- Shopping Bag ( 0 items )
Explore Yourself-in Spirit and Faith
In this updated reissue of his 1984 classic, James Fowler-author of the bestselling Stages of Faith-applies his groundbreaking research on the development of faith to Christianity. Grounded in developmental psychology, theology, and practical ethics, Becoming Adult, Becoming Christian shows how communities of shared beliefs can support and nurture individuals as they shape themselves, and are shaped, in spirit and faith. In his revised first chapter Fowler locates his approach to the study of human and faith development in relation to the contemporary conversation about identity and selfhood in postmodernity. Fowler invites readers to explore what it means to find and claim vocation: a purpose for one's life that is part of the purposes of God. Reclaiming covenant and vocation as ideals for responsible, mature, Christian selfhood, Fowler shows how a dynamic understanding of what vocation involves can both inform and transform lives.
ADULT DEVELOPMENT THEORIES AND THE CRISIS OF VOCATION.
The Crisis in Vocational Ideals.
Images of Good Man/Good Woman in Ferment.
From "Ideological Tepee" to Proteus: Continuity and Change.
Adult Developmental Theorists as Philosophers and Gossips.
DEVELOPMENTALISTS AS PHILOSOPHERS AND GOSSIPS.
Developmentalists as Philosophers and Gossips (E. Erikson).
Toward Generative Adulthood (D. Levinson).
Seasons and Wisdom (C. Gilligan).
TWO PATHS TO RESONSIBLE SELFHOOD.
Faith Development Theory and the Human Vocation.
Faith, A Focus for Research.
Stages of Faith and Human Becoming.
FAITH DEVELOPMENT THEORY AND THE HUMAN VOCATION.
Adulthood, Vocation, and the Christian Story.
The Christian Classic and Its Narrative Structure.
Christian Faith and the Human Vocation.
The Challenge of Vocation.
Vocation vs. Destiny, or Beyond Self-Actualization.
CHRISTIAN COMMUNITY AND ADULTHOOD.
Self and Others: From Destiny to Vocation.
Vocational Existence and Christian Community.
The Christian Story, Passion, and Affections.
Virtues and Vocation.
BECOMGING ADULT, BECOMING CHRISTIAN.
Studies in Vocation.
Vocation and the Stories of Our Lives.
Development and Conversion.
Becoming Adult, Becoming Christian.
The Shorter Westminster Catechism begins with this question: "What is the chief end of man?" Its answer, learned by twenty generations of the theological heirs of John Calvin, states: "The chief end of man is to glorify God and to enjoy him forever."1 In a broad sense, this is a vocational question. It asks, "What are human beings here for? In what consists the virtue or excellence of human beings?" Any culture shaped around this vision of the human calling would have clear notions and practices regarding childrearing, schooling, ethical norms, and what occupations and services are most valuable. A vital key to understanding any society can be found in the range of vocational ideals it recognizes and toward which it nurtures its members. For tightly knit societies like Calvin's Geneva (or the Scottish villages where the Westminster catechism has been taught for hundreds of years), one central vocational ideal may be enough. This is true particularly for those who, like Calvin, partake deeply of the belief in a sovereign yet merciful God. But consider some other cultural vocational ideals from different societies and eras.
In Aristotle's Nichomachean Ethics, we encounter the ideal of the megalopsychos, the "high-minded man," whose virtues, habits of taste, and artistry in friendship and public life make him truly worthy of respect and emulation. The Stoics offered the image of the rational man, who avoids excesses of passion by aligning his reason with the rational structure believed to lie at the heart of the universe. Early Christianity offered the model-which included women as well as men-of the disciple and member of the ekklesia (church) and, eventually, the models of the monastic, the martyr, and the saint.
When the divine right of kings found new life in Europe as a rationale for royal authority, there emerged with it the vocational ideal of the Christian King. (In the seventeenth century this model began to include Christian Queens as well.) Among the Knights Templar of the Crusades, the model of the "soldier-saint" became vital, and it has reappeared regularly, in many guises, in Western nations. Likewise, in Eastern societies we see, in many forms, the vocational ideals of the sage, the monastic, and the bureaucratic statesman. In Japan we see an Eastern version of the soldier-saint, the samurai.
Such vocational ideals not only shaped the lives of those who played those roles. They also gave rise to more or less hierarchical patterns by which the lives of persons who played many roles in society were given shape, coherence, and accountability by the culture's leading images of the purposes and virtues of human life.
In modern societies we notice everywhere the proliferation and secularization of vocational ideals. From the emergence of the traditional professions (religion, medicine, law, education, and government), with their prescribed roles and privileges, we see both a diversification and a flattening of life models. In societies like our own, the moral and religious overtones of vocational ideals have largely been secularized and relativized. An obvious shift has occurred. Previously people were notable and admired as vocational models because of their virtues and usefulness in service to society. Now, admiration is more likely inspired by the appearance or reality of success and wealth, by fascination with power and its exercise, and by the name and face recognition that comes with celebrity or notoriety. With such shifts in the focus of social recognition and rewards, the value systems, roles, and exemplars that receive respect and interest change significantly. Hence in our society we accord powerful (if transient) fascination and rewards to the vocational models of the rock star, the movie celebrity, the sports idol, the eccentric billionaire-adventurer, and, not infrequently, the spiritual guru or guide.
To say that we are experiencing a societal crisis of vocational ideals is to suggest that our normative models of what it means to be a good woman or a good man are in considerable ferment. Let's consider some of the principal reasons for this. Across many parts of the planet, people today can count on living much longer lives than could adults who lived a century ago. In 1900 in the United States the average life expectancy of males was forty-seven, and for women it was only slightly higher-fifty. There were as many single-parent households in this country then, per capita, as there are today. Rather than separation and divorce, however, the most frequent cause was the death of one of the spouses.2
Because of greater life expectancy, coupled with astonishing breakthroughs in dietary science and medicine, many citizens of affluent countries have the sense today that human life is less precarious than it was in earlier times. For reasonably well off folk, a long and healthy life is now expected as an entitlement. This mood leads to a decided shift toward "this-worldly" criteria for describing a well-lived life.
It is difficult to know to what degree the hope of eternal life or the desire for personal immortality has waned as a source of courage and a motive for ethical and religious seriousness. It does seem certain, however, that as the new millennia opens, many persons carry the hope and expectation of living a long and full life. Relative to previous centuries, this expectation has modified, both personally and culturally, people's sense of time and timing in the "seasons" of our lives. It has also shifted the horizons of values and possibilities that frame our reflections on a good or well-lived life.
A second factor that contributes to ferment regarding our cultural images of the good man or good woman derives from mobility. Physical mobility has characterized this nation of immigrants from its beginnings. From the development of steam power for trains and ships to the laying of the interstate highways beginning in the 1950s, this has been a nation of travelers and movers. In addition, a worldwide military presence and the globalization of business and commerce have rotated American men and women onto virtually every continent since World War II. Opportunities for American students and faculties of American colleges and universities to study and teach abroad have proliferated in the last quarter century, and the literature and religious traditions of other societies have become readily available in good, inexpensive translations.
More pervasive in evoking changed outlooks and experience, however, have been developments in electronics and telecommunication. Most homes have multiple televisions, many served by cable or direct satellite connections, making accessible literally hundreds of local and international channels. The four or five channels of the seventies have given way to specialty channels of all sorts, bringing a pluralistic riot of possible stopping places, national and international, to choose from. Such channels as MTV and CNN, with their around-the-clock availability and their mixture of creative commercials and programming, have blurred traditional distinctions between news and entertainment.
In recent years, access to personal computers in the workplace, in homes, and in libraries and schools has opened up multiple ways to enter various spheres of virtual reality. The Internet offers an extraordinary variety of niches in cyberspace where one can create virtual personas, shape virtual games and ventures, and bring to vivid life the most sublime, ambitious, or destructive imaginations of the heart. Making these created spaces accessible to others-from a few individuals to millions of them-provides virtual meeting places for unaccountable interactions that can reshape consciousness and reframe the patterns of ordinariness.3
As we register the impact of new technology on our image of the good woman or man, we need also to recall developments in the last half of the twentieth century that provided new contexts and methods for intentional work with the self. Daniel Yankelovich, in his 1981 book New Rules: Searching for Self-Fulfillment in a World Turned Upside Down,4 documented a cultural shift from an ethic of self-denial, which characterized American society through the end of World War II, to an ethic of self-fulfillment, which became dominant beginning in the fifties. An entire people, having come through the stress of the Depression and having mobilized in an unprecedented and heroic way to defeat a truly worldwide threat in World War II, began to find the space, the time, and the means to get on with the tasks of marrying, establishing families, continuing or completing educations, and pursuing life's satisfactions and happinesses. And in the course of following those pursuits, people began to come to terms with long-postponed tasks and challenges of their inner lives.
When we reflect on the flood of techniques for self-examination and growth that became available to Americans in the sixties and seventies, we recognize that there must have been a deep and pervasive social readiness for new dimensions of intimacy. Sensitivity training, therapy sessions, encounter groups, and a variety of other techniques inspired by humanistic psychology created a new ethos of introspection, instant self-disclosure, and the sharing of oneself. The spoken and unspoken promise underlying intensive self-reflective experiences and marathons was that of transformation: new dimensions of self-discovery, new depths of honesty in relationships with others, new sensitivity and authenticity. There was the promise of new richness for lives capsulated in routine and jaded by conventional values and lifestyles.
During this same period, with psychiatric services increasingly paid for by health insurance plans, access to many forms of personal therapy was opened up for middle-class America. In this era, theological seminaries, in cooperation with the burgeoning field of clinical pastoral education, built pastoral counseling into their curricula as a central focus. This not only contributed to the development of a new specialization in ministry but also opened, through churches and pastoral counseling centers, new contexts for personal growth and change, sanctioned by the church. Through courses in organizational and personal development, industry widely appropriated and taught the new techniques and value systems of what we came to call the human potential movement.
