793

CHAPTER 14

The Bioecological Model ofHuman DevelopmentURIE BRONFENBRENNER and PAMELA A. MORRIS

OVERVIEW 795DEFINING PROPERTIES OF THE

BIOECOLOGICAL MODEL 796Proposition I 797Proposition II 798FROM THEORY TO RESEARCH DESIGN:

OPERATIONALIZING THEBIOECOLOGICAL MODEL 799

Developmental Science in the Discovery Mode 801Different Paths to Different Outcomes: Dysfunction

versus Competence 803The Role of Experiments in the Bioecological Model 808HOW DO PERSON CHARACTERISTICS INFLUENCE

LATER DEVELOPMENT? 810Force Characteristics as Shapers of Development 810Resource Characteristics of the Person as Shapers

of Development 812Demand Characteristics of the Person as

Developmental Inf luences 812THE ROLE OF FOCUS OF ATTENTION IN

PROXIMAL PROCESSES 813

PROXIMAL PROCESSES IN SOLO ACTIVITIESWITH OBJECTS AND SYMBOLS 814

THE MICROSYSTEM MAGNIFIED: ACTIVITIES,RELATIONSHIPS, AND ROLES 814

Effects of the Physical Environment onPsychological Development 814

The Mother-Infant Dyad as a Context of Development 815BEYOND THE MICROSYSTEM 817The Expanding Ecological Universe 818Nature-Nurture Reconceptualized:

A Bioecological Interpretation 819TIME IN THE BIOECOLOGICAL MODEL:

MICRO-, MESO-, ANDMACROCHRONOLOGICAL SYSTEMS 820

FROM RESEARCH TO REALITY 822THE BIOECOLOGICAL MODEL: A

DEVELOPMENTAL ASSESSMENT 824REFERENCES 825

The bioecological model, together with its correspon-ding research designs, is an evolving theoretical systemfor the scientific study of human development over time(Bronfenbrenner, 2005). In the bioecological model, de-velopment is defined as the phenomenon of continuityand change in the biopsychological characteristics ofhuman beings, both as individuals and as groups. Thephenomenon extends over the life course, across succes-

We are especially grateful for the thoughtful criticisms ofearlier drafts of the manuscript generously provided by thefollowing colleagues: Jay Belsky, Rick Canfield, Nancy Dar-ling, Glen H. Elder Jr., Steven F. Hamilton, Melvin L. Kohn,Kurt Lscher, Phyllis Moen, Donna Dempster-McLain, Lau-rence Steinberg, and Sheldon H. White. We owe particularthanks to Professor Susan Crockenberg and her students at the University of Vermont who, in the course of a graduate

sive generations, and through historical time, both pastand future. The term future raises a question: How is itpossible to scientifically investigate phenomena thathave not yet taken place? This question is hardly new;indeed, it pervades every field of scientific endeavor.However, we are the only species that, over historicaltime, has developed the capacity to engage successfullyin scientific inquiry, and thereby, in many respects, has

seminar, carefully reviewed a draft of this chapter, and mademany constructive suggestions. We have done our best tomeet the high standards that they commendably set. We wishto express gratitude to Richard M. Lerner and WilliamDamon, the editors of the 1998 Volume and of that series as awhole, for their wise advice, encouragement, and patience.Finally, a special thanks goes to our most severe and mostconstructive critic, Liese Bronfenbrenner.

794 The Bioecological Model of Human Development

been able to change the nature of the world in which welive. As a result, within certain limits, we humans havealtered the nature and course of our own development asa species (Bronfenbrenner & Evans 2000; Bronfenbren-ner & Morris 1998).

To place bioecological theory of human developmentinto a larger context, it is important to recognize thatmany of the general perspectives advanced and elabo-rated in this theory are also parts of other related linesof theoretical and empirical inquiry into human devel-opment. Examples include life-span psychology (Baltes,Lindenberger, & Staudinger, Chapter 11, this Handbook,this volume), cultural psychology (Cole, 1995; Shwederet al., Chapter 13, this Handbook, this volume), Magnus-sons developmental theory of contextual-interactiveholism (Magnusson & Stattin, Chapter 8, this Hand-book, this volume), and, especially, the work of RobertCairns (Chapter 3, this Handbook, this volume), whothrough communications and publications extendingover 3 decades, has played a major role in the evolutionof the four defining properties of the bioecologicalmodel: (1) Process, (2) Person, (3) Context, and (4)Time. Cairns is best known as the founder and principalprotagonist of developmental science, and there are sev-eral excellent examples of his books and articles thathave been most relevant to the evolution of the bioeco-logical model (Bergman, Cairns, Nilsson, & Nysted,2000; Cairns, 1970; Cairns & Cairns, 1994). The spe-cific profile of the bioecological model of human devel-opment is its interdisciplinary and integrative focus onthe age periods of childhood and adolescence and its ex-plicit interest in applications to policies and programspertinent to enhancing youth and family development.

