answers to both these questions are affirmative, and I shall attempt to confirm the observation further in subsequent chapters.

Toward the end of his discussion of establishing sociological proofs, Durkheim ventured a number of observations on the different types of comparisons necessary to explain data comparatively. Some comparative analysis is possible within a single society, if "facts are widely distributed" and statistical information is "extensive and varied." As an example Durkheim noted the possibility of arriving at "genuine laws" by examining the differences in suicide rates over time according to provinces, classes, age, sex, and the like. This method cannot suffice when studying "an institution, a legal or moral regulation, or an established custom which functions in the same manner over the entire extent of the country and which changes only in time." In such a case the available data would amount to "only a single pair of parallel curves, namely the curve which shows the progression in history of the phenomenon considered and the curve of the supposed cause in this single society". What Durkheim appeared to say is that the latter case does not provide as much variability within the society as the former. Or in other words, the latter case yields only an N of I because of the uniformity of the association within the society, whereas the former produced a larger N of groups and categories showing variation in suicide rates throughout a society.

In the latter case, then, it is necessary to extend the field of comparison and include "several people of the same species." If a parallelism is observed between, say, the social milieu of Rome, Athens, and Sparta and the family systems of these city-states at corresponding stages of their development, one has increased the number of cases in which the concomitant variation occurs and thus increased the confidence in the presumed causal association. Even more, it becomes essential to compare the phenomenon (for example, family systems) in other species as well, so that the more fundamental character of family life can be established, and so that its step by step evolution can be traced. In so doing not only is the number of societies increased but the presumed effects (types of domestic organization) and causes (social milieu) are permitted to vary over a wider range.

In concluding, Durkheim warned of a possible error in such extended comparisons. The error consists in comparing "what occurs at the decline of each species with what happens at the beginning of the succeeding species," regarding them as identical phenomena, and thereby drawing erroneous conclusions. For example, a scholar might observe the religious traditionalism of a society in its late stage of development, and note the religious traditionalism of the society succeeding it, and conclude that "decline in religious traditionalism" is always transitory. This would be an error, Durkheim argued, because the social context of the traditionalism is different. The religious life of a young society is a function of "the special conditions in which every young society is placed." Thus the explanation of apparent "transitoriness" is superseded by another explanation. On the basis of this reasoning, Durkheim admonished that in order to arrive at proper comparisons, "it will suffice to consider the societies compared at the same period of their development"

But in thus admonishing, is not Durkheim paying more homage to Mill's canons than he intended? Is not the insistence that societies be "controlled" for stage of development in order to make just comparisons an invocation of a version of Mill's method of difference, in which cases are made to agree in circumstances (in this case youth, or stage of development) other than the presumed causal connection to be discovered in the concomitant variation of the suspected social facts? The control is gained not by experimental manipulation, as Mill insisted it should be, but rather by the conceptual manipulation of features of similarity and difference among societies in order to rule out stage of development as a causal factor.

Later in his career, Durkheim was even more forceful in enunciating the dependence of "social facts" on social context. Such facts, he argued,

Lamenting the errors that have been committed by "scattering . . . researches over all the societies possible," Durkheim called for a concentration on a clearly determined type. If one includes "all sorts of societies and civilizations," he draws facts hastily from different contexts, and ends up with "tumultuous and summary comparisons." To Durkheim this signaled the need to limit the number of societies studied so that greater precision could be gained. The methodological significance of this point, however, is that if the similarity or differences among facts is a function of the similarity or difference of social context, then social contexts must be made similar if the facts are to be judged so. Or to put it even more directly, the operation Durkheim suggests controls - by means of classification into types - unwanted sources of variation in the phenomena under study.