Traditional images of good manhood and womanhood underwent "disestablishment" in the contexts of sensitivity and encounter groups. In the interest of spontaneity and the discovery of one's deeper or "true" self, people were led to loosen their hold on images of the self fashioned after traditional models and roles. The goal was to learn to recognize and claim one's "needs," or as the clichŽ went, "to get my needs met." Experiences designed to alter people's relations to their bodies and to widen the range of their emotional and intuitive responses to self and others sanctioned a frankly open and experimental approach to sexuality and to issues of commitment and lifestyle.
As we enumerate background movements that contributed to contemporary experiences of ferment regarding vocational ideals and cultural images of good womanhood and manhood, we need to include the successive (and simultaneous) waves of liberation movements that transformed older patterns of sexual, racial, gender, political, and religious relations. Taken individually and as a whole, movements for inclusion and justice for minorities, women, and gays and lesbians have fundamentally altered and enriched this society's previous consensual images of beauty, excellence, and the virtues of good womanhood and manhood.
Finally, I must say something, in a preliminary way, about the changed and changing role of religious authority with respect to our definitive images of vocational ideals. The interplay of all the factors suggested in this brief historical discussion create a complex picture. Nowhere is that complexity more evident than in the areas of religion and spirituality. Certainly the starting point and most visible symbol for sketching changes in this area was the Second Vatican Council of the Roman Catholic Church. Protestants and Reform and Conservative Jews have experienced equivalent transformations, but theirs were less visible because they were more institutionally plural. But with Vatican II we beheld the spectacle-truly remarkable-of an international communion of faith solemnly and publicly going through the anguish of fundamentally altering its self-definition and its structures of authority. From a church defined by the hierarchy and its authoritative control of scripture and tradition we saw a move toward a church defined as the "people of God." Within limits that have been continually contested, "the people of God" were given access to scripture and tradition and called to personal responsibility, along with the hierarchy, for shaping faithful lives and institutions in an acknowledged pluralistic world. Confusing to many Catholics, liberating and exhilarating to many others, the normative images of Christian adulthood fostered by the church became more pluralistic. Individual Catholics and local congregations were freed-and burdened-to take responsibility for implementing the teachings and guidelines of significantly reworked doctrinal and social-ethical teachings of the church. This experience, for many, led to a kind of "vertigo of relativity." As many laypersons put it, "If the Church can change the rules about not eating meat on Fridays and about saying the Mass in Latin, what can't be changed?" Many devout Catholic families found that, intergenerationally, they were loyal to two different churches-one pre-Vatican II, the other post-Vatican II. Since his installation, Pope John Paul II has labored consistently and effectively to reinstate more traditional, centralized authority on doctrine and practices in the church. While he has had some success in reestablishing centralized authority and maintaining a male-dominated church, this has not been accomplished without the alienation of millions of thoughtful and faithful Catholics around the world whose faith had developed to the individuative-reflective stage or beyond.5
I have used the Catholic experience to illustrate a broader movement among all churches and synagogues. A significant number of adherents in all of these communities have claimed personal responsibility for their appropriation of the traditions of which they are a part and have taken authority for how they integrate religious teachings into their lives. They have based their faith stance to a significant degree on their own reflective judgments and experiences. The recent emergence of the term spiritual as an adjective to describe one's personal religious or faith orientation in large measure reflects a commitment to the ideal of a postconventional claiming of authority for one's own religious and ethical orientation and practices.
Inevitably, these developments have altered the role of clergy and theologians as authorities. They have opened the way for struggles over what will be the normative images in religious communities of the virtues and character of the good woman and good man.
From "Ideological Tepee" to Proteus: Continuity and Change The interreacting factors discussed in the previous section, plus many others I did not describe, help to explain a fundamental shift that has occurred in this culture's images of adulthood in the past half century. Many persons over fifty can remember childhoods in stable environments-whether urban, rural or small-town-where adulthood appeared to be and was experienced as a fairly static condition. Some years ago my mentor and friend Carlyle Marney offered a graphic image of this older, more static image of adulthood. He called it the "tepee model" of adulthood.6
When you build a tepee, you fix a number of poles in a circle and then bring them together, resting one against another, at the top. You bind the poles together firmly where they meet and then stretch skins or canvas tightly over the frame. The tent poles, in this analogy, represent a set of ideological commitments or assumptions that one assembles in the process of becoming an adult.
By the time most people are in their late teens or early twenties, they had begun settling a number of matters regarding values and lifestyle. They may have put in place a tent pole for their career or occupation. They may have established a pole for marriage and place of residence. They may have embraced a particular ideology or identified themselves with a particular region, integrating the meaning and loyalties related to being a Democrat or a Republican, a Down East Yankee, a Hoosier, or a Texan into their adult identity. They may have erected another pole dealing with sex roles, providing them with clear values and images concerning what it means to be a man or a woman. Other poles in the tepee are derived from their economic and educational status, and so on. Typically there is also a religious pole (or an antireligious one), made visible by membership and identification with (or rejection of) a church or temple. There are also poles for people's racial and ethnic identifications and the accompanying social meanings.
Once a young adult had assembled these ideological tent poles, Marney suggests, he or she binds them together in an often tacit, or unexamined, unity. Then a covering is stretched over the structure, and the young woman or man enters this ideological hut to dwell, never needing to come out again.
In this older, more static image of adulthood, it was expected that the essential structure would be in place by the mid to late twenties. If one reached thirty without having built it, people would say, "He has not found himself" (implying that the self is analogous to a Platonic ideal form, waiting somewhere in the darkness to be found and embraced). Or if in the late thirties or forties the ideological tepee should begin to be buffeted by the changes and struggles of life-the tent poles rattling, the cover flapping-people would say, "She seems to be having a nervous breakdown." Character, in this view of adulthood, meant consistency, predictability, and stability of values, attitudes, commitments, and lifestyle. Often it also meant the expectation of continuity in one's place of residence and relationships.
Of course, the tepee model is a caricature, too extreme in its depiction of an older, more static image of adulthood. But it may prove a helpful image in our effort to grasp and characterize the present ferment surrounding normative images of adulthood. Let me invite you to consider another, more contemporary caricature of adulthood, this time from the opposite end of a continuum of stability and change.
Robert Jay Lifton pioneered the study of psychohistory, which he defines as inquiry into the relationship between large-scale historical change and individual patterns of adaptation. Among other interests, in his early career Lifton studied the survivors of catastrophic disasters. He coined the now familiar term survival guilt to describe a complex of feelings characteristic of those who have the burden of making sense of terrible events in which friends, loved ones, and others were obliterated but they themselves were spared. Lifton, who conducted much of his early research in the Far East, spent considerable time in Japan in the early sixties. There he devoted many hours to interviewing young adults who as children had survived the atomic bombing of Hiroshima and Nagasaki.
As Lifton analyzed the stories of these young adult survivors in Japan, he began to note a pattern of commitment and change in their lives that arrested him. Again and again he saw that in their teens and early adulthood they seemed to have rapidly made, and then changed, a series of fundamental relational and ideological commitments. He was surprised by the intensity of their commitments and then by their frequent withdrawals and turns, often in diametrically opposed directions. As Lifton pursued research with other survivors-victims of Chinese brainwashing or thought control experiments who escaped to Hong Kong-and with American students who had been caught up in the turbulence of the sixties, he began to see that this pattern could be generalized. To link this pattern to a name and an image, Lifton reached into Greek mythology. Proteus was a minor god in the court of Poseidon, the god of the sea. "Proteus," said Lifton, "was able to change his shape with relative ease-from wild boar to lion to dragon to fire to flood. But what he did find difficult, and would not do unless seized and chained, was to commit himself to a single form, a form most his own, and carry out his function of prophecy."7 In the sixties, some social commentators, including Lifton, began to suggest that perhaps Proteus and the protean style of malleability and continual change held together the elements of an adaptive way of responding to an era of unrelenting and accelerating change. According to that view, it is virtuous to be fluid, flexible, and ready to change fundamental convictions and ideological outlooks frequently. Lifton felt that two major historical factors helped account for the emergence of the protean pattern:
The first is the worldwide sense of ... historical (or psychohistorical) dis-location, the break in the sense of connection which men have long felt with the vital and nourishing symbols of their cultural tradition-symbols revolving around family, idea-systems, religions, and the life-cycle in general.... The second large historical tendency is the flooding of imagery produced by the extraordinary flow of post-modern cultural influences over mass-communication networks. These ... cause (contemporary persons) to be overwhelmed by superficial messages and undigested cultural elements, by headlines and by endless partial alternatives in every sphere of life.8
In 1993 Lifton embarked on a longer and more systematic examination of the protean self and of proteanism in postmodern societies. In The Protean Self he suggests that in the thirty years since he wrote about survivors of the atomic holocaust at Hiroshima and Nagasaki and about survivors of the Chinese government's "thought reform" program, the protean self has appeared in many settings around the world, in both developed and developing societies. There are features of protean selfhood that represent resilient, adaptive responses to disequilibrating experiences of trauma or to contexts of radical change and discontinuity. Protean selfhood may be called for in response to experiences of radical confrontation with evil or with "otherness." At points he acknowledges that the protean self is adaptive and is, in fact, "appropriate to the restlessness and flux of our time."9 He makes an important disclosure, however, in response to those who might be inclined to take the protean self as normative and inevitable for the coming era:
I must separate myself, however, from those observers, postmodern or otherwise, who equate multiplicity and fluidity with disappearance of the self, with a complete absence of coherence among its various elements. I would claim the opposite: proteanism involves a quest for authenticity and meaning, a form-seeking assertion of self. The recognition of complexity and ambiguity may well represent a certain maturation in our concept of self. The protean self seeks to be both fluid and grounded, however tenuous that combination. There is nothing automatic about the enterprise, no "greening of the self," but rather a continuous effort without clear termination. Proteanism, then, is a balancing act between responsive shape shifting, on the one hand, and efforts to consolidate and cohere on the other.10
Like the image of the ideological tepee, the protean self, as a pure type, is something of a caricature. But perhaps the two caricatures, taken as extremes on a continuum, will serve to suggest something about the shift, within the lifetime of many of us, from a largely stable, not to say static, image of adulthood to one that, in its extreme form, manifests great fluidity and instability. I believe that the tepee and the protean model depict, in polar tension, two yearnings and realities that pull powerfully at each other in the hearts of each of us and in the culture of which we are a part. These yearnings and realities-the experience of relentless change and the longing for continuity and stability-persist and conflict in the midst of the shifting tectonic plates of values and convictions underlying our societal and cultural life.
