Multiple Levels of Analysis and the Limitations of Methodological Individualisms


  • Address correspondence to: Ronald Jepperson, Department of Sociology, University of Tulsa, Tulsa, OK 74120. Tel.: 918-631-3900. E-mail: We have benefited from comments provided by Albert Bergesen, John Boli, Marion Fourcade, David Frank, Kyriakos Kontopoulos, Peter Meyer, Francisco Ramirez, Evan Schofer, Ann Swidler, and anonymous reviewers. We also draw upon past conversations with Shmuel Eisenstadt, Edgar Kiser, and Morris Zelditch Jr.


This article discusses relations among the multiple levels of analysis present in macro-sociological explanation—i.e., relations of individual, structural, and institutional processes. It also criticizes the doctrinal insistence upon single-level individualistic explanation found in some prominent contemporary sociological theory. For illustrative material the article returns to intellectual uses of Weber's “Protestant Ethic thesis,” showing how an artificial version has been employed as a kind of proof text for the alleged scientific necessity of individualist explanation. Our alternative exposition renders the discussion of Protestantism and capitalism in an explicitly multilevel way, distinguishing possible individual-level, social-organizational, and institutional linkages. The causal processes involved are distinct ones, with the more structural and institutional forms neither captured nor attainable by individual-level thinking. We argue more generally that “methodological individualisms” confuse issues of explanation with issues about microfoundations. This persistent intellectual conflation may be rooted in the broader folk models of liberal individualism.


Sociology is partly constituted by ideas of “emergent properties” and multiple levels of analysis. These ideas have been discussed throughout the history of the discipline. We refer to the expansive literatures on “social facts,” structural effects, micro- and macro-levels, and structures and agents—epitomized in reviews (e.g., Alexander and Giesen 1987; Blau 1974; Knorr-Cetina 1981; Mayntz 2003; Münch and Smelser 1987; Turner and Boyns 2002) and reflected upon by philosophers (e.g., Bhargava 1992; Bunge 1996; Fodor 1997; Garfinkel 1981; Kincaid 1996; Little 2007).

Yet substantial ambiguity still surrounds the central analytical issue that motivates this article. The issue, in a phrase, is: “Is there such a thing as macro-macro causation?” (Little 2007:362). Spelled out: Are there fully macroscopic explanations—explanations linking one “macro-” property to another macro-property—that are separate from explanations that link one macro-property to another via micro-causal pathways? If so, what is the nature of the difference?

The issue is illustrated recently by the familiar “Boudon-Coleman diagram” (Bunge's [1996] label, reflecting Boudon [1981]). In Figure 1, does the direct macro-macro link—arrow 4—represent a distinctive and legitimate set of causal linkages? Or is that arrow a stand-in to be dispensed with when a macro-to-micro-to-macro pathway—arrows 1, 2, and 3—is provided?

Figure 1.

“Boudon-Coleman diagram.”
(After Coleman[1986a, 1986b] and Hedström and Swedberg [1998]).

Coleman himself sometimes diagrams a direct macro-to-macro connection, but then drops it to focus upon the macro-to-micro-to-macro pathway. The relationship of the direct macro-to-macro arrow to the individualistic pathway is not discussed (see especially Coleman 1987). A related treatment by Hartmut Esser on “explanatory sociology” (Esser 1994) also includes a direct macro-to-macro arrow. Sociology is multilevel, Esser says, but causation occurs only at the bottom. What this means is left unexplained. In contrast, Hedström and Swedberg in an oft-cited treatment do not include any macro-level connection in their Boudon-Coleman variant (Hedström and Swedberg 1998:21ff). System-level properties are linked exclusively via the micro-pathway.

Given the centrality of the issue for sociology as a discipline, the murkiness regarding the direct macro-level linkage is striking. Many discussions acknowledge that a direct macro-macro connection is empirically present in some sense. But rarely does one find a clear sense of what that means, or a clear treatment of the differences—if any—between the different pathways pictured in the Boudon-Coleman diagram. If anything, the methodological individualism offered by Hedström and Swedberg has become a kind of default interpretation.

The main overviews of the “macro-micro connection,” cited above, do not directly address this core issue by working with actual sociological explanations—they typically are philosophical in character. Sawyer (2002:219) in a review can then indict sociologists for failing to provide a clear account of structural causation. Philosophers, in turn, can claim that the discussion remains confused (Zahle 2007:334–36) and that the whole issue requires “fresh thinking” (Little 2007:344). (Treatments that do partly address the specific issue using substantive examples include Collins [1981, 1988]; Hannan [1992], Kontopoulos [1993], Münch [1987], and Stinchcombe [1991].)

We address the central theoretical lacuna in this article. We attempt to provide a clear exposition of multilevel explanation with substantive examples, showing that macro-level causal pathways are distinct from micro-level ones. We seek to provide specification and illustration of structural levels of analysis, of the nature of the causal arguments at these levels, and of the relations of more structural (and institutional) levels to the individual level of analysis. In so doing we also provide a rebuttal to a fundamental element of methodological individualism: the idea that more structural arguments are at best merely “explanation sketches,” or temporary stand-ins for a later proper individualist analysis.

We take up a discursive opportunity provided by Coleman (e.g., 1987, 1990, and elsewhere) and his intellectual colleagues. We refer to Coleman's use of Figure 1 to render Weber's “Protestant Ethic thesis” ([1904/1905] 1996) as a proof text for individualistic explanation. We criticize this usage, and offer an alternative multilevel schema for summarizing the substantive historical discussion of Weber's topic—the connections between Protestantism and capitalist development.