In this chapter, we undertake to present the ecologi-cal model of human development that has been intro-duced over the course of the prior two editions of thisHandbook (Bronfenbrenner & Crouter, 1983; Bronfen-brenner & Morris, 1998). The main focus of the 1983chapter was on the empirical and theoretical roots of amodel already in use, which centered on the role of theenvironment in shaping development. In contrast, thischapter is oriented toward the future. The present modelintroduces major theoretical innovations from the 1983chapter in both form and content. The present formula-tion makes no claim as a paradigm shift (if there be sucha phenomenon); rather, it continues a marked shift in thecenter of gravity of the model, in which features of ear-lier versions are first called into question but then re-

combined, along with new elements, into a more com-plex and more dynamic structure.

The transition in the form and content of the modelactually took place over an extended period of time, anexpression that will become all too familiar to the reader(Bronfenbrenner, 2005). The transition from a focus onthe environment to a focus on processes was first intro-duced in the context of Bronfenbrenners unpublishedlectures, colloquium presentations, and contributionsto symposia. Not until 1986, did reference to an emer-gent new model first appear in print (Bronfenbrenner,1986b). The following extended excerpt conveys both itsspirit and intended substance. Because both of these at-tributes are relevant to the gradual evolution of themodel to its present form, we quote from the 1986 state-ment at some length:

It is now more than a decade ago that, being somewhatyounger, I presumed to challenge the then-prevailing con-ventions of our field by describing the developmental re-search of the day as the study of the strange behavior ofchildren in strange situations for the briefest possible pe-riod of time (Bronfenbrenner, 1974). Instead, I argued (asif it were simply a matter of choice), we should be studyingdevelopment in its ecological context; that is, in the actualenvironments in which human beings lived their lives. Ithen proceeded to outline, in a series of publications, a con-ceptual framework for analyzing development in context,and to offer concrete examples of how various elements ofthe schema might be applied both to past studies and tostudies yet-to-come. I also emphasized the scientific andpractical benefits of a closer linkage, in both directions, be-tween developmental research and public policy (Bronfen-brenner, 1974, 1975, 1977a, 1977b, 1979a, 1979b, 1981).Now, a dozen years later, one might think that I have goodreason to rest content. Studies of children and adults inreal-life settings, with real-life implications, are now com-monplace in the research literature on human development,both in the United States and, as this volume testifies, inEurope as well. This scientific development is taking place,I believe, not so much because of my writings, but ratherbecause the notions I have been promulgating are ideaswhose time has come. . . .

Clearly, if one regards such scientific developments asdesirable, there are grounds for satisfaction. Yet, alongwith feelings of gratification, I must confess to some dis-content. My disquiet derives from two complementaryconcerns. The first pertains to one of the main roads thatcontemporary research has taken; the second, to somemore promising pathways that are being neglected.

Overview 795

Alas, I may have to accept some responsibility for whatI regard as the wayward course. It is an instance of whatmight be called the failure of success. For some years, Iharangued my colleagues for avoiding the study of devel-opment in real-life settings. No longer able to complain onthat score, I have found a new bte noir. In place of toomuch research on development out of context, we nowhave a surfeit of studies on context without development.

One cannot presume to make so brass an allegationwithout being prepared to document ones case. I am pre-pared. (Bronfenbrenner 1986a, pp. 286288)

What followed was an early version of the newly evolv-ing theoretical framework, but the purpose of the pres-ent chapter is better served by presenting the model inits current, albeit still-evolving, form now called thebioecological model. The term evolving highlights thatthe model, along with its corresponding research de-signs, has undergone a process of development during itslife course (Bronfenbrenner, 2005). The bioecologicalmodel addresses two closely related but fundamentallydifferent developmental processes, each taking placeover time. The first process defines the phenomenonunder investigationcontinuity and change in thebiopsychological characteristics of human beings. Thesecond focuses on the development of the scientifictoolstheoretical models and corresponding researchdesigns required for assessing continuity and change.