One of the principal reasons for the present widespread acceptance and embrace of psychological theories of adult development, I believe, is that they provide us with narrative frameworks for holding together our profound experiences of change and continuity-and the attendant tensions. In this and in other ways that I shall consider with you now, they provide normative and descriptive images of adulthood that uniquely and powerfully speak to the situation of cultural ferment and confusion over vocational ideals in our society today.
Theorists of adult development have come to play the role in our society that storytellers and mythmakers once played in primitive and classical cultures. They have taken on many of the functions that philosophers and theologians performed in the twelfth through the nineteenth centuries. In our time of fractured images of the human vocation and fragmented religious and cultural symbols, a group of philosophical psychologists has helped us gain a holistic grasp of the course of human life. Using the organic root metaphor of development in a variety of ways, their research and theories offer empirically grounded chartings of predictable patterns and turnings in human life cycles. Seventeenth-century Protestant theologians wrote and taught about what they called the ordo salutis-the path or steps to salvation. It may not be going too far to suggest that philosophical theorists of developmental psychology are offering, in formal and mainly secular terms, contemporary versions of an ordo salutis. This parallel makes more sense if you recall that in Latin, salus, the root of salutis (and the English word salvation), means "wholeness" or "completion."
Consider these further contributions of philosophical adult developmental psychologists. They name and map our experiences of personal change, providing reassurance that many of the crises we experience can be understood in developmental terms. They help us reduce our reliance on classifications and diagnostic terminology derived from the study of pathology. They enable us to see that much of our dis-ease can be understood as "sickness unto new health"-a developmental transition-rather than as "sickness unto death." Providing a language for our experiences of change, they also offer normative depictions of the telos or goals of human life. Their theories provide benchmarks or blazes by which we can determine where we are on the human life course. They provide guidance and encouragement regarding the direction and challenges of the next step. In Chapters Two and Three I invite you to examine with me the perspectives of four philosophical theories of adult development. In Chapter Two, the theories of Erik Erikson, Daniel Levinson, and Carol Gilligan will receive our attention. In Chapter Three, I shall try to deal in a similar way with the theory of faith development for which I and my associates have been the principal researchers. Although we shall give attention to the research and empirical foundations of these theories, our principal interest, for the purposes of this book, will focus elsewhere. Our main concern will be to clarify the theories' images of the human being struggling toward psychological and ethical maturity-their images of human movement toward wholeness and completion. We shall treat these developmental theories as narrative structures, as myths of becoming, that elaborate both the quality and the direction of human growth and development.
As contemporary myths and philosophical statements, current theories of developmental psychology deserve both our attention and our critical evaluation. Our evaluation must focus on their claims to descriptive adequacy-that is, their claims to be the results of scientific research. But equally important, and more so for this particular project, we must engage these myths of becoming with criteria of ethical and theological adequacy. In other words, we must treat these theorists as quasi-ethicists and quasi-theologians.
But why do I refer to adult developmental theorists, including myself, as gossips? This reference began as a humorous recognition of the fact that part of what makes these writings attractive to wide audiences is their interesting and extensive use of case materials from research. In our transient contemporary life we miss the leisure and continuity of community and relationship with neighbors that made gossip not only a dubious form of entertainment but also an important source of life insight and wisdom. One day, I decided to look up the derivation of the word gossip in the Oxford English Dictionary. There I found reason to take my reference to adult developmental theorists as gossips more seriously. Gossip derives from the two roots god and sib, the latter meaning "akin," or "related." Put together, these roots meant "one who has contracted spiritual affinity with another by acting as a sponsor at baptism." Hence a gossip was originally a godfather, godmother, or other sponsor of someone about to be baptized. From this rather surprising sixteenth-century understanding of gossip, the term degenerated to today's common usage: "a person of light and trifling character, especially one who delights in idle talk, a rumor-monger, a tattler." One final use of gossip refers to gossip as a medium: "easy, unrestrained talk or writing, especially about persons or social incidents."11
To refer to adult developmental theorists as gossips is to reclaim the original meaning of the word, to refer to one who serves as a sponsor in a critical rite of passage. It is to suggest that the narrative frameworks of developmental theories provide support and clarification for the passages of our lives. It is also to acknowledge that we do learn a great deal about life from looking deeply into the lives of our contemporaries and forebears. The disciplined manner of investigating and reporting their findings that characterizes these theorists maximizes the possibility that the study of lives can yield wisdom. Moreover, the methods of research employed by adult developmental theorists, involving lengthy interviews and relationships of co-inquiry with their subjects, provide unique and humane modes of access to sources of insight for living our lives.0787952281.txt |
Organizations are multilevel systems. This axiom-the foundation of organizational systems theory-is reflected in the earliest examples of organizational theory, including the Hawthorne Studies (Roethlisberger & Dickson, 1939), Homans's theory of groups (1950), Lewin's field theory (1951), sociotechnical systems theory (Emery & Trist, 1960), Likert's theory of organizational effectiveness (1961), Thompson's (1967) theory of organizational rationality, and Katz and Kahn's (1966) social organizational theory, to name but a few. Further, this axiom continues to provide a foundation for virtually all contemporary theories of organizational behavior. Yet, despite the historical tradition and contemporary relevance of organizational systems theory, its influence is merely metaphorical. The system is sliced into organization, group, and individual levels, each level the province of different disciplines, theories, and approaches. The organization may be an integrated system, but organizational science is not.
General systems theory (GST) has been among the more dominant intellectual perspectives of the twentieth century and has been shaped by many contributors (e.g., Ashby, 1952; Boulding, 1956; Miller, 1978; von Bertalanffy, 1972). Systems concepts originate in the "holistic" Aristotelian worldview that the whole is greater than the sum of its parts, in contrast with "normal" science, which tends to be insular and reductionistic. The central goal of GST is to establish principles that generalize across phenomena and disciplines-an ambitious effort that is aimed at nothing less than promoting the unity of science.
Fundamental to the levels perspective is the recognition that micro phenomena are embedded in macro contexts and that macro phenomena often emerge through the interaction and dynamics of lower-level elements. Organizational scholars, however, have tended to emphasize either a micro or a macro perspective. The macro perspective is rooted in its sociological origins. It assumes that there are substantial regularities in social behavior that transcend the apparent differences among social actors. Given a particular set of situational constraints and demographics, people will behave similarly. Therefore, it is possible to focus on aggregate or collective responses and to ignore individual variation. In contrast, the micro perspective is rooted in psychological origins. It assumes that there are variations in individual behavior, and that a focus on aggregates will mask important individual differences that are meaningful in their own right. Its focus is on variations among individual characteristics that affect individual reactions.
Early efforts to conceptualize and study organizations as multilevel systems were based in the interactionist perspective (Lewin, 1951) and focused on the construct of organizational climate.2 Those early efforts played a significant role in developing a "levels" perspective. Interactionists see behavior as a function of both person and situation, with the nature of the combined effect broadly conceived (as, for example, additive, multiplicative, and reciprocal; see Schneider, 1981; Terborg, 1981). Thus behavior is viewed as a combined result of contextual and individual-difference effects. The interactionist perspective has had a pervasive influence on organizational research. It has played a dominant role in shaping research on climate, first posited by Lewin, Lippitt, and White (1939). It continues to exert influence through research on person-organization fit.
Although interest in the development and testing of multilevel theoretical models has increased dramatically in the past decade, there have been relatively few efforts to provide multilevel theoretical frameworks for organizational researchers (e.g., House et al., 1995; Klein et al., 1994; Rousseau, 1985). Multilevel theory building presents a substantial challenge to organizational scholars trained, for the most part, to "think micro" or to "think macro" but not to "think micro and macro"-not, that is, to "think multilevel." Our goal is to explain fundamental issues, synthesize and extend existing frameworks, and identify theoretical principles to guide the development and evaluation of multilevel models.
The principles we derive are intended to be general guidelines applicable to most circumstances; they are not immutable laws. We acknowledge at the onset that the complexity of the issues involved in multilevel theory makes exceptions to the general principles inevitable. In such cases, theory takes precedence-that is the one overarching principle.
This section describes fundamental theoretical processes that provide the underpinnings for developing multilevel theories. We hope to assist readers in emulating and extending the best of current multilevel thinking. Toward this end, we highlight established principles and consider provocative new possibilities for multilevel theory building and research. For ease of presentation, we present central principles of multilevel theory building and research organized around the what, how, where, when, and why (and why not) of multilevel theoretical models.
On what should multilevel theory building and research focus? The possibilities are virtually endless, reflecting the full breadth of organizational processes, behavior, and theory. Nevertheless, a few guidelines regarding the process of choosing a focus for study are possible. First, we urge scholars to begin to fashion their theoretical models by focusing on the endogenous construct(s) of interest: What phenomenon is the theory and research attempting to understand? The endogenous construct, or dependent variable, drives the levels, constructs, and linking processes to be addressed by the theory. Too frequently, researchers begin theory development with the antecedents of interest: "These are interesting constructs; I wonder how well they predict generic outcomes." Such an approach invites the development of a trivial or misspecified theory. Without careful explication of the phenomenon of interest, it is exceedingly difficult to specify a meaningful network of potential antecedents. Principle: Theory building should begin with the designation and definition of the theoretical phenomenon and the endogeneous construct(s) of interest.
level relationships among human resource practices, aggregate employee outcomes, and firm financial performance, but what are the cross-level and emergent processes-the linkages of individual responses to human resource practices-that mediate the relationship between organizational human resource practices and organizational performance? The time is now ripe for such multilevel theory building (Ostroff & Bowen, Chapter Five, this volume). Principle: Multilevel theoretical models are relevant to the vast majority of organizational phenomena. Multilevel models may, however, be unnecessary if the central phenomena of interest (a) are uninfluenced by higher-level organizational units, (b) do not reflect the actions or cognitions of lower-level organizational units, and/or (c) have been little explored in the organizational literature. Caveat: Proceed with caution!