Our effort begins with Eisenstadt's reminder that the Reformation involved broad changes in entire systems of rules and entire sets of roles, as well as changes in motivations (Eisenstadt 1965, 1968b). That is, for this explanatory problem, it is entirely reasonable to hypothesize that changed religious ideas affected individual economic goals, and subsequently changed whole economies. (The famous individual-level argument.) But changed religious rules and roles also altered the social organization of economic life—via a distinct set of causal processes. And, at an institutional level, the changed religious institutions and cosmologies of a whole society altered the entire set of rules governing economic life—another distinct set. That is, causal processes at multiple levels appear to have independent explanatory import.

We argue that methodological individualisms avoid or deny multiple levels of analysis because they confuse explanation with attention to the microfoundations of social life. Of course all causal social processes work through the behaviors and ideas of individual persons—this “ontological truism” (Watkins 1952) is a basic premise of all post-Hegelian naturalist social science. But this premise (sometimes called “ontological individualism”) in no way necessitates an explanatory (or “methodological”) individualism. (See Bhargava [1992], Bunge [1996], Fodor [1997], Kincaid [1996], and Sawyer [2002] for this point.)

Developing this idea, we also build upon the related principle, stressed by Stinchcombe and reinforced by Sawyer, that the levels of analysis featured in a sociological explanation should be an empirical rather than doctrinal matter (Sawyer 2002; Stinchcombe 1991). For some explanatory issues individual-level explanations will be central. But for others they may be ineffective or irrelevant. Here are eight reasons that have appeared in the literature, presented to motivate the discussion that follows. (1) The presence of purposive self-interested actors in a social situation does not indicate that an explanatory model limited to such actors will be sufficient (since studying the “lower-level” composition of any structure does not guarantee explanation) (Bunge 1996:241–81). (2) “Higher-level” explanations (more structural, more macroscopic) are sometimes better than lower-level ones—or the only potential explanations available (Mayhew 1980; Sawyer 2002:23). (3) In many cases a social process, precisely because it represents “organized complexity,” may have individual-level “realizations” that are too heterogeneous or complex to theorize (Fodor 1997; Goldstein 1973; Simon 1962; Stinchcombe 1991). (4) In any case, structural arguments remain possible even when individualist ones are not available (Sawyer 2002:220). In addition, (5) an undue focus on lower-level mechanisms can bring in causally irrelevant material (Sawyer 2003:215). Further, (6) many social explanations do not require much attention to (for example) psychological states (Goldstein 1973). And (7) many structural arguments do not assume or require any particular or special theory of agency or subjectivity in the first instance (Hannan 1992; Sawyer 2002). Finally, (8) in some instances a focus upon lower-level mechanisms may be an outright waste of time, if one is truly interested in explanation (Stinchcombe 1991).


Nonetheless in recent years doctrinal methodological individualisms have become more codified and conventional. A wide range of arguments in sociology focus on the aggregation of the behavior of persons or collectivities conceived as bounded and purposive “actors,” or upon the strategic interaction of putative actors. As Kincaid (1996) notes: “Much contemporary social science makes methodological individualism its official methodology” (1996:7).

The intended scope of the newer individualistic imageries is expansive. For Coleman, for instance, a primary objective is to develop explanations of “the macro-level of system behavior,” of “institutional structure” and of “historical phenomena often large in scale” (Coleman 1990:1–26; Coleman and Fararo 1992:x, xvi). Sometimes the newer individualistic imagery is represented in an ecumenical fashion, as only one of a set of various complementary-and-competing theoretic imageries (e.g., Coleman and Fararo 1992; Fararo 1992). Different imageries generate distinct arguments to be adjudicated (or integrated) via empirical analysis.

In other formulations, however, the representation of individualist theorizing is more exclusivist and doctrinal. In these cases methodological individualism is presented as an all-purpose and self-sufficient matrix for social theory. The posture is “microchauvinist” (Turner and Boyns 2002), present to varying degrees in, for example, the work of Abell (1997), Coleman (1990), Elster (1989), Hechter (1989), and Hedström and Swedberg (1998). Here the claim is that social scientific explanation must reason through causal processes that reach to the presumed bottom-line of the person, conceived as a sharply defined purposive “actor.” As Michael Hannan notes, “many [sociologists] assume without reflection that sociological theories that do not direct explicit attention to natural persons are inherently flawed” (Hannan 1992:127). A philosopher of social science has suggested that many social scientists now treat methodological individualism as a “categorical imperative,”“unconditionally binding for all social scientists, because it is based on certain self-evident truths about society” (Udehn 2002:501).

In stark contrast, recent empirical discussions of macro-social outcomes—such as long-term economic development in the historical West or in the contemporary world—have been patently “multilevel” in character, employing various structural and institutional arguments as well as individual-level ones. In the literature on Anglo-European modernization, for example, explanatory ideas have expanded to consider broad institutional frameworks over long time periods (Baechler et al. [1988] and Engerman [2000] provide overviews). Scholars have been driven to develop various structural and institutional arguments—as well as individual-level ones—out of pragmatic explanatory necessity. Most of the structural hypotheses they have had to employ cannot realistically be converted into any individualist form. A whole range of useful hypotheses could not even be considered if one insisted upon employing individualist imagery in an exclusive way.