These two tasks cannot be carried out independently,for they are the joint product of emerging and converg-ing ideas, based on both theoretical and empiricalgroundsa process called developmental science in thediscovery mode (Bronfenbrenner & Evans 2000,pp. 9991000). In the more familiar verification mode,the aim is to replicate previous findings in other settingsto make sure that the findings still apply. By contrast, inthe discovery mode, the aim is to fulfill two broader butinterrelated objectives:1. Devising new alternative hypotheses and correspon-

ding research designs that not only question existingresults but also yield new, more differentiated, moreprecise, replicable research findings and, thereby,produce more valid scientific knowledge.

2. Providing scientific bases for the design of effectivesocial policies and programs that counteract newlyemerging developmentally disruptive influences.This has been an explicit objective of the bioecologi-cal model from its earliest beginnings. To orient the

reader to the present formulation of the biologicalmodel, a preview follows.

OVERVIEW

We begin with an exposition of the defining propertiesof the model, which involves four principal componentsand the dynamic, interactive relationships among them.The first of these, which constitutes the core of themodel, is Process. More specifically, this construct en-compasses particular forms of interaction between or-ganism and environment, called proximal processes, thatoperate over time and are posited as the primary mech-anisms producing human development. However, thepower of such processes to influence development ispresumed, and shown, to vary substantially as a func-tion of the characteristics of the developing Person, ofthe immediate and more remote environmental Contexts,and the Time periods, in which the proximal processestake place.

The sections that follow examine in greater detail eachof the three remaining defining properties of the model,beginning with the biopsychological characteristics ofthe Person. This domain was given sequential priority tofill a recognized gap in earlier prototypes of the ecologi-cal model. Thus, at midstage in the development of thepresent model, Bronfenbrenner criticized its theoreticalpredecessors and acknowledged his share of responsibil-ity for failing to deliver on an empirical promise:

Existing developmental studies subscribing to an ecologi-cal model have provided far more knowledge about the na-ture of developmentally relevant environments, nearand far, than about the characteristics of developing indi-viduals, then and now. . . . The criticism I just made alsoapplies to my own writings. . . . Nowhere in the 1979monograph, nor elsewhere until today, does one find aparallel set of structures for conceptualizing the charac-teristics of the developing person. (Bronfenbrenner,1989a, p. 188)

Three types of Person characteristics are distin-guished as most influential in shaping the course offuture development through their capacity to affectthe direction and power of proximal processes throughthe life course. First, dispositions can set proximalprocesses in motion in a particular developmentaldomain and continue to sustain their operation. Next,

796 The Bioecological Model of Human Development

bioecological resources of ability, experience, knowl-edge, and skill are required for the effective function-ing of proximal processes at a given stage ofdevelopment. Finally, demand characteristics invite ordiscourage reactions from the social environmentthat can foster or disrupt the operation of proximalprocesses. The differentiation of these three formsleads to their combination in patterns of Person struc-ture that can further account for differences in the di-rection and power of resultant proximal processes andtheir developmental effects.

These new formulations of qualities of the personthat shape his or her future development have had theunanticipated effect of further differentiating, expand-ing, and integrating the original 1979 conceptualiza-tion of the environment in terms of nested systemsranging from micro to macro (Bronfenbrenner, 1979b).For example, the three types of Person characteristicspreviously outlined are also incorporated into the defi-nition of the microsystem as characteristics of parents,relatives, close friends, teachers, mentors, coworkers,spouses, or others who participate in the life of the de-veloping person on a fairly regular basis over extendedperiods of time.

The bioecological model also introduces an evenmore consequential domain into the structure of the mi-crosystem that emphasizes the distinctive contributionto development of proximal processes involving inter-action not with people but with objects and symbols.Even more broadly, concepts and criteria are introducedthat differentiate between those features of the environ-ment that foster versus interfere with the development ofproximal processes. Particularly significant in the lattersphere is the growing hecticness, instability, and chaosin the principal settings in which human competenceand character are shapedin the family, child-carearrangements, schools, peer groups, and neighborhoods.