By definition, multilevel models are designed to bridge micro and macro perspectives, specifying relationships between phenomena at higher and at lower levels of analysis (for example, individuals and groups, groups and organizations, and so on). Accordingly, a multilevel theoretical model must specify how phenomena at different levels are linked. Links between phenomena at different levels may be top-down or bottom-up. Many theories will include both top-down and bottom-up processes.
Top-down processes: contextual influences. Each level of an organizational system is embedded or included in a higher-level context. Thus individuals are embedded within groups, groups within organizations, organizations within industries, industrial sectors within environmental niches, and so on. Top-down processes describe the influence of higher-level contextual factors on lower levels of the system. Fundamentally, higher-level units may influence lower-level units in two ways: (1) higher-level units may have a direct effect on lower-level units, and/or (2) higher-level units may shape or moderate relationships and processes in lower-level units.
Principle: Virtually all organizational phenomena are embedded in a higher-level context, which often has either direct or moderating effects on lower-level processes and outcomes. Relevant contextual features and effects from the higher level should be incorporated into theoretical models. Bottom-up processes: emergence. Many phenomena in organizations have their theoretical foundation in the cognition, affect, behavior, and characteristics of individuals, which-through social interaction, exchange, and amplification-have emergent properties that manifest at higher levels. In other words, many collective constructs represent the aggregate influence of individuals. For example, the construct of organizational culture-a particularly broad and inclusive construct-summarizes the collective characteristics, behaviors, and values of an organization's members. Organizational cultures differ insofar as the characteristics, behaviors, and values of organizational members differ.3
tem. Thus, for example, interactions among atoms create molecular structure, or interactions among team members yield team effectiveness. This perspective views an emergent phenomenon as unique and holistic; it cannot be reduced to its lower-level elements (e.g., Dansereau et al., 1984).
Principle: Many higher-level phenomena emerge from characteristics, cognition, behavior, affect, and interactions among individuals. Conceptualization of emergent phenomena at higher levels should specify, theoretically, the nature and form of these bottom-up emergent processes.
Virtually inseparable from the question of how is the question of where-that is, precisely where do top-down and bottom-up processes originate and culminate? The answers to these questions specify the focal entities-the specific organizational levels, units, or elements-relevant to theory construction. Suppose, for example, that a theorist is interested in the influence of unit climate on individual actions. What is the level of interest? For example, is it group climate? division climate? organizational climate? the climate of the informal friendship network? In the passages that follow, we will first explore the nature of organizational units as evoked by multilevel theory and then describe processes that determine the strength of the ties that link organizational levels or units.
Nature of organizational units. All but the smallest organizations are characterized by differentiation (horizontal divisions) and integration (vertical levels). These factors yield myriad entities, units, or levels. In organizational research, levels of theoretical interest focus on humans and social collectivities. Thus individuals, dyads, groups, subunits, and organizations are relevant levels (units, or entities) of conceptual interest. The structure is hierarchically nested so that higher-level units encompass those at lower levels. Many writers (Brown & Kozlowski, 1997; Freeman, 1980; Glick, 1985; Hannan, 1991; Simon, 1973) assert the importance of using formally designated units and levels for specification; for example, leadership research typically defines the "leader" as the formal unit manager. Generally speaking, formal units can be defined with little difficulty, although there can be exceptions, where unit boundaries or memberships are fuzzy.
Principle: Unit specification (formal versus informal) should be driven by the theory of the phenomena in question. Specification of informal entities that cut across formal boundaries, or that occur within formal units and lead to differentiation, requires careful consideration.
Determinants of the strength of ties linking organizational levels or units. One overgeneralization of the systems metaphor is that everything is related to everything. In reality, some levels and units are much more likely than others to be strongly linked, through what Simon (1973) refers to as bond strength. The theorist needs to chose appropriate units and levels or risk a misspecified or ineffective theory. Bond strength and related concepts help to explain what is likely to be connected across levels, and why.
Principle: Linkages across levels are more likely to be exhibited for proximal, included, embedded, and/or directly coupled levels and entities.
Principle: Linkages are more likely to be exhibited for constructs that tap content domains underlying meaningful interactions across levels.
Time is rarely a consideration in either single-level or multilevel organizational models (House et al., 1995), yet it is clearly the case that many if not most organizational phenomena are influenced and shaped by time. Here we explore three ways in which time may be incorporated into a multilevel model, increasing the rigor, creativity, and effectiveness of multilevel theory building.
Time as a boundary condition or moderator. Many organizational phenomena have a unidirectional effect on higher- or lower-level organizational phenomena, but multilevel relationships are not always so simple; instead, over time the relationship between phenomena at different levels may prove bidirectional or reciprocal. A given phenomenon may appear to originate at a higher or lower level according to the theorist's assumption about the current time point in a stream or cycle of events. The failure, quite common, to make such assumptions explicit can lead to apparently contradictory models of the same phenomenon and to debates about its "true" level.
Principle: The temporal scope, as well as the point in the life cycle of a social entity, affect the apparent origin and direction of many phenomena in such a way that they may appear variously top-down, bottom-up, or both. Theory must explicitly specify its temporal reference points.
Time-scale variations across levels. Differences in time scales affect the nature of links among levels (Simon, 1973). Lower-level phenomena tend to have more rapid dynamics than higher-level and emergent phenomena, which makes it is easier to detect change in lower-level entities. This is one reason why top-down models predominate in the literature. For example, efforts to improve organizational outcomes (for example, quality) through training (for example, total-quality management, or TQM) assume emergent effects that originate at the individual level. Models of training effectiveness focus on the transfer of trained skills to the performance setting. Higher-level contextual support (for example, a transfer climate; see Rouiller & Goldstein, 1993) enhances transfer in such a way that the effects of TQM training on quality are relatively immediate. However, the effect of individual-level TQM training on organizational outcomes is emergent and requires a much longer time scale. Individual cognition, attitudes, and behaviors must combine through social and work interactions. Depending on the nature of the vertical transfer process, individual outcomes will compose or compile to the group level and, over longer time frames, will yield organizational outcomes (Kozlowski & Salas, 1997; Kozlowski, Brown, Weissbein, & Cannon-Bowers, Chapter Four, this volume). Thus contextual or top-down linkages can be manifest within short time frames, whereas emergent, bottom-up linkages necessitate longer time frames.
Principle: Time-scale differences allow top-down effects on lower levels to manifest quickly. Bottom-up emergent effects manifest over longer periods. Research designs must be sensitive to the temporal requirements of theory. homogeneous groups, heterogeneous groups, or independent individuals-can be influenced by factors that, over time, change the level of the relationship (Dansereau, Yammarino, & Kohles, 1999).
Entrainment: changing linkages over time. The term entrainment refers to the rhythm, pacing, and synchronicity of processes that link different levels (Ancona & Chong, 1997; House et al., 1995). Coupling across levels or units is tightened during periods of greater entrainment. Entrainment is affected by task cycles and work flows, budget cycles, and other temporally structured events that pace organizational life (Ancona & Chong, 1997). For example, the concept of entrainment has been used in the group and team performance literature to capture the idea that work-flow interdependence is not necessarily uniform over time; rather, the degree of interdependence or coupling can vary significantly depending on the timing of events or acts that require a synchronous and coordinated response (e.g., Fleishman & Zaccaro, 1992; Kozlowski, Gully, McHugh, Salas, & Cannon-Bowers, 1996; Kozlowski et al., 1999; McGrath, 1990). Thus levels or units that ordinarily are loosely coupled will be tightly coupled during periods of synchronicity.
Principle: Entrainment can tightly couple phenomena that ordinarily are only loosely coupled across levels. Theories that address entrained phenomena must specify appropriate time cycles and must employ those cycles to structure research designs.
Argument by assertion is invariably a poor strategy for theory building. Argument by logical analysis and persuasion-argument that explains why-is always preferable. In multilevel theory building, explaining why is not merely preferable but essential. A great deal of organizational multilevel theory building spans organizational subdisciplines (industrial/organizational psychology and organizational theory, for example). Therefore, the unstated assumptions in a multilevel theory may be obvious to the members of one subdiscipline but not to the members of another, who are also interested in the new multilevel theory. Furthermore, multilevel theories often incorporate novel constructs (for example, team mental models, or organizational learning). The meaning of such constructs may well be obscured in the absence of thorough explanations concerning why. Finally, multilevel data analysis has been the subject of considerable and continuous debate. Conflicts regarding the best way to analyze multilevel models abate considerably, however, in the presence of carefully and fully explicated theoretical models (Klein et al., 1994) that make the choice of analytical strategy clear (Klein, Bliese et al., Chapter Twelve, this volume). Thus multilevel theorists must not only specify what, how, where, and when but also why: Why are relationships in the model conceptualized as top-down rather than bottom-up? Why are constructs conceptualized as compositional rather than compilational? Why are predictors assumed to have immediate rather than long-term consequences for the outcomes of interest?
in an organizational subunit? Why might predictors, hypothesized to be influential over time, prove instead to have immediate consequences? In exploring why not, theorists may refine their models, incorporating important insights and nuances. This adds diversity and depth to theory; it is how a science is built.
Principle: Multilevel theoretical models must provide a detailed explanation of the assumptions undergirding the model. Such explanations should answer not only the question of why but also the question of why not.