For Weber, it will be recalled, the principal source of sociological explanation lies in the generation of one or more ideal-type constructions of the subjective-meaning complex of actors and the comparison of these expectations with the best available data. In comparing such "models" with historical data, it is possible "to arrive at a causal explanation of the observed deviations [from the course of action specified in the ideal type] which will be attributed to such factors as misinformation, strategical errors, logical fallacies, personal temperament, or considerations outside the realm of [the posited course of action]". Weber remained skeptical about the level of scientific generality that could be attained in sociology on several counts: first, he was suspicious of highly generalized systems of deductive laws in general; second, he was continuously aware of - and reminding the reader of "historical accidents and the plurality of historical factors [that] make it impossible to predict the actual course of events" and third, he was aware of imperfections in the data with which sociology must deal.

With respect to the latter Weber regarded the empirical verification of hypotheses by experimentation to be very limited in sociology, possible "only in the very few special cases susceptible of psychological experimentation." A second mode of verification is feasible "in the limited number of cases of mass phenomena which can be statistically described and unambiguously i," no doubt because of the large number of cases and the capacity to treat them by techniques of quantitative analysis. Third, "[for] the rest, there remains only the possibility of comparing the largest possible number of historical or contemporary processes which, while otherwise similar, differ in the one decisive point of their relation to the particular motive or fact the role of which is being investigated". In the end, Weber, like Durkheim, laid the heaviest burden in sociology on the comparative analysis of empirical data generated in the historical process.

The second and third types - statistical and comparative analysis - evidently shade into one another. In certain cases, such as the data that might be available to assess the validity of a principle like Gresham's law, "the correspondence between the theoretical interpretation of motivation and its empirical verification is entirely satisfactory and the cases are numerous enough so that verification can be considered established. For other analyses the number of historical cases are so few as to reduce confidence in the results. As an example, Weber cited Eduard Meyer's interpretation of the causal significance of the battles of Marathon, Salamis, and Platea in terms of typical attitudes of the Greek oracles and prophets toward the

Persians. Because analysis had to rest only on the few cases in which the Persians were victorious, such an interpretation, while plausible, must necessarily remain a hypothesis because of the difficulties of verification.

Weber did not develop even as limited a statement of the strategies of comparative analysis as did Durkheim. Insight may be gained into his reasoning, however, by examining what he described as the "imaginary experiment." Listing this "uncertain procedure" after describing the experimental, statistical, and comparative methods, he characterized it as a process of "thinking away certain elements of a chain of motivation and working out the course of action which would then probably ensue, thus arriving at a causal judgment", What sort of methodology underlies this procedure?

In one of his methodological essays, published in 1905, Weber resumed his polemic against those who argued for a "pre-supposition-less" approach to history. Rather, he argued, historical explanation - the attribution of effects to causes - involves a series of abstractions. The decisive abstraction occurs when "we conceive of one or a few of the actual causal components as modified in a certain direction and then ask ourselves whether under the conditions which have been thus changed, the same effect . . . or some other effect 'would be expected". Would the relevant chain of historical events have been otherwise if a given battle had had a different outcome, if a political leader had not been assassinated, and so on? To analyze these possibilities is the essence of the mental experiment. It involves disregarding what actually happened and the "mental construction of a course of events which is altered through modification in one or more 'conditions'.

Weber further described this process in terms of a series of "isolations" and "generalizations." The first process is to decompose the given historical situation into components or factors, and then, by an "empirical rule," determine what effects each of these "conditions" could be expected to have. The "generalization" aspect lies in the "empirical rule," by which Weber meant the store of general knowledge we have about the historical process which permits us to assess the effect of the altered conditions. Finally, and also on the basis of our general historical knowledge, it is possible to assign a judgment of the relative probability of different historical outcomes. "We can . . . estimate the relative 'degree' to which [an] outcome is 'favored' by the general rule by a comparison involving the consideration of how other conditions operating differently 'would' have 'favored' it." Thus, while the general mission of social (including historical) science is "to understand on the one hand the relationship and the cultural significance of individual events in their contemporary manifestations and on the other the causes of their being historically so and not otherwise," much knowledge can be gained by a systematic analysis of the "otherwise" through a series of mental experiments.