The latter theme speaks to the fourth and final defin-ing property of the bioecological model and the one thatmoves it farthest beyond its predecessorthe dimensionof Time. The 1979 Volume scarcely mentions the term,whereas in the current formulation, it has a prominentplace at three successive levels: (1) micro-, (2) meso-,and (3) macro-. Microtime refers to continuity versusdiscontinuity in ongoing episodes of proximal process.Mesotime is the periodicity of theses episodes acrossbroader time intervals, such as days and weeks. Finally,Macrotime focuses on the changing expectations andevents in the larger society, both within and across gen-

erations, as they affect and are affected by, processesand outcomes of human development over the lifecourse. The treatment of this last topic draws on Elderand Shanahan, Chapter 12, this Handbook, this volume.Our primary emphasis, however, is on the role of devel-opmental processes and outcomes in producing large-scale changes over time in the state and structure of thebroader society over time, and the implications of thosechanges for the societys future.

Before turning to the task at hand, it is important tomake explicit three overarching orientations that definethe content and the structure of the chapter as a whole.First, we use the term development to refer to stabilityand change in the biopsychological characteristics ofhuman beings over the life course and across genera-tions. There are no restrictive assumptions of changefor the better or of continuity in the characteristics ofthe same person over time. Rather, these are issues tobe investigated.

Second, from the perspective of the bioecologicalmodel, the forces producing stability and change in thecharacteristics of human beings across successive gen-erations are no less important than stability and changein the characteristics of the same person over his orher lifetime.

The third orientation is perhaps the most essential,and the most difficult to achieve. It was Kurt Lewin(cited in Marrow, 1977) who said that there is nothing sopractical as a good theory. But to be good, a theorymust also be practical. In science, a good theory is onethat can be translated into corresponding research de-signs that match the defining properties of the theory. Inthe absence of such research designsor worse yet, inthe application of research designs that fail to match oreven violate the defining properties of the theorysci-ence cannot move forward. Hence, we have sought, as weproceed through successive stages of theoretical formula-tion, to specify, and, wherever possible, to illustrate theproperties of a research design that corresponds with, orat least approximates, the proposed theoretical structure.

DEFINING PROPERTIES OF THEBIOECOLOGICAL MODEL

An early critical element in the definition of the bioeco-logical model is experience, which indicates that thescientifically relevant features of an environment forhuman development not only include its objective prop-

Defining Properties of the Bioecological Model 797

erties but also the way in which the properties are sub-jectively experienced by the person living in that envi-ronment. This equal emphasis on an experiential as wellas an objective view springs neither from an antipathy tobehaviorist concept nor from a predilection for existen-tial philosophic foundations but is dictated simply by thefact that very few of the external influences signifi-cantly affecting human behavior and development canbe described solely in objective physical conditions andevents (Bronfenbrenner & Evans 2000; Bronfenbrenner& Morris 1998).

Critical to the foregoing formulation is the wordsolely. In the bioecological model, both objective andsubjective elements are posited as driving the course ofhuman development; neither alone is presumed suffi-cient. Moreover, these elements do not always operate inthe same direction. It is therefore important to under-stand the nature of each of these two dynamic forces,beginning on the phenomenological or experiential side.Both of the terms are relevant because, while related toeach other, they are typically applied to somewhat dif-ferent spheres. Experiential is more often used in rela-tion to cognitive development and pertains mainly tochanges in how the environment is perceived at succes-sive stages of the life course, beginning in early infancyand proceeding through childhood, adolescence, adult-hood, and, ultimately, old age.

By contrast, experience pertains more to the realm offeelingsanticipations, forebodings, hopes, doubts, orpersonal beliefs. Feelings, emerging in early childhoodand continuing through life, are characterized by bothstability and change: They can relate to self or to others,especially to family, friends, and other close associates.They can also apply to the activities in which we engage;for example, those that we most or least like to do. But themost distinctive feature of such experiential equalities isthat they are emotionally and motivationally loaded, en-compassing both love and hate, joy and sorrow, curiosityand boredom, desire and revulsion, often with both polar-ities existing at the same time but usually in differing de-grees. A significant body of research evidence indicatesthat such positive and negative subjective forces, evolvingin the past, can also contribute in powerful ways to shap-ing the course of development in the future (Bronfen-brenner & Evans 2000; Bronfenbrenner & Morris 1998).

But these forces are not the only powerful ones atwork, other forces are more objective in nature. Thispresence does not mean, however, that the forces arenecessarily either more or less influential, mainly be-

cause the two sets of forces are interdependent and af-fect each other. Like their subjective counterparts, thesemore objective factors also rely on their assessment ofcorresponding theoretical models and associated re-search designs, which evolved over time. These moreobjective relationships are documented propositionspresented later (see too Bronfenbrenner & Evans 2000;Bronfenbrenner & Morris 1998). The first propositionspecifies the theoretical model, and provides concreteexamples; the second foreshadows a corresponding re-search design for their assessment.