Many of the controversies and problems associated with multilevel research are based on misspecifications or misalignments among the theoretical level of constructs, their measurement, and their representation for analysis. Misalignment is a problem for any research design that incorporates mixed levels, but it is also a problem for single-level research that incorporates emergent constructs. The nature of these misalignments is well documented elsewhere (Burstein, 1980; Firebaugh, 1979; Freeman, 1980; Hannan, 1991; Robinson, 1950; Rousseau, 1985; Thorndike, 1939). The following are some common problems: blind aggregation of individual-level measures to represent unit-level constructs, use of unit-level measures to infer lower-level relations (the well-known problems of aggregation bias and ecological fallacies), and use of informants who lack unique knowledge or experience to assess unit-level constructs.
Misalignments degrade construct validity and create concerns about generalizability. To build theoretical models that are clear and persuasive, scholars must explicate the nature of their constructs with real care. Precise explication lays the foundation for sound measurement. Constructs that are conceptualized and measured at different levels may be combined in a variety of distinctive multilevel models. Research design and analytical strategies need to be aligned with the levels inherent in these models. Principles relevant to these concerns are considered in the remainder of this section.
Construct level and origin. Constructs are the building blocks of organizational theory. A construct is an abstraction used to explain an apparent phenomenon. The level of a construct is the level at which it is hypothesized to be manifest in a given theoretical model-the known or predicted level of the phenomenon in question. Although organizational theorists have often discussed "the level of theory," we prefer to use the phrase level of the construct because mixed-level models, by definition, include constructs that span multiple levels; that is, generalizations are constrained by the level of the endogenous construct ("the level of the theory"), but other constructs in a model may be at higher or lower levels. Thus, in mixed-level research, the theoretical explanation will span several levels in the effort to understand an endogenous construct at a given focal level.
Principle: The theorist should explicitly specify the level of each construct in a theoretical system.
Principle: When higher-level constructs are based on emergent processes, the level of origin, the level of the construct, and the nature of the emergent process must be explicitly specified by the theory.
Types of unit-level constructs. Unit-level constructs describe entities composed of two or more individuals: dyads, groups, functions, divisions, organizations, and so on. In the organizational literature, many problems and controversies revolve around the definition, conceptualization, justification, and measurement of unit-level constructs. The "level" of many higher-level constructs (culture, leadership, or participation, for example) is often debated. The debate is due in part to the potential for these constructs to emerge from lower-level phenomena.
Global unit properties differ from shared and configural unit properties in their level of origin. Global unit properties originate and are manifest at the unit level. Global unit properties are single-level phenomena. In contrast, shared and configural unit properties originate at lower levels but are manifest as higher-level phenomena. Shared and configural unit properties emerge from the characteristics, behaviors, or cognitions of unit members-and their interactions-to characterize the unit as a whole. Shared and configural unit properties represent phenomena that span two or more levels. Shared unit properties are essentially similar across levels (that is, isomorphic), representing composition forms of emergence. In contrast, configural unit properties are functionally equivalent but different (that is, discontinuous), representing compilation forms of emergence. Configural unit properties capture the variability or pattern of individual characteristics, constructs, or responses across the members of a unit. We elaborate in what follows, and then we discuss how the nature of a unit construct influences its measurement.4
Global unit properties. Global constructs pertain to the relatively objective, descriptive, easily observable characteristics of a unit that originate at the unit level. Global unit properties do not originate in individuals' perceptions, experiences, attitudes, demographics, behaviors, or interactions but are a property of the unit as a whole. They are often dictated by the unit's structure or function. Group size and unit function (marketing, purchasing, human resources) are examples of global properties. There is no possibility of within-unit variation because lower-level properties are irrelevant; indeed, any within-unit variation is most likely the result of a procedure that uses lower-level units to measure the global property. If, for example, group members disagree about the size of their group, someone has simply miscounted. Unit size has an objective standing apart from members' characteristics or social-psychological processes. In contrast, "perceived group membership" is an entirely different type of construct.
Shared unit properties. Constructs of this type describe the characteristics that are common to-that is, shared by-the members of a unit. Organizational climate, collective efficacy, and group norms are examples of shared unit-level properties. Shared unit properties are presumed or hypothesized to originate in individual unit members' experiences, attitudes, perceptions, values, cognitions, or behaviors and to converge among group members as a function of attraction, selection, attrition, socialization, social interaction, leadership, and other psychological processes. In this way, shared unit properties emerge as a consensual, collective aspect of the unit as a whole. Shared unit properties are based on composition models of emergence, in which the central assumption is one of isomorphism between manifestations of constructs at different levels; the constructs share the same content, meaning, and construct validity across levels. When researchers describe and study shared unit properties, they need to explain in considerable detail the theoretical processes predicted to yield restricted within-unit variance with respect to the constructs of interest: How does within-unit consensus (agreement) or consistency (reliability) emerge from the individual-level characteristics (experiences, perceptions, attitudes, and so on) and interaction processes among unit members?
Configural unit properties. Constructs of this type capture the array, pattern, or configuration of individuals' characteristics within a unit. Configural unit properties, like the shared properties of a unit, originate at the individual level. Unlike shared unit properties, however, configural unit properties are not assumed to coalesce and converge among the members of a unit. The individual contributions to configural unit properties are distinctly different. Therefore, configural unit properties have to capture the array of these differential contributions to the whole. Configural unit properties characterize patterns, distribution, and/or variability among members' contributions to the unit-level phenomenon. Configural unit properties do not rest on assumptions of isomorphism and coalescing processes of composition but rather on assumptions of discontinuity and complex nonlinear processes of compilation. The resulting constructs are qualitatively different yet functionally equivalent across levels.
Principle: Theorists whose models contain unit-level constructs should indicate explicitly whether their constructs are global unit properties, shared unit properties, or configural unit properties. The type of unit-level construct should drive its form of measurement and representation for analyses.
Basic issues. The level of measurement is the level at which data are collected to assess a given construct. Individual-level constructs should, of course, be assessed with individual-level data. Unit-level constructs, in contrast, may be assessed with either unit-level or individual-level data. When unit-level constructs are assessed with unit-level measures, an expert source (a subject matter expert, for example, or an objective archive) provides a single rating of each unit. When unit-level constructs are assessed with individual-level measures, unit members provide individual-level data (for example, individual ratings of climate, or individuals' reports of their own demographic characteristics), which are subsequently combined in some way to depict the unit as a whole. Rousseau (1985, p. 31) advises researchers to measure unit-level constructs with global (that is, unit-level) data whenever possible: "Use of global data is to be preferred because they are more clearly linked to the level of measurement, avoiding the ambiguity inherent in aggregated data." Klein and colleagues (1994, p. 210) note that when a researcher uses "a global measure to characterize a group, he or she lacks the data needed to test whether members are, indeed, homogeneous within groups on the variables of interest." Accordingly, Klein and colleagues (1994, p. 210) recommend that researchers use global measures to capture unit-level constructs only when the level of the construct is "certain" or "beyond question." Here, we elaborate on Rousseau's (1985) and Klein and colleagues' (1995) admonitions, advising that the level of measurement should be determined by the type of the unit-level construct.
Individual-level constructs. Individual-level constructs should, as already noted, be assessed at the individual level. For example, individuals may complete measures of their own job satisfaction, turnover intentions, self-efficacy, psychological climate, and so forth. In some cases, one or more experts may provide assessments of the characteristics of other individuals. This procedure can be used when the characteristic is observable, or when the informant has unique access to relevant information (Campbell, 1955; Seidler, 1974). A supervisor may describe his or her individual subordinates' performance behavior, an observer may record individual demographic characteristics, or a researcher may use archival records to assess individuals' ages, tenure, or experience. In each case, data are assigned to individuals and are considered individual-level data. Issues of measurement quality are, of course, still relevant.
Global properties. The measurement of unit-level variables is often more complex and more controversial. Least complex and l east controversial is the measurement of the global properties of a unit. By definition, global properties are observable, descriptive characteristics of a unit. Global properties do not emerge from individual-level experiences, attitudes, values, or characteristics. Accordingly, there is no need to ask all the individuals within a unit to describe its global properties. A single expert individual may serve as an informant when the characteristic is observable, or when the informant has unique access to relevant information. Thus a vice president for sales may report his or her company's sales volume, a CEO may report a firm's strategy, or a manager may report a unit's function. Although these examples each use an individual respondent, the data are considered global unit-level properties.
Shared properties. In contrast, shared properties of a unit emerge from individual members' shared perceptions, affect, and responses. The theoretical origin of shared properties is the psychological level, and so data to assess these constructs should match the level of origin. This provides an opportunity to evaluate the composition model of emergence underlying the shared property; that is, the predicted shared property may not in fact be shared, in which case the data cannot be averaged to provide a meaningful representation of the higher-level construct. Therefore, the data to measure shared unit properties should be assessed at the individual level, and sharedness within the unit should be evaluated. Given evidence of restricted within-unit variance, the aggregate (mean) value of the measure should be assigned to the unit. Several empirical examples of this approach to the conceptualization, assessment, and composition of unit-level constructs can be found in the literature (e.g., Campion, Medsker, & Higgs, 1993; Hofmann & Stetzer, 1996; Kozlowski & Hults, 1987). This approach ensures both that the data are congruent with the construct's origin and that they conform to the construct's predicted form of emergence, thereby avoiding misalignment.
Configural properties. When a construct refers to a configural property of a unit, the data to assess the construct derive from the characteristics, cognitions, or behaviors of individual members. Individual-level data are summarized to describe the pattern or configuration of these individual contributions. As before, theory-the conceptual definition of the emergent construct-drives the operationalization of the measure. Configural properties emerge from individuals but do not coalesce as shared properties do. Thus a researcher, in operationalizing the configural properties of a unit, need not evaluate consensus, similarity, or agreement among individual members except to rule out coalescence. The summary value or values used to represent the configural property are based on the theoretical definition of the construct and on the nature of its emergence as a unit-level property. A variety of data-combination techniques may be used to represent, capture, or summarize configural properties, including the minimum or maximum, indices of variation, profile similarity, multidimensional scaling, neural nets, network analyses, systems dynamics and other nonlinear models, among others. The mean of individual members' characteristics is generally not an appropriate summary statistic to depict a configural unit property, although it may be combined with an indicator of variance or dispersion (Brown et al., 1996). In the absence of within-unit consensus, means are equifinal, ambiguous, and questionable representations of higher-level constructs.