From a methodological standpoint, the "imaginary experiment" is a species of comparative analysis. It involves increasing the number of cases under consideration, though in this particular instance the new case or cases are invented rather than observed. Furthermore, by decomposing the historical situation into factors and systematically varying one and then another, Weber was in effect making a conceptual effort to realize the conditions set forth in Mill's method of difference - that is to say, comparing cases that are similar in all respects except one, and attempting to trace the effects of this one difference. The reasons why Weber correctly regarded this procedure as "uncertain" are (1) because the manipulation of the factor is imaginary (not rooted in empirical variation), no actual historical data or variations are produced; and (2) as a result, the imagined "otherwise" data must be posited on the investigator's general knowledge of "laws" and "principles," which are not sufficiently developed in the social sciences to permit the assumption of particular results from certain imaginary situations. Despite the fragile status of the imaginary experiment, however, its logical structure and its strategic significance place it on a continuum with the other methods of generating knowledge in the social sciences.

Durkheim's and Weber's discussion of the logic of verification and proof differ a great deal from one another, because their general programs for sociology differ. Durkheim, approaching social science more from a model of natural science, attempted to modify and adapt the logic and procedures of the natural sciences to sociological inquiry; Weber, approaching social science in a manner which allowed him to escape the pitfalls of historicism, attempted to devise procedures to permit more generalizable inferences than historians typically permitted themselves. Yet the two also approximated one another in significant ways. Both settled on the centrality of comparative sociology - the comparative analysis of similarities and differences in as many empirical instances as could be assembled. Both were sensitive, moreover, to the problems of taking into account controlling, if you will - sources of empirical variation that could "contaminate" suspected causal associations, though neither produced anything like a systematic strategy designed to overcome these problems.

Rather than reiterate in condensed form the comparisons and contrasts between Durkheim and Weber on the methodology of the social sciences, let me instead indicate, in summary form, the issues that emerged in the thought of each as they turned their minds to considering fundamental principles and how they ought to guide scientific inquiry.

1. What are the respective roles of the actor and observer in the generation of sociological knowledge? To what extent does the outlook of the actor have to be taken into account? In what ways does - and should - the observer influence the concepts and data with which he deals?

2. At what level of generality should sociological explanations be pitched statements of laws, statements of probabilistic tendencies, or the explanation of historically unique constellations and events?

3. What kinds of data should the investigator consult in the generation of sociological knowledge? In particular, should he seek for quantifiable indices to be standardized for all cases studied, or should he rely on data that somehow reflect the "uniqueness" of each case being studied? Or should he attempt some compromise between these two strategies?

4. At what conceptual level should knowledge be generated - the psychological, the social, in some way that relates the two, or at some other level?

5. How does the investigator deal with the complexity of his subject matter, particularly when he does not control the recording of that data? What are the major methods that are available to isolate, control, and manipulate variables? What are the relations of these methods to one another?

6. With respect to the comparative method as such - the comparison of limited numbers of cases that differ from one another in some respects what are the available strategies for isolating, controlling, and manipulating variables? What is the relative effectiveness of these methods in terms of generating valid inferences?

7. What is the role of abstract "models" in empirical investigation?

As we have seen, Durkheim and Weber faced all these questions in one way or another, though they occurred to them often in quite different form than I have phrased them, and they differed from one another both in the degree to which they attempted solutions to all and in the kind of solutions they generated. In particular, we observed that both had the habit of proposing a definite, sometimes polemical solution that showed them to be in extreme opposition to one another; then, as they elaborated, qualified, or equivocated, they would gradually move toward a position on each issue that emphasized points of agreement more than opposition.

As we shall see, these issues are far from resolved to the present day, and, in fact, they continue - though often in different form - to dominate the concerns of those currently engaged in comparative analysis. And despite the diversity of solutions generated in the past decades, the same issues arise and re-arise. We shall observe this in our reference to more contemporary developments in the last two chapters. Before undertaking that, however, I should like to look at some of the substantive work of Durkheim and Weber, to see what kind of comparative methodology emerged in their own practice of sociology, as contrasted, perhaps, to how they argued the sociologist should practice.