However, before proceeding with formal definitions,it may be useful to point out that traditionally such phe-nomena as parent-child interactionor, more generally,the behavior of others toward the developing personhave been treated under the more inclusive category ofthe environment. In the bioecological model, a criticaldistinction is made between the concepts of environ-ment and process, with the latter not only occupying acentral position, but also having a meaning that is quitespecific. The construct appears in Proposition I stipulat-ing the defining properties of the model. To place itsmeaning in context, we cite Proposition II as well.

Proposition I

Especially in its early phases, but also throughout the lifecourse, human development takes place through processesof progressively more complex reciprocal interaction be-tween an active, evolving biopsychological human organismand the persons, objects, and symbols in its immediate ex-ternal environment. To be effective, the interaction mustoccur on a fairly regular basis over extended periods oftime. Such enduring forms of interaction in the immediateenvironment are referred to as proximal processes. Exam-ples of enduring patterns of proximal process are found infeeding or comforting a baby, playing with a young child,child-child activities, group or solitary play, reading, learn-ing new skills, athletic activities, problem solving, caringfor others in distress, making plans, performing complextasks, and acquiring new knowledge and know-how.

For the younger generation, participation in such in-teractive processes over time generates the ability, moti-vation, knowledge, and skill to engage in such activitiesboth with others and on your own. For example, throughprogressively more complex interaction with their par-ents, children increasingly become agents of their owndevelopment, to be sure only in part.

798 The Bioecological Model of Human Development

Proximal processes are posited as the primary en-gines of development (see Gottlieb, Wahlsten, & Lick-liter, Chapter 5, this Handbook, this volume; Tobach,1981; Tobach & Schneirla, 1968). A second definingproperty, the fourfold source of these dynamic forces isidentified in Proposition II.

Proposition II

The form, power, content, and direction of the proximalprocesses effecting development vary systematically as ajoint function of the characteristics of the developing per-son, the environmentboth immediate and more remotein which the processes are taking place, the nature of thedevelopmental outcomes under consideration, and the so-cial continuities and changes occurring over time throughthe life course and the historical period during which theperson has lived.

Propositions I and II are theoretically interdepend-ent and subject to empirical test. An operationalresearch design that permits their simultaneous inves-tigation is referred to as the Process-Person-Context-Time (PPCT) model.

Characteristics of the person actually appear twicein the bioecological modelfirst as one of the four ele-ments influencing the form, power, content, and directionof the proximal process, and then again as developmentaloutcomesqualities of the developing person thatemerge at a later point in time as the result of the joint,interactive, mutually reinforcing effects of the four prin-cipal antecedent components of the model. In sum, in thebioecological model, the characteristics of the personfunction both as an indirect producer and as a product ofdevelopment (see Lerner, 1982, 2002; Lerner & Busch-Rossnagel, 1981).

Finally, because in the bioecological model the con-cept of proximal process has a specific meaning, it isimportant that its distinctive properties be made ex-plicit. For present purposes, the following features ofthe construct are especially noteworthy:

1. For development to occur, the person must engage inan activity.

2. To be effective, the activity must take place on afairly regular basis, over an extended period oftime. For example, this means that with young chil-dren, a weekend of doing things with Mom or Dad

does not do the job, nor do activities that are ofteninterrupted.

3. Why not? One reason is that, to be developmentallyeffective, activities must continue long enough tobecome increasingly more complex. Mere repeti-tion does not work.

4. Developmentally effective proximal processes arenot unidirectional; there must be influence in both directions. For interpersonal interaction, thismeans that initiatives do not come from one sideonly; there must be some degree of reciprocity inthe exchange.

5. Proximal processes are not limited to interactionswith people; they also can involve interaction withobjects and symbols. In the latter circumstance, forreciprocal interaction to occur, the objects and sym-bols in the immediate environment must be of akind that invites attention, exploration, manipulation,elaboration, and imagination.