Principle: There is no single best way to measure unit-level constructs. The type of a unit-level construct, in addition to its underlying theoretical model, determine how the construct should be assessed and operationalized. As a general rule, global properties should be assessed and represented at the unit level. Shared and configural properties should be assessed at the level of origin, with the form of emergence reflected in the model of data aggregation, combination, and representation.
Establishing the construct validity of shared properties. The assumption of isomorphism that is central to the conceptualization of shared constructs requires explicit consideration. There are two primary issues relevant to testing models with one or more shared unit properties:
The issue of the measurement model addresses the construct validity of aggregated lower-level measures as representations of higher-level constructs. It is generally addressed through examining patterns of within-group variance. Consensus- or agreement-based approaches-for example, rwg(j)-evaluate within-group variance against a hypothetical expected-variance (EV) term. Agreement is examined for each shared property measure for each unit: a construct-by-group approach. Consistency- or reliability-based approaches-for example, ICC(1), ICC(2), and within-and-between analysis (WABA)-evaluate between-group variance relative to total (between and within) variance, essentially examining interrater reliability for each shared property across the sample: a construct-by-sample approach (Kozlowski & Hattrup, 1992; Bliese, Chapter Eight, this volume).
Principle: The assumption of isomorphism of shared unit properties should be explicitly evaluated to establish the construct validity of the aggregated measure. The selection of a consensus- or consistency-based approach should be dictated by theory and data; no approach is universally preferable.
Data source, construct, and measurement levels. Individuals as sources of data play different roles in measuring the three different types of unit constructs. This observation highlights the distinction between the data source, on the one hand, and the level of the construct and its measurement, on the other. For example, a knowledgeable individual may act as the data source for a global unit property such as size, function, or strategy, but in such a case the level of measurement is not considered the individual but rather the unit as a global entity.
Principle: Individuals may serve as expert informants for higher-level constructs when they can directly observe or have unique knowledge of the properties in question. As a general rule, expert informants are most appropriate for the measurement of global unit-level properties and observable (manifest) configural properties. They are least appropriate for the measurement of shared properties and unobservable (latent) configural properties.
Item construction. Several authors have provided guidelines for item construction, primarily for the measurement of shared properties. In general, the advice is to focus respondents on description as opposed to evaluation of their feelings (James & Jones, 1974) and to construct items that reference the higher level, not the level of measurement (James, 1982; Klein et al., 1994; Rousseau, 1985). In practice, research has tended to use items framed at both the individual level (data source) and at higher levels. Recently, Chan (1998) distinguished these practices as representing different composition models of the constructs in question. For example, Chan views climate items referencing self-perceptions (for example, "I think my organization ...") as constructs distinct from items that tap the same content but reference collective perceptions (for example, "We think the organization ...")-what he refers to as "reference shift consensus."
Theoretical models describe relationships among constructs. A multilevel perspective invites-indeed, necessitates-special attention to the level of the constructs united within a theoretical model. In this section, we build on the preceding section by describing broad types of models distinguished by the levels of the constructs they encompass, as well as by the links they propose among constructs. Model specifications are illustrated in Figure 1.1. Following our description of basic models, we note further complexities in the creation of multilevel models.
Single-level models. Single-level models, as their name suggests, specify the relationship between constructs at a single level of theory and analysis. Such models are common in our literature and generally represent particular disciplinary perspectives. Psychologists are likely to find individual-level models the most familiar and straightforward type of single-level model. Individual-level models may be conceptually complex, specifying intricate interactional relationships among numerous constructs. However, individual-level models, by definition, ignore the organizational context of individual perceptions, attitudes, and behaviors. Thus the simplicity of individual-level models is in many cases a major limitation. Indeed, ignoring the context when it is relevant will lead to biases in the examination of construct relations (that is, the standard-error estimates of parameters will be biased).
Cross-level models. Cross-level theoretical models describe the relationship between different independent and dependent constructs at different levels of analysis (Rousseau, 1985). Typically, organizational cross-level models describe the top-down impact of higher-level constructs on lower-level constructs (outcomes and processes). Although theory often conceptualizes the potential impacts of lower-level constructs on higher levels (the impact of newcomers on group cohesion, for example), bottom-up cross-level modeling is a distinct rarity in the empirical literature because of its analytic limitations. We should note, however, that recent work is beginning to address this problem (Griffin, 1997). Here, we outline three primary types of top-down cross-level models:
predictor. Thus, for example, an organization's adoption and implementation of a new computerized technology may engender changes in the image of the organization to outsiders, in the extent to which distinct groups within the organization coordinate their work tasks, and in individual employees' feelings of job security as a function of their technical expertise and trust in the organization. Mixed-determinant and mixed-effect models may be combined to create complex cross-level models of antecedent and outcome networks.
The "group average" specified in a frog-pond model is not conceptualized as a shared property of the unit. Indeed, were the construct predicted to be shared within each group, then it would make no conceptual or empirical sense to assess individual standing on the construct relative to the mean-the hallmark of frog-pond models (Xi - the group mean of X). Nor is the "group average" considered a global property of the unit; perhaps the group average, in combination with deviations, may be considered a configural property of the unit. This insight is subtle and complex, but it may help clarify why the frog-pond effect has been classified by some scholars as a distinct phenomenon or even as a distinct level of analysis. Just as we have created a distinct category for configural unit-level properties- unit properties that are characteristics of the unit but are neither global nor shared (isomorphic)-so others (e.g., Klein et al., 1994; Dansereau & Yammarino, Chapter Ten, this volume), in their conceptualizations, have designated frog-pond (heterogeneous or parts) models as a distinctive level.
Homologous multilevel models. These models specify that constructs and the relationships linking them are generalizable across organizational entities. For example, a relationship between two or more variables is hypothesized to hold at the individual, group, and organizational levels. Such models are relative rarities. The most commonly cited example of such a model is Staw, Sandelands, and Dutton's (1981) model of threat rigidity. Staw and his colleagues posit that the way in which individuals, groups, and organizations respond to threat is by rigidly persisting in the current response. By arguing for parallel constructs and homologous linking processes, they have developed a homologous multilevel model of threat-rigidity effects. However, the model has not been tested empirically, its propositions are open to debate (e.g., House et al., 1995), and its attention to construct composition is limited. Lindsley, Brass, and Thomas's model (1995) of efficacy-performance spirals is an excellent example of a homologous multilevel model that carefully attends to the composition of its constructs. However, we know of no empirical test, in the published organizational literature, of a fully homologous multilevel model.
Sampling within and across units. When testing individual-level theoretical models, researchers endeavor to ensure that their samples contain sufficient between-individual variability to avoid problems of range restriction. Sampling issues in multilevel research are more complex but comparable. In testing unit-level theoretical models (for example, the relationship between organizational climate and organizational performance) and mixed-level models containing unit- and individual-level variables (for example, the relationship of organizational human resources practices and individual organizational commitment), researchers must endeavor to ensure that their samples show adequate variability on the constructs of interest, at all relevant levels in the model. Thus, for example, it may be inappropriate to test a cross-level model linking a group construct to an individual outcome in a single-organization sample. If a higher-level organizational characteristic constrains between-group variability, it will yield range restriction on the measure of the group construct and preclude a fair test of the model. Unfortunately, this problem is all too common in levels research. Principle: In the evaluation of unit-level or mixed unit-level and individual-level theoretical models, the sampling strategy must allow for between-unit variability at all relevant levels in the model. Appropriate sampling design is essential to an adequate test of such models.
Sampling across time. In the section on theoretical principles (see "Principles for Multilevel Organizational Theory Building," pp. 21- 25), we highlighted the importance of time, as well as its general neglect in theory construction for processes that link different levels. However, temporal considerations are important not only for theory; they are also essential to research design. Two issues are central: differential time scales across levels, and entrainment.
Principle: Time-scale differences allow top-down cross-level effects to be meaningfully examined with cross-sectional and short-term longitudinal designs. Bottom-up emergent effects necessitate long-term longitudinal or time-series designs.
Principle: Entrainment tightly links phenomena that are ordinarily only loosely connected across levels. Sampling designs for the evaluation of theories that propose entrained phenomena must be guided by theoretically specified time cycles, to capture entrainment and its absence.
Several techniques are available for the analysis of multilevel data: analysis of covariance (ANCOVA) and contextual analysis using ordinary least squares (OLS) regression (e.g., Mossholder & Bedeian, 1983); cross-level and multilevel OLS regression; WABA (Dansereau et al., 1984); multilevel random-coefficient models (MRCM), such as hierarchical linear modeling (HLM; Bryk & Raudenbush, 1992); and multilevel covariance structure analysis (MCSA; Muthen, 1994). The techniques differ in their underlying theoretical assumptions and are designed to answer somewhat different research questions. Therefore, no single technique is invariably superior in all circumstances; rather, the choice of an analysis strategy is dependent on the nature of the researcher's questions and hypotheses. Here we see again the primacy of theory in dictating the resolution of levels issues. The best way to collect and the best way to test multilevel data will depend on the guiding theory. The more explicit and thorough the guiding theory, the more effective data collection and analysis are likely to be. We provide a brief overview of these analytic approaches here but direct the reader to later chapters in this volume for in-depth consideration of contextual and regression analysis (James & Williams, Chapter Nine), WABA (Dansereau & Yammarino, Chapter Ten), and multilevel random-coefficient models (Hofmann, Griffin, & Gavin, Chapter Eleven).