Consider the history of the “Weber thesis” itself. In the actual historical discussion of socioeconomic development in the modern world, Weber's thesis was entirely absorbed in broader analyses of institutional conditions for development. There are the Roman legal ideas transmitted through the Church and propelling political and organizational rationalization (Anderson 1974; Berman 1983). There is the long-term Christian demystification of nature (Eisenstadt 1999; Huff 2003), supported by a highly rationalized monasticism (Collins 1986, 1999; Werner 1988). There is the extraordinary intellectual and technological dynamism of the Medieval period, once conceived as dark and static (Landes 1998; Mokyr 1992; White 1978). There are the institutional sources of the states system within Christendom, and the effects of that competitive system on economic development (Mann 1986; Strayer 1973; Tilly 1990; Wallerstein 1974). Then there are all sorts of ecological factors (the physical layout of Europe, the Baltic and Atlantic economies) that fostered accumulation (Jones 1981; Landes 1998; McNeill 1974; Pomeranz 2000) and weakened the controls of traditions (allowing the success of both Protestant heresy and proto-capitalist trading systems).

In these substantive lines of thought, the textbook Protestant ethic thesis is unrecognizable. Current discussions of European development present a broad series of institutional forces operating at a macro-sociological level, supporting some structures (the mobilized and competitive national state) and cultural models (rationalized society, moral individualism), and weakening others (feudal aristocracy, the direction of resources into the Church). In fact, as a substantive argument, the narrow thesis was absorbed quite early, starting with Weber's own work. As Collins (1980) emphasizes, by the time Weber gave his lectures on “General Economic History” in 1919–1920, the Protestant ethic thesis as we know it played a minor role. The legal and institutional changes aided by the Reformation were given great importance. But the social psychological effects of Calvinism were given a much more circumscribed place (see Weber [1923] 1950:365ff).

Nonetheless, the thesis survived and evolved in the last century as a meta-theoretical instrument rather than a substantive theory of development. It provided a compelling story of how prior institutional changes, possibly triggering an intriguing psychological process, might have produced great historical change. Accordingly, one finds a long and continuing intellectual history of breathtaking claims like those of David McClelland, in The Achieving Society, suggesting that “[t]he Protestant Reformation might have led to earlier independence and mastery training, which led to greater n Achievement [need for achievement], which in turn led to the rise of modern capitalism” (1967:47).


Coleman and others, picking up this individualistic narrative, employed the Protestant ethic thesis in an influential tutorial for sociological explanation (Coleman 1986a, 1986b, 1987, 1990). Others have made similar efforts: Boudon (1981), Hartmut Esser (1994), Hernes (1976, 1989), Wippler (1978), Wippler and Lindenberg (1987), and Zetterberg (1965, 1993). In each case, an individualistic explanatory scheme is offered, and Weber's more structural and institutional arguments are set aside.

Figure 2 shows the way in which the Protestant ethic thesis is presented by Coleman. We insert the labels of types of “mechanisms” suggested by Hedström and Swedberg (1998:21–22; cf. Esser 1994). The figure presents a very simple explanatory structure. Causal relations between historical and institutional structures occur primarily or exclusively through micro-processes: i.e., through similar situational effects on a large number of individuals, who then generate new lines of action, together transforming society. Coleman et al. do not use the scheme to grapple with substantive arguments about Protestantism or the rise of the modern economy—there is no reference to the substantive literature. Instead, they are advocating a certain style of explanation, and arguing (to varying degrees) that it should have something close to monopoly status.

Figure 2.

Typical reconstruction of Weber's “Protestant ethic thesis.”
(After Coleman[1986a, 1986b] and Hedström and Swedberg [1998]).

Coleman's initial discussions accompanying his schema indicated that causal effects can occur at multiple levels of analysis (1986a, 1986b, 1987). Coleman mentioned examples of hypothesized collective effects, referring to work by Hannan and Freeman (1986) on organizational change and by Skocpol and Orloff (1986) on welfare states (Coleman 1986b:355ff). Coleman raised some questions about these collective-level arguments but he did not rule them out. Instead, Coleman indicated that his own primary interest was to develop a better “theory of transitions” between the level of behavior of individual actors and the level of social system behavior (Coleman and Fararo 1992:ix; cf. Coleman 1987).

At other points, however, Coleman's treatment slides into a more doctrinal individualism. In his fuller treatment, Coleman (1990) dispenses with any exclusively system-level interconnections (i.e., the macro-to-macro arrow). The examples of possible collective effects are dropped. Here Coleman is disposed to claim that “the central theoretical problem in sociology is the transition from the level of the individual to a macro level” (1986b:347). He also suggests that macro-level propositions are temporary stand-ins for individual-level explanations (1990:20).

With these latter arguments the “macro-micro-macro” sketch became interpretable as an all-purpose form of explanation. Hedström and Swedberg (1998) seem to present the schema in just this way. Wippler and Lindenberg (1987) have made similar arguments. Here is Hedström and Swedberg's treatment:

[P]roper explanations of change and variation at the macro level entails showing how macro states at one point in time influence the behavior of individual actors, and how these actions generate new macro states at a later time. That is, instead of analyzing relationships between phenomena exclusively on the macro level, one should always try to establish how macro-level events or conditions affect the individual (Step 1), how the individual assimilates the impact of these macro-level events (Step 2), and how a number of individuals, through their actions and interactions, generate macro-level outcomes (Step 3). (1998:21–22, emphases added)

[They add a footnote that says:] The logic of Coleman's argument also suggests that any kind of continuous social action can be conceptualized as a long chain of successive macro-micro-macro transformations… the analytical point is precisely to explain this cumulative social action of a large number of macro-micro-macro transitions. (1998:22, emphases added)

This conceptualization reflects the exclusivist form of methodological individualism that we critically highlight.


If one insists upon individualist formulations, whole sets of social processes will not be considered for possible hypotheses. To make this point we reconstruct the substantive discussion of Protestantism and capitalism in an explicitly multilevel way. Beyond the individual level, we note distinct sets of causal processes working through social-organizational arrangements, and then processes occurring at a very macro-sociological, or institutional, level.

Levels of Analysis

“Levels of analysis” in scientific explanation typically refer to sets of causal processes, each representing different degrees of organizational complexity (hence the idiom of “levels of organization” and “levels of complexity”). This is the conceptualization present in the physical sciences (e.g., Anderson 1972; Gell-Mann 1994; Wilson 1998), and in the philosophy of science (e.g., Bunge 1996; Sawyer 2002; Simon 1962; Thomson 2003:89–100; cf. Kontopoulos 1993 for the social sciences). Philosophers and scientists refer to quantum-level processes, molecular-level ones, cellular ones, and so on through ecological processes and beyond—the ontology of the sciences is “hierarchical” (Anderson 1972; Simon 1962). “Higher-level” processes are currently stable “emergent” configurations of lower-level ones; these configurations then themselves can have causal powers (Bunge 1996:241–81; Thomson 2003:89–100).

Herbert Simon referred to a continuation of such levels into the “architecture of complexity” of the social world (Simon 1962): a hierarchy of levels of organization ranging from social-psychological processes and “elementary social behavior” (Homans 1961) through various social-organizational processes, to cultural-organizational (“institutional”) complexes. This social architecture is bounded on the “lower” side by the psychological and biological processes seen as subsocial ones.

In the multilevel exposition here we distinguish between three different sets of causal processes: individual-level ones, social-organizational ones, and institutional ones.1 For example, when scholars think about economic development they consider individual-level processes, such as the possibility that societies may develop more rapidly because their individual persons are socialized as educated and more productive actors. They also consider social-organizational ones, such as Wallerstein's ideas about how particular trade dependencies may affect development opportunities (Wallerstein 1974). Other potentially relevant processes may be more institutional in character, as when expanded education legitimates and thus fosters technical and social rationalization, sometimes eventuating in economic development (Meyer 1977).

Furthermore, there are consequential interactions among processes operating at different levels. For instance, the individual-level effect of education depends on institutional rules giving credentialing advantages to educated persons (Collins 1979; Meyer 1977). The effects of schooling and other socialization efforts are likely to be weak in social contexts that do not support or legitimate the new values. Instruction in knowledge and values is likely to be more effective if the possession of the new knowledge or values is required in valued roles in society. In any case, the causally operative levels must be decided via substantive and empirical adjudication, not dictated by theoretical precommitments.

Individual-Level Explanation

One can categorize the main “individual-level” processes invoked by social scientists as follows:

  • • strictly individual-level, subsocial, processes, i.e., those considered to be primarily biological and psychological in origin and organization;
  • • social-psychological processes (processes in which both intrapsychic and social mechanisms are inextricably engaged, such as in personality formation);
  • • “elementary social behavior” in the sense of Homans (1961) (e.g., interaction in small groups), or Schelling (e.g., in his “tipping” models of residential segregation, invoking so-called simple social processes involving very little social organization (Schelling 1971));
  • • organizationally rudimentary exchange relations, “simple” group dynamics, “collective behavior” (e.g., crowd dynamics), and “simple” forms of strategic interaction (such as those captured by game theory).2

In Weber's Protestant Ethic there were two main individual-level arguments. The “salvation anxiety” argument—the idea that Calvinist predestination produced uncertainty and compelled self-discipline and hard work—is the individual-level proposition that has been especially attractive to meta-theoretical users of Weber. Yet this idea has not fared well in substantive discussion (Parkin 1982:45; Poggi 1983:47). Contemporary empirical social science is critical of the kind of psychology involved in the “salvation anxiety” proposition. It turns out that fear is often a poor motivator of complex behavior (e.g., ceasing to smoke, driving safely, avoiding drugs), and it is difficult to imagine that people can be frightened into long-run capitalist behavior.

In his own work, Weber emphasized a second more general individual-level dynamic. “Psychological impulses originating in religious belief …[give] direction to the individual's everyday way of life and [prompt] him to adhere to it” (Weber in Bendix 1960:63–64). Protestant sects, especially Calvinism (he thought), extended a more “rational mode of life” outside of monastic circles—in effect, bringing religious discipline to the masses (Weber [1923] 1950:365; see Collins 1980; Gorski 2003). Religious strictures encouraged the “rationalization of the conduct of life in general” (Weber [1923] 1950:354). Over time, more individuals became responsive to economic discipline and economic motivation. This second more general formulation ended up as the main individual-level component in Weber's broader explanatory framework (Collins 1980). The “world historical transformation”—the development of economic rationalism—“was not the product of Puritanism; rather, Puritanism was a late development that reinforced tendencies that had distinguished European society for a long time past” (Bendix 1960:71–72).

Thus Weber's individual-level arguments were highly contextualized—in contrast to the impression given by the Boudon-Coleman sketches. More recent empirical research has taken the same direction. Researchers are increasingly aware that each part of the causal sequence of Figure 2 requires special conditions. Without strong institutionalized support for transformed individual outcomes: (1) The socializing effects of Protestantism on individual attitudes are not likely to be strong. (2) Whatever attitudes may be changed are unlikely to be stable over decades. Personality characteristics are known to be unstable over long periods. (3) Any changed attitudes are unlikely to lead to changed behavior, absent strongly supportive conditions. Changed workplace behavior requires strong structural support. (4) Finally, changed individual attitudes or behavior is unlikely to aggregate into shifts in the institutional structure of society. Weber himself stressed repeatedly that strong desires for profit characterize many societies without producing anything like capitalism (e.g., Weber [1920] 1996:17ff).