ANCOVA and contextual analysis. Among the earliest approaches to the analysis of cross-level data were adaptations of ANCOVA and the use of OLS regression to conduct contextual analysis (Firebaugh, 1979; Mossholder & Bedeian, 1983). The ANCOVA approach is used to determine whether there is any effect on an individual-level dependent variable that is attributable to the unit, beyond the effect accounted for by individual differences. Essentially, this approach treats the individual-level variables as covariates and then uses unit membership as an independent variable to determine how much variance is attributable to the unit. Unit membership as a variable accounts for all possible remaining differences across units. Therefore, this approach cannot identify the specific constructs relevant to unit membership that are actually responsible for observed differences among groups; such effects are unexplained. Nevertheless, to the extent that there are any differences attributable to the grouping characteristic, this approach will capture it (Firebaugh, 1979).
Cross-level and multilevel regression. In the organizational literature, OLS regression has been adapted to examine cross-level and multilevel effects and is quite flexible with respect to the type of model it can evaluate. Contemporary uses of this approach treat aggregation as an issue of construct validity (James, 1982; Kozlowski & Hattrup, 1992) so that a model of emergence is first evaluated before individual-level data are aggregated to the group level (e.g., Kozlowski & Hults, 1987; Ostroff, 1993). Therefore, with respect to the specification and measurement of construct types, this approach is relevant to the issues we have discussed in this chapter. Once the measurement model of the higher-level (aggregated) constructs is established, the analysis proceeds to test substantive hypotheses. For example, if the theory assumes shared perceptions of unit climate as predictors of individual satisfaction, then one establishes restricted within-unit variance on climate, aggregates the data to the unit level (that is, computes means), and then disaggregates to the individual level of analysis (that is, assigns the means to individuals in the unit). The analysis then estimates the amount of variance in individual satisfaction that is attributable to unit climate. Individual-level analogues of the contextual construct are not necessarily controlled (as in contextual analysis) unless the question is of substantive interest (James & Williams, Chapter Nine, this volume).
Within-and-between analysis. The basic WABA equation (Dansereau et al., 1984) is modeled on the classic decomposition of within-and-between variance terms formulated by Robinson (1950) to model individual-level and aggregate group-level correlations. The "classic" WABA analysis examines bivariate relationships, assumes measures at the lowest level of analysis for all constructs, and proceeds in two phases. The first phase, WABA I, establishes the level of the variables. The second phase, WABA II, evaluates the level of relations between all the variables in the analysis (Dansereau et al., 1984). WABA I is designed to assess whether measures, treated one at a time, show variability in the following ways: both within and across units (as typically with an individual-level construct), primarily between units (as typically with a unit-level construct), and primarily within units (as with a frog-pond, parts, or heterogeneous construct). WABA II is designed to assess whether two measures covary in the following ways: both within and across units (as typically with individual-level relationships), primarily between units (as typically with unit-level relationships), and primarily within units (as typically with a frog-pond, parts, or heterogeneous relationship; see Klein et al., 1994). Although WABA was originally developed to examine bivariate relations at multiple levels, it has been extended to address multivariate relations (Schriesheim, 1995; Dansereau & Yammarino, Chapter Ten, this volume).
Multilevel random-coefficient modeling. The MRCM analysis strategy is represented by several packages of statistical software (for example, PROC MIXED in SAS; MLn; lme in S-PLUS), of which HLM is probably the most familiar. HLM analysis assumes hierarchically organized, or nested, data structures of the sort that are typically encountered in organizations: individuals nested in units, units nested in organizations, and organizations nested in environments. Models of theoretical interest typically represent multiple levels of data. For instance, many cross-level models involve an outcome variable at the lowest level of analysis, with multiple predictors at the same and higher levels. HLM is well suited to the handling of such data structures.
Principle: There is no one, all-encompassing multilevel data-analytic strategy that is appropriate to all research questions. Particular techniques are based on different statistical and data-structure assumptions, are better suited to particular types of research questions, and have different strengths and weaknesses. Selection of an analytic strategy should be based on (a) consistency between the type of constructs, the sampling and data, and the research question; and (b) the assumptions, strengths, and limitations of the analytic technique.
Some of the most engaging and perplexing natural phenomena are those in which highly structured collective behavior emerges over time from the interaction of simple subsystems [Crutchfield, 1994, p. 516].
A central theme woven throughout this chapter is the need for a more extended understanding of emergence as a critical multilevel process in organizational behavior. There is evident dissatisfaction with the overreliance on isomorphism-based composition as the primary model for conceptualizing collective constructs (House et al., 1995; Rousseau, 1985). Indeed, there is increasing recognition that emergence based on isomorphism may well be the exception rather than the rule. Although isomorphic emergence is a very powerful conceptual model, it is but one possible model. Emergent phenomena are not necessarily shared, uniform, and convergent. In their discussion of dispersion theory, a precursor to our typology, Brown and Kozlowski (1997, p. 7) note that nonuniform "phenomena marked by differentiation, conflict, competition, coalition formation, and disagreement are common" in organizations.
Our purpose is to take a step toward this elaboration, describing forms of emergence that until now have received little attention in the organizational literature on levels of analysis. In preceding sections of this chapter, we contrasted composition (shared unit properties) and compilation (configural unit properties) as distinctive, ideal types of emergence. This contrast was useful in making salient the important differences that affect conceptualization, measurement, and sampling. However, composition and compilation are not necessarily clear-cut dichotomous categories; rather, they are end points for a diverse set of emergence alternatives, with some forms of emergence being more akin to composition and some forms being more akin to compilation.
There are three primary conceptual contributions of this effort. First, our intent is to be inclusive, encompassing multiple perspectives. Several recent theoretical efforts have started to explore emergence and the ways in which it may be manifest (Brown & Kozlowski, 1997, 1999; Brown et al., 1996; Chan, 1998; Kozlowski, 1998, 1999; Morgeson & Hofmann, 1999a, 1999b). Although these efforts are for the most part compatible, they have also chosen different points of theoretical departure, different language, and different organizing structures. It is not our goal to explicitly integrate these efforts, but we believe our framework makes their compatibilities more explicit. We build on the strong theoretical and research foundation provided by isomorphism-based composition and elaborate it to embrace different, alternative, and neglected forms of emergent organizational phenomena that follow from a consideration of discontinuity-based compilation. Because compilation entails less restrictive assumptions, it allows for many more possible emergent forms relative to composition. We argue that a broader range of alternatives, from composition to compilation, is necessary to more fully capture complex emergence.
Emergence is bottom-up and interactive. The concepts undergirding emergence have broad expression in the biological, social, and physical sciences and are represented in theories of chaos, self-organization, and complexity (Arthur, 1994; Gell-Mann, 1994; Kauffman, 1994; Nicolis & Prigogine, 1989; Prigogine & Stengers, 1984) which address the dynamics of emergence. Our focus is on emergent phenomena that occur within the boundaries and constraints of organizational systems. Emergence is particularly relevant in the continuing effort of our science to understand how individuals contribute to organizational effectiveness. This is a central theme in several of the chapters of this book, including those focused on selection (Schneider, Smith, & Sipe, Chapter Two), performance appraisal (DeNisi, Chapter Three), training effectiveness (Kozlowski et al., Chapter Four), and human resources management (Ostroff & Bowen, Chapter Five). Emergence plays an important role in the linkages involved in interorganizational relationships (Klein, Palmer, & Conn, Chapter Six) and cross-cultural relations (Chao, Chapter Seven).
Emergence is shaped and constrained. Although emergent phenomena have their origins in lower levels, the process of emergence is shaped, constrained, and influenced by higher-level contextual factors. Interaction in organizations is constrained by a hierarchical structure that defines unit boundaries. The individuals in a unit tend to interact more dynamically and intensely with each other than with individuals outside their unit (Simon, 1973). Moreover, work-flow transactions-the ways in which people are linked to accomplish the work of the unit (Thompson, 1967)-pattern interactions and exchanges. Individuals directly linked by the work flow tend to interact more with each other than with individuals who are only linked indirectly (Brass, 1995). Thus, for example, professors tend to interact more intensely with the students who are involved in their research than with the other students in their programs, and they interact more with students in their programs than with students in other programs. This patterning of interaction by formal structure and work flow shapes emergence.
Emergence varies in process and form. As already noted, interaction dynamics can lead to variation in the ways in which a higher-level phenomenon emerges; that is, a given phenomenon, such as team performance, can arise in a variety of different ways, even in the same organization. Individual characteristics, cognition, affect, and behavior are constrained by their context. Over time, interaction dynamics acquire certain stable properties; stable structure emerges from a dynamic process. Katz and Kahn (1966) describe this as recurrent patterns of interaction. Thus the emergence of a collective phenomenon is the result of a dynamic unfolding of role exchanges (Katz & Kahn, 1966), ongoings (Allport, 1954), or compilation processes (Kozlowski et al., 1999) among individuals. It is from these dynamics that a stable collective pattern emerges.
Our framework is formulated around theoretical distinctions between ideal forms of composition and compilation, considered in earlier sections of this chapter. Here we turn our attention to three sets of overlapping assumptions, shown in Figure 1.2, that are useful for more finely distinguishing these alternative forms of emergence. The assumptions include the following elements:
At the risk of some redundancy, we will outline these assumptions and apply them to the contrasting of composition and compilation forms of emergence. We will then present a typology, using the assumptions to distinguish alternative forms of emergence ranging between composition and compilation ideals.
Model and elemental contribution. Composition and compilation are distinguished by their underlying theoretical models. Composition is based on a model of isomorphism, whereas compilation is based on a model of discontinuity.7 Isomorphism and discontinuity represent differing conceptualizations with respect to the nature and combination of the constituent elements that constitute the higher-level phenomenon.
Interaction process and dynamics. The hallmark of composition forms of emergence is convergence and sharing. In climate theory, for example, a variety of constraining forces have been proposed that are thought to shape the emergence of a shared collective climate. Individuals are exposed to homogeneous contextual constraints-common organizational features, events, and processes (James & Jones, 1974). They develop individual interpretations of these characteristics, yielding psychological climate. ASA processes operate to narrow variation in psychological climate (Schneider & Reichers, 1983). Interpretations are filtered and shaped by leaders (Kozlowski & Doherty, 1989). Individuals interact, communicate perspectives, and iteratively construct a common interpretation. Variations in individual interpretations dissipate as a collective interpretation converges. This is an incremental process that, over time, promotes stability, characterized by reduced dispersion as outliers are trimmed and by increased uniformity as perceptions are pushed to a convergent point. An equilibrium is achieved.
et al., 1999).