This necessary contextualization presents a problem for those who would privilege individual-level explanations. For Figure 2's sequence to be credible, it must be embedded in a society that legitimates Protestant beliefs, that continually reinforces them in individuals, that facilitates their implementation in capitalist action, and that validates capitalist action (formerly stigmatized, as in most societies) as producing the collective good. But if all this supportive social structure is in place, one must consider whether it has effects of its own, quite over and above any effects on individual motives. And if all this supportive structure exists for capitalist activity, why would this activity depend on the private fears and hopes of individuals?

We are not criticizing individual-level explanation per se. For example, for the topic at hand, David Landes's recent rendition of an individual-level hypothesis seems plausible: that Protestantism eventually “encouraged the appearance in numbers of a personality type that had been exceptional and adventitious before,” a “new kind of businessman” that was an important ingredient in the expansion of the new manufacturing enterprises (Landes 1998:178, 175). We are only criticizing the exclusive insistence upon individual-level formulations.

Explanation in Terms of Social-Organizational Processes

Sociologists distinguish two “structural” levels of analysis above the individual or elementary social level. One is social-organizational, the other more institutional.3 Social-organizational processes refer to the causal influences attributed to (for example) hierarchic, network, market, and ecological formations. In each case, specific structural features are the properties of interest: for instance, Simmel's idea that triadic structures generate distinctive dynamics wherever found. Contemporary exemplars are Granovetter (1973) on effects of weak versus strong network ties, Burt (1982) on effects of “structural holes,”White (1970, 1992) on dynamics of “vacancy chains,” and ecological ideas emphasizing the effects of properties of competitive niches upon organizational survival (Hannan and Freeman 1989). Sociologists also invoke a number of “emergent” and more collective social-organizational properties that have a cultural character: group values, typified social positions (roles), prestige patterns, emotional dynamics, interaction rituals, organizational cultures.

Weber's main social-organizational idea about Protestantism has to do with the formation of a new collective actor (Poggi 1983). Protestantism facilitated the transformation of a preexisting collective actor—the urban commercial estate—into the legitimated burgher class of entrepreneurs. Protestantism enhanced the class solidarity among Protestant burghers, by providing a religiously based class morale. In addition, rising commercial classes embraced Protestantism for purposes of political and economic legitimation (Bendix 1960:49–82; Parkin 1982:60–62). Networks of the new collective actors could expand, facilitated by common ideology and the trust generated by shared community (Parkin 1982:60–62; also Mann 1986).

In a more structural imagery, the capitalist is produced less by the inner motives of some individuals than by an evolving religious and political structure that defines a new role or collective actor. The new structures give prestige and legitimacy to the role, and set it in effective relation to other structures, including the state, the religious system, property ownership, and the family system. Quite apart from any motivational effects, the Reformation certainly modified roles and organizational structures throughout society. It weakened the legitimacy of many important roles, and dramatically strengthened others. In the religious arena, the authority of state elites was expanded, and the old roles undercut—priests and bishops were subordinated, and the authority of the cathedral and monastery and papacy destroyed. The religious autonomy of the individual, and of the church as an association of individuals, was enhanced. Politically, landowners and aristocrats were weakened, and the rationalized state greatly empowered. Economically the old roles of the pariah bankers and merchants were undermined, along with the whole complex of landowner-serf relations. The authority and legitimacy of free individuals functioning in markets were greatly strengthened.

Structural explanations, in other words, give accounts of how changed distributions of social values and opportunities occur—in this case supported by dramatic religious changes. Much less causation is located in individual personality or actorhood, and structural arguments do not typically rely upon strong (and empirically questionable) assumptions about the competence, coherence, and boundedness of persons. It is one thing to be a capitalist if that implies the social role of pariah or alien or thief—perhaps some very distinctive values and orientations are necessary, requiring much socialization (religious or otherwise). It is quite another thing to be admired as a capitalist carrier of progress, and quite routine motives may suffice to encourage individuals to want to be recognized as successful. One does not need to assume very much about values or personality to understand that young persons in the 1990s might have seen investment banking as an attractive career.

Explanation in Terms of Institutional Processes

The structural imagery in American sociology often features networks, typically with limited cultural meaning and few supra-individual properties. A recent example is provided by Padgett and McLean (2006), who represent social context by “multiple-network architectures” (2006:1468). People are treated as “cross-domain composites of roles” (2006:1469). This kind of reasoning has been very productive. However, it does not capture the actual imagery about roles used in substantive macro-sociological explanation. For instance, the destruction of the “priest,” and of the magical values attached to formerly stigmatized roles of the entrepreneur or the usurer, are imperfectly captured by pure network imagery. These roles embody dramatic cultural meanings—grand social models and ideologies extending across time and space. They involve cosmologies as well: metaphysical claims about the nature of humans, the physical world, and the moral world.

Historically, sociology has pulled together conceptions of phenomena at this level of analysis using the concept “institution” (Berger and Luckmann 1967; Buckley 1967; Eisenstadt 1968a). Institutions are chronically reproducing complexes of routines, rules, roles, and meanings. They have organizational aspects—structures of authority and responsibility, often integrated in something like control systems. They also have cultural aspects—generalized models formulating and justifying rules, built up into systems of thought and analysis (legal models, political models, religious models, knowledge systems, professional discourses). Instances of institutional structures such as the Catholic Church, national state, monetary system, constitutional legal system, limited-liability corporation, scientific academy, and research university exemplify both organizational and cultural dimensions.

Sociologists propose distinctive causal processes at the institutional level. For instance, substantive treatments of the Reformation (by Weber and others) routinely feature institutional-level arguments. We call attention to five such arguments in the literature (cf. Eisenstadt's early reviews [1965, 1968b] and Gorski's recent complementary discussion [2005]).