Combination rules and representation. The representation of an emergent construct is an effort to capture or freeze the result of a dynamic process. The assumptions identified earlier provide the basis for different combination rules-guidelines for summarizing or capturing a collective representation from the elemental content. For composition, similar types and amounts of elemental content that evidences relative stability, uniformity, and low dispersion will generally be summarized with linear additive or averaging rules. This procedure will yield a single indicator-a convergent point capturing the shared unit property. Collective climate, based on composition assumptions, is generally represented by unit means (Kozlowski & Hattrup, 1992). Homogeneous perceptions of worker participation are likewise represented as unit means (Klein et al., 1994).
Summary of distinctions between composition and compilation. The key assumptions that distinguish composition and compilation, respectively, involve the question of whether the following elements are present:
The purpose of our typology is to promote a more expansive conceptualization of the theoretical mechanisms that characterize different forms of emergence. Our typology of emergence, shown in Figure 1.3, juxtaposes composition and compilation. The theoretical underpinnings derived previously are used to distinguish a variety of exemplars-specific emergence models. We discuss each exemplar, illustrating the exemplars with examples regarding collective performance, learning-cognition-knowledge, and other phenomena. We include exemplars for the following types of emergence: convergent, pooled constrained, pooled unconstrained, minimum/maximum, variant, and patterned. Each exemplar decribes a different emergence process, based on contextual constraints and interaction processes, for how a lower-level phenomenon is manifested at a higher level. The nature of elemental contributions, in type and amount, and the combination rules applicable to each exemplar are indicated. Although we have used the individual and group levels to make the examples easier to explain, the models are applicable to higher levels as well. The typology is intended to help elaborate the theoretical underpinnings that shape the conceptualization of alternative forms of emergence.
The exemplar for this type of emergence represents the ideal form of composition that we have discussed throughout this chapter. The model is based on the assumption that contextual factors and interaction processes constrain emergence in such a way that individuals contribute the same type and amount of elemental content. Therefore, the phenomenon converges around a common point that can be represented as a mean or a sum. For example, the performance of a crew rowing a scull is dependent on each individual providing the same amount and type of physical thrust at precisely the same time. Synchronized swimmers must execute the same movements, in the same amount, at the same time. Similarly, the notion of team mental models is predicated on all team members sharing the same amount and type of knowledge (Klimoski & Mohammed, 1995). Ideal composition is also illustrated by theory and research on collective climate and collective efficacy. Group members' perceptions converge on the referent construct. Sharing is evaluated on the basis of consensus or consistency. Variability in elemental content and individual contributions is very low and uniform in distribution across members. Therefore, aggregation to the group mean eliminates the small amount of error variance and effectively represents the group on the higher-level construct.
This exemplar relaxes the assumptions for the amount of elemental contribution, but the type of content remains similar. The underlying model is based on the assumption that contextual factors and interaction processes shape emergence in such a way that some minimum amount of contribution is required of each individual. Therefore, there will be restricted variability within the group, yielding a pattern across individuals that is relatively uniform and moderate in dispersion. An additive or averaging model combines the elemental contributions.
This exemplar fully relaxes the requirement on the amount of elemental contribution, but, as before, the type of content remains similar. Here, variation in the amount of elemental contribution can be quite high. For example, research demonstrates that performance in pooled tasks can be plagued by social loafing and free riding: some individuals contribute far less to the collective when the amount of their contributions cannot be identified (Harkins, Latane, & Williams, 1980). In such circumstances, the group product may be represented as a sum or mean. However, in contrast with the previous exemplar, the group representation and the individual contribution may be dramatically different. Similarly, one conceptualization of organizational climate is based on the assumption that within-group variation in climate perceptions is random measurement error (Glick, 1985, 1988). No restriction is placed on how much variability can be eliminated through averaging.
This exemplar represents a shift from linear combination rules (that is, additive models) to nonlinear rules. Elemental contribution is based on similar content, but the amount of contribution is qualitatively distinct. Contextual factors and interaction processes constrain emergence so that the pattern across individuals is discontinuous. The standing of one individual on the phenomenon in question determines the standing of the collective. Therefore, dispersion and uniformity are not directly applicable to the conceptualization of this exemplar.
Unlike the other exemplars, which focus on representative values to capture the emergent characteristic of the collective, this form of emergence represents the phenomenon as variability within the group. Conceptually, this form of emergence is related to heterogeneity (Klein et al., 1994), parts (Dansereau & Yammarino, Chapter Ten, this volume), and uniform dispersion (Brown & Kozlowski, 1997; Brown et al., 1996; Chan, 1998). The elemental contribution may be similar in type and amount (for example, norm crystallization) or different in type and amount (for example, demographic diversity). Therefore, individuals may make contributions that are similar or different, but the substantive focus is on the variance of contribution (Roberts et al., 1978). It is important to emphasize that this one form captures different types of emergence that may range from low dispersion to high dispersion.
For example, one form of creativity can be characterized by the diversity, or lack thereof, of the knowledge or perspectives that are brought to bear on a problem (Wiersema & Bantel, 1992). Demographic diversity captures the extent to which individual members of a unit differ in their demographic characteristics (Tsui, Egan, & O'Reilly, 1992; Jackson et al., 1995). Homogeneity of charisma (that is, the extent to which a leader has equally charismatic relationships with all of his or her subordinates; see Klein & House, 1995), norm crystallization (Jackson, 1975), and culture strength (Koene, Boone, & Soeters, 1997) are based on variability within a collective. Homogeneity, crystallization, and strength are predicated on low variance, whereas the absence of homogeneity, crystallization, and strength is indicated by high variance. Klein and colleagues (Chapter Six, this volume) explore the antecedents and consequences of variability in organizational boundary spanners' trust in and commitment to their organization's interorganizational partner. Variance, of course, is a key operationalization of variability. Variance can capture emergence that differs across groups, contexts, and time. Therefore, it represents a shift in conceptual focus, from the content of the phenomenon to the nature of emergence itself.
This model is based on the widest variability in the type and amount of elemental contribution, and in the patterns by which those differences combine to represent emergent phenomena. This model incorporates the assumption that emergence may manifest itself as different forms, and it views nonuniform patterns of dispersion as meaningful substantive phenomena.
We introduced this third and last section of the chapter with three intentions: to be inclusive and expansive in our consideration of alternative forms of emergence, to focus on building a theoretical foundation for different forms of emergence, and to use typology as a vehicle for explicating and elaborating on the theoretical underpinnings of emergence. We hope that we have, in some measure, accomplished these goals. We believe, as we shall describe, that our framework is largely consistent with other efforts to explore emergence. We also believe that our particular attention to the underlying processes and dynamics that shape different forms of emergence can enhance understanding of the moderator effects and boundary conditions affecting emergence. An appreciation of the influence of these processes will lead to more precise specification of the theory addressing emergent phenomena. We see our effort as a point of departure for guiding and pushing further theoretical elaboration.
It is interesting to us that when our effort was originally conceived, we viewed our focus on different forms of emergence, and on the processes that shape those forms, as novel. However, a number of other researchers, contemporaneous with the development of this chapter, have also started to explore emergence (Brown & Kozlowski, 1997, 1999; Brown et al., 1996; Chan, 1998; Kozlowski, 1998, 1999; Morgeson & Hofmann, 1999a, 1999b). Although this chapter is not intended as an integration of these efforts, we believe that our framework helps to make explicit the compatibilities across these apparently disparate efforts to explore emergence. For example, Brown and Kozlowski (1997, 1999) posit dispersion theory, which focuses on patterns of within-group variability or the dispersion of phenomena, as opposed to the more common focus on means or convergent points. In dispersion theory, uniform patterns that evidence low dispersion are consistent with composition processes, whereas subgroup bifurcation that creates nonuniform patterns of dispersion are consistent with compilation processes. Similarly, Morgeson and Hofmann (1999a, 1999b) have made a strong case for distinguishing construct structure and function. Structural and functional identity across levels is consistent with composition processes, and functional but not structural identity across levels is consistent with compilation processes.
We would be remiss if we did not note that there are also apparent inconsistencies between the contemporary treatments of emergence (just noted) and other treatments with a tradition in the literature. We see the treatments as compatible yet different efforts to understand the same general class of phenomena. For example, the varient paradigm (Dansereau et al., 1984) treats emergence as a relationship between variables that exists at a higher, collective level but that does not hold between similar variables at a lower level. Thus, for example, a relationship between two variables is said to emerge at the group level of analysis if the two variables are significantly related (both statistically and practically) at the group level of analysis but the relationship between the two variables is not significant at the individual level of analysis. The varient perspective on emergence and our perspective are related but distinct. Dansereau and his colleagues focus on the emergence of relationships between variables at higher unit levels and on the statistical detection of such relationships. In contrast, we have focused primarily on the emergence of higher-level constructs, endeavoring to show the variety of ways in which a higher-level construct may emerge from lower-level entities and interaction processes. Measurement and analysis are important but separable issues. Ultimately, specific theories that assume particular emergent forms will need to be tested empirically. The varient paradigm, other analytic approaches, and even new techniques will be useful in this process.
As the next millennium approaches, we are poised to witness a renaissance in organizational theory and research. There is increasing recognition that the confines of single-level models-a legacy of primary disciplines that undergird organizational science-need to be broken. A meaningful understanding of the phenomena that comprise organizational behavior necessitates approaches that are more integrative, that cut across multiple levels, and that seek to understand phenomena from a combination of perspectives. There is a solid theoretical foundation for a broadly applicable levels perspective, for an expanding, empirically based research literature, and for progress toward the development of new and more powerful analytic tools. A levels perspective offers a paradigm that is distinctly organizational.