1. Reconstruction of polity and community: The Reformation obviously weakened the roles and ideas linking society in a liturgical manner, via the Church and a nominally hierarchical social order (Eisenstadt 1965, 1968b; Walzer 1965; Zaret 1989). It led to the collapse of the ideal of imperial unity as well (McNeill 1963). Instead, the eventual Reformation settlements ended up strengthening new ideas of corporate territorial “societies” independent of the Church (McNeil 1963; Parsons 1977). Moreover, the main institutional innovation of Protestantism—the sect—represented a new model of polity and community.4 This model constituted individuals as empowered agents, responsible for bridging the gap between spiritual ideals and mundane realities. People were supposed to throw off what Marx called “the idiocy of rural life” and to enact an entirely new identity, eventually producing the “cult of individualism” (Durkheim [1898] 1969) that we now naturalize.

2. Legitimating and reconstructing economic activity: By investing economizing with moral significance, Protestant ideology contributed to the collective legitimation of practical economic activity, and of profit and wealth. This line of argument was at the core of Weber's Protestant Ethic, but commentators have focused upon the more individualistic formulations. Private gain became a public good in the new cultural models. This change was a striking departure from the ideological stigmatization of commerce typical in agrarian societies and pronounced in Catholic natural law doctrines. Ideas of just price and just wage, for example, had left economic activity in a state of “intellectual, moral, and legal chaos” (Lüthy 1968:106). Protestant reform groups directly contributed to reformulations of ideas of covenant, contract, and natural law. Calvin himself worked on ideas for rationalizing capital and interest accounts (1968:107). Banking was reenvisioned as part of an autonomous economic sphere, governed by its own distinctive economic calculus. Restrictions on entrepreneurs and corporations were eased. Eisenstadt talks about a general “incorporation of Protestant premises into law” in England, Scandinavia, the Netherlands, and (later) the United States, making eventually for more “flexible,”“open,” and “autonomous” institutions (Eisenstadt 1968b:12). Counter-Reformation zones, in contrast, tended to stifle the economic developments of the time (Lüthy 1968; Trevor-Roper 1967).

3. Legitimating the national state: The attack on the Church chartered the national state as the primary legitimate polity, and subsequently proto-national development efforts (Mann 1986:467ff; Parsons 1977; Wuthnow 1989). Calvinism appears to have provided new models of social discipline and executive administration, central in national state-building efforts in some zones of Europe (e.g., in the Netherlands and Prussia) (Gorski 2003).

4. Restructuring collective action: Protestant sects made new voluntarist organizational forms possible. The very idea of sect members running their own congregations allowed for ongoing experiments with new models of organized action (Thompson 1963; Walzer 1965). People no longer required clans or guilds to act collectively. In addition, Protestant cultural models gave concrete legitimacy to a wide range of collective projects for intendedly rational action in a mundane world—and not just economic ones. Protestant ideas “opened” up the world as a thoroughly objectified, profaned, manipulable field for purposeful “action” (Poggi 1983:70), while desacralizing older forms of authority and related traditions (Walzer 1965). Ideas about the potential for and obligation of social progress intensified. (“It was the Reformation that reestablished the ethical rigorism and instrumental activism of the Old Testament”[Münch 1994:183].) The idea that this progress could occur partly through the action of ordinary persons became far more prominent—after all, Protestantism greatly expanded the number of imagined empowered souls in the world. But notions of these kinds were put in place not by role or dispositional evolution, but by central authorities carrying broad theoretical discourses (e.g., the English, French, and American revolutionaries).

5. Legitimating science: The institutionalization of religious pluralism in Europe, after the settlements of the Religious Wars, indirectly contributed to the conventionalization of intellectual pluralism (McNeill 1963; Zaret 1989). In so doing the Reformation promoted “the [ideological] skepticism and refusal of authority that is at the heart of the scientific endeavor” (Landes 1998:179). Recent work has backed Weber's hypothesis that while “scientific progress and Protestantism must not at all be unquestioningly identified,” Protestantism helped to “[place] science in the service of technology and economics” (Weber [1923] 1950:368; Merton 1970; cf. Mann 1986:471; Landes 1998:178ff). Hence Protestantism not only encouraged “high science” (Mann 1986:471), but promoted the gradual rise of science as a central modern institution engaged with every collective purpose (Drori et al. 2003).

In summary, the Reformation precipitated broad institutional changes in dominant models of individual and society. The consequences were not simply economic. And as institutional changes they had long histories in Western development—long preceding and long following the Reformation itself.

Multilevel Analysis

The point of the above section was simply to call attention to lines of likely distinctive causal influence. We epitomize the discussion in Figure 3, which provides a multilevel alternative to Figure 2. We show causal pathways operating at three levels of analysis. If one reifies the Boudon-Coleman diagram (Figure 2)—and the way of thinking associated with it—one is likely to fixate on possible causal processes such as the “Protestant ethic thesis.” One might not seek out causal possibilities such as the social-organizational and institutional ones we have discussed (see Garfinkel 1981:164ff for the general analytical point). Yet discussion of the possible relations of Protestantism and capitalism would remain entirely superficial unless one does consider them.

Figure 3.

Possible pathways linking Protestantism and capitalism at different levels of analysis.


The more doctrinal methodological individualists might persist in arguing that one should seek to convert the pathways of Figure 3 into those of Figure 2. Hedström and Swedberg (1998) seem to make this argument, in the passage quoted earlier. The idea that structural arguments are at best explanation sketches, to be resolved into and replaced by more adequate individual-level explanations, is now relatively commonplace. This idea depends upon three separate notions, two legitimate but one fallacious and invalidating. First, there is the (legitimate) requirement of adducing specific causal processes if one purports to offer a causal explanation. Second, there is the (legitimate) requirement of being able to illustrate micro-instantiations of any process (i.e., examples of the constituent people and activities involved). Third, there is the alleged necessity of formulating arguments exclusively in terms of micro-processes. This allegation reflects a conflation of issues of explanation with issues of microfoundations, as we now explain. (See Kincaid [1996:142–90] for extended development of this point; also Bhargava [1992] and Sawyer [2002].)

The distinction between microfoundations and explanation has great import across the sciences. Consider current theories of psychological depression for illustration. The availability of medications targeting neurotransmitters has given recent discussions a neurophysiological focus. However, analysts obviously consider multiple possible etiologies of depression, loosely differentiated by levels of analysis (and categorized as, e.g., neurology, psychiatry, psychology and family dynamics). The explanatory role of neurotransmitter balance can vary from central to entirely peripheral in these accounts. Perhaps some depressions can be attributed primarily to genetic or ontogenetic neurotransmitter imbalance. But in other proposed etiologies, neurotransmitters may be envisioned only as carrying microfoundations, without having explanatory centrality. For instance, it has been proposed that some depressive syndromes are adaptive responses to trauma among less “resilient” individuals (Kessler 1997). It has been proposed that women may inhabit more depressogenic social networks than men (Kessler and McLeod 1984). It has been proposed that some depressive syndromes are culturally scripted personality styles (an allegation about Central Europe, for instance [e.g., Kurzweil 1989, Townsend 1978]). These etiologies invoke different levels of causation. In each of these putative etiologies, neurotransmitters will of course be present and operative—and hence causally implicated from an ontological standpoint—but they will have greatly varying explanatory relevance.

This example is representative of countless others across the sciences, wherein microfoundations must be distinguished from explanation. In social science, the “microfoundations” of social-organizational and institutional causal pathways are not equivalent to causal arguments at the level of individuals conceived as actors. For example, the micro-composition of a truly collective-institutional process is not equivalent to one of Coleman's macro-micro-macro explanations.

To drive home this point for sociology, one can consider the different “micro-samples” that make up different causal processes. A micro-sample refers to the constituent people and activities isolated in a specific setting and time (we adapt language from Collins [1981, 1988]). The micro-samples of any social process, suggests Collins, can be imagined as “film-strips”—that is, what specific scenes might one film in an attempt to capture some imagined social process? The scenes (people and activities) that one would film in order to capture an institutional process are different from what one would film to capture a social-organizational process or—most relevant here—an individual-level one. Any of the possible processes connecting Protestantism and capitalism at the different levels could be filmed—with sufficient creativity, of course—and in that sense one would capture “micro-samples” in time and space. But the scenes, players, and activities would be quite different for each level of analysis.

The filmstrips (micro-samples) of an hypothesized process working through individual actors might in fact capture some version of Coleman's macro-micro-macro sequence—for example, Landes's (1998) version mentioned earlier. (Different religion, different socialization, different identity, different aggregated behaviors reshaping community.) But the filmstrips of an institutional-level process—its micro-samples—would not capture a Protestant ethic process or any other process at the level of the person seen as an individual actor aggregating with others, or interacting in a strategic game. For instance, if one considers Lüthy's (1968) account of changed banking laws—an institutional story—the micro-samples would be drawn from legal guilds and court ministries, featuring legal and religious professionals enacting highly institutionalized roles rather than functioning as individual actors. One would be sampling how Protestant theory modified the collectively dominant model of society. Priests and lawyers and state elites, operating as carriers of the changed model, worked to change various rules of the social game. They worked to legitimate usury, for instance. Among other consequences, rich people eventually became more likely to use their money to engage in capital investment, rather than to buy status and protection.

Of course collective processes are produced and reproduced by persons’ behaviors (the “ontological truism”). But the “microfoundations” of many collective processes do not involve a mass of similar individuals, operating as choosing actors, affected by a situation, taking new actions, and changing society via some aggregation or assembly. One would not be able to work a truly collective process into such a framework: the micro-realizations of institutional processes are likely to be multiple and heterogeneous—since institutions are chronically reproduced complexes of disparate rules, roles, routines, and meanings.

Put another way: individual-level causal pathways capture effects produced by relatively unorganized people. Here causation is generated by their subsocial or elementary social characteristics. In contrast, more structural imageries capture causation generated by more and more collective and complexly organized activities. In social-organizational imageries, for example, the featured causal influences might be those generated by (say) relatively durably organized networks of social roles, or group cultural and religious commitments. These patterns will have distinct micro-instantiations, having to do with (for instance) the opportunities or communications available to people (or other units). But the causality lies with the effects of the role networks themselves, not with variations in the unit-level properties. Similarly, in more institutional imagery, the featured causal influences are those generated by broad complexes of organization and meaning. The micro-instantiations might have to do with (for instance) complementary sets of role enactments, or with enactments of broad ideological scripts. But the causality emerges from the complex organization of roles, routines, and meanings that we call institutional structure—and not from variations in role enactments or in the individuals involved.

The recent emphasis on “mechanisms” has obscured these necessary distinctions. It is not only that the term is used in many different ways (Gerring [2007] finds nine definitions; Mahoney [2001] finds 24). The problem is more deeply rooted. First, the discourse about mechanisms conflates the idea of causal process in general—a causal pathway or “chain”—with the idea of the lower (or lowest-relevant-level) causal processes within a multilevel system. Second, some analysts conflate microfoundations with causality in general (Hedström and Swedberg [1998] seem to do this, as does Little [2007:358] in his recent philosophical review).

The conflations ensure the elision of the core analytical distinction illustrated by this article. That distinction, to summarize, is the following. Two macro-properties can be connected via a micro-level causal pathway, as the Boudon-Coleman diagram depicts. But two macro-properties may also be connected via distinct macro-level causal pathways—or distinct meso-pathways—as our outline of Weber's topic has shown (and as Figure 3 depicts). In the latter cases, one may go on to seek microfoundations if it is analytically useful to do so—Stinchcombe's (1991) original point. The recent literature on mechanisms misses this fundamental distinction, and hence avoids the theoretical task of attending to the multiple levels of analysis involved in sociological explanation.


We have emphasized the intellectual costs of assigning privileged status to individualist explanation, using interpretations of Weber's Protestant ethic as our focus. Any levels essentialism, micro or macro, will distort and impoverish social science.

From a scientific standpoint, pure “atomism” and pure “holism” are both fictions (Bunge 2000:396). All the sciences must attend both to composition and to emergent systemic properties, and only some of the latter result from aggregation (Bunge 1996:256, 241–81; Thomson 2003:89–100). The physical sciences during the contemporary period have experienced a robust development of multilevel theorizing, displacing doctrinal atomisms—doctrinal holisms (such as vitalism in biology) having been displaced much earlier (Bunge 2000:384; also Gell-Mann 1994; Wilson 1998).

It is worth reflecting, then, why social scientists in the modern period have recurrently returned, often with a good deal of passion and even moralism, to micro-reductionist themes. (In contrast, “macrochauvinisms” have faded, remaining only in weakened forms in some Continental circles.)

Social science has never been far removed from underlying folk and policy models—national models, into the late twentieth century and more global ones now (Münch 1986, 1994, 2001). In the Anglo-American context, methodological individualism has reemerged repeatedly as a kind of revitalization movement—“bringing men back in” and the like—offering for some scholars a great simplification that might finally lead to a proper and pure social science. Many analysts of social science have detected a constitutive individualism in especially American social science (Bunge 1996:241–63; Hawthorn 1987; Levine 1995; Münch 1986, 2001; Sawyer 2003:221ff; Udehn 2002; Varenne 1984; Zahle 2007:319ff). Liberal and especially American cultural models of society notoriously dramatize and valorize purposive individual action. In the contemporary period these models have intensified and rapidly globalized, weakening notions of states and communities as primordial collective actors, and driving the “cult of individualism” to heights Durkheim could never have envisioned (Durkheim [1898] 1969). Now there are strong folk and professional visions of a whole world of equal, free, educated individuals making all the choices that drive a global economy, polity, society, and cultural system. The individual, now, is the one certain ontological element left in social life. It is no wonder that social science reifies these visions.

In this reification, a powerful preference for “rock bottom” accounts is created (Watkins’s phrase [1957]). Images of modal individuals aggregating in plebiscitary or market-like ways are highly conventionalized, and provide a cultural matrix—in effect, a set of defaults—for social scientific explanation. It is natural to formulate explanations that feature plebiscitary and market-like aggregations of individuals, imagined as modal citizens or consumers, and to focus upon the psychological states of such individuals. (Varenne [1984] has made this argument about American “anthropological conversations,” and Münch [2001] has discussed the [different] matrix of German social theory.) Even the most individualistically inclined social scientists are of course aware of institutions and culture. However, it may seem natural to neglect such forces in professional work, or to think that such forces cannot be measured or theorized, or to assume that individual-level analyses will implicitly capture them.

There is a broader epistemological issue here. There is an underlying tension within modernity's ideological system between its romance of human agency (Sawyer 2003:221ff) and its scientific project. The scientific telos leads necessarily toward biological, ecological, structural, and cultural causation and hence to displacement of action theories. But a culture that descriptively and normatively stresses decisions and decision-based “actions” will tend to support isomorphic professional models, whatever violence to reality such models may do (Brunsson 1985; Kahneman et al. 1982). Such models have in fact proliferated wherever dramas centered on empowered actors are central—today the world of interstate relations, of interorganizational relations, and the market world of relations among persons seen as actors.

The most rapidly developing contemporary industries are those that prop up the individual as actor—educators, therapists, consultants, and professional advisors. Similarly expanded industries service the other more derivative actors of the modern system—the rationalized organization, the expanded service-oriented state. So exaggerated intellectual individualisms become possible because of extraordinarily elaborated institutional structures—ones that methodological individualisms would disregard, at substantial cost to the project of social science.


  • 1

    The terms “macro,”“micro,” and (sometimes) “meso” make it easy to conflate complexity with scale—that is, spatial size or extension in time. But scale and complexity are analytically independent dimensions. In social science one might refer to the “collective level” (for example, the polity) of, say, families—a fairly “micro” unit in scalar terms. One might also refer to a “micro-level” (perhaps specific commodity chains) of the world system, a “macro” scalar unit. Here, we focus on levels of causal processes differentiated by complexity—the fundamental denotation in the philosophy of science.

  • 2

    The example of game theory shows that aggregation of behaviors is not always an element in more individual-level and “elementary social” imagery (as Coleman stressed [1987]). For instance, a game theoretic model of “class conflict,” such as in Przeworski and Wallerstein (1982), employs economic classes as units. Their interaction does not primarily affect society via aggregation. Yet the core model is a bargaining game between two stylized unitary actors, hence a relatively “simple” social interaction in the technical sense just defined.

  • 3

    The boundaries between the three “levels” necessarily blur. Social-organizational ideas blur on the one side with individual and elementary social ones. For instance, ideas about social movements blur with ideas of “collective behavior,” the latter having a more individualist cast. Ideas of organizational and professional cultures blur with ideas about cultures having longer and more institutional histories.

  • 4

    We are indebted to Ann Swidler for this point.