6. The powerful moderating factors specified in Propo-sition II produce substantial changes in the content,timing, and effectiveness of proximal processes. Inparticular:a. As children grow older, their developmental

capacities increase both in level and range;therefore, to continue to be effective, the corre-sponding proximal processes must also becomemore extensive and complex to provide for thefuture realization of evolving potentials. At thesame time, in view of the ongoing developmentaladvance, the intervals between periods of pro-gressively more complex activity can be in-creasingly longer, although they must still occuron a fairly regular basis. Otherwise, the paceof development slows, or its course may even reverse direction.

b. The principal persons with whom young childreninteract on a fairly regular basis over extendedperiods of time are parents, but especially aschildren get older, other personssuch as care-givers, relatives, siblings, and peersalso func-tion in this role. These are soon followed byteachers or mentors in other activities, and thenby close friends of the same or opposite sex,spouses or their equivalents, and coworkers, supe-riors and subordinates at work. As the examplesindicate, the involvement of persons functioning

From Theory to Research Design: Operationalizing the Bioecological Model 799

in this role is not limited to the formative years.Borrowing a term from G. H. Mead (1934), werefer to such persons as significant others.

The foregoing constitute the principal elements of theemergent theoretical model. If so, the question arises inwhat sense is the model bioecological? Where and howdoes biology come into the picture? We present three an-swers to that question in an order of decreasing cer-tainty about their validity. The first is an unqualifieddisclaimer. Little in the pages that follow speaks to theoperation of biological systems within the organism.By contrast, considerable scientific attention is ac-corded to characteristics of the person generally re-garded as biologically based that influence proximalprocesses and their developmental outcomes. Finally,the present model rests on the assumption that biologi-cal factors and evolutionary processes not only set limitson human development but also impose imperatives re-garding the environmental conditions and experiencesrequired for the realization of human potentials. The po-sition is taken that, to the extent that the necessary con-ditions and experiences are not provided, such potentialswill remain unactualized (Bronfenbrenner & Ceci,1993, 1994a, 1994b).

It is our belief that, when applied, the bioecologicalparadigm is scientifically productive. At the presenttime, however, its most distinguishing characteristic isnot its proven scientific power, but its rarity. To be sure,the rarity is hardly surprising, given the fact that suc-cessive revisions of the emerging model began to bepublished only in the past several years (Bronfenbren-ner, 1989a, 1990, 1993, 1994, 1995; Bronfenbrenner &Morris, 1998; Bronfenbrenner & Ceci, 1994a). Paradox-ically, some concrete examples nevertheless existedmuch earlier. They were the product of what Bronfen-brenner and Crouter referred to in the 1983 edition ofthis Handbook as latent paradigms; that is, theoreticalmodels that were not explicitly stated, but were implicitin the research designs used in analyzing the data(Bronfenbrenner & Crouter, 1983, pp. 373376). In-deed, a partial precursor of the bioecological model ap-peared in the 1983 Handbook chapter under the rubric ofa person-process-context model. In that chapter, how-ever, what is meant by process is never specified, andthe overwhelming majority of the examples cited do notinclude a proximal process component as defined inProposition I. The same holds true for developmentally

relevant characteristics of the Person. The 1983 chapteralso made no reference to Time as a defining property ofthe theoretical model. In these and other respects to fol-low, todays bioecological model goes far beyond itspredecessors both with respect to basic constructs andtheir bidirectional, synergistic interrelationships.

FROM THEORY TO RESEARCH DESIGN:OPERATIONALIZING THEBIOECOLOGICAL MODEL

We have come to the point where it is both possible andnecessary to examine the requirements imposed by thebioecological model for corresponding research designs.We begin with a concrete example of the latter.

In the 1950s and 1960s, Cecil Mary Drillien (1957,1964), a physician and professor of child life and healthat the University of Edinburgh, carried out a 7-year lon-gitudinal investigation of psychological development intwo groups: 360 children of low birthweight and a con-trol group selected by taking the next mature birthfrom the hospital admission list (1957, p. 29). In herfollow-up assessments, the investigator found that chil-dren of low birthweight were more likely to exhibitproblems in physical growth, susceptibility to illness,impaired intellectual development, and poorer class-room performance, with all of these tendencies beingmore pronounced in boys (1964). In a comparison ofchildrens school performance with what would havebeen expected on the basis of their scores on an intelli-gence test, Drillien found that those of low birthweightwere especially likely to be working below their mentalcapacity. In relation to this finding, the author com-ments as follows: In most cases, failure to attain a stan-dard commensurate with ability was associated withproblems of behavior, which were found to increase withdecreasing birthweight [and] to be more common inmales (1964, p. 209).

Figure 14.1 depicts the results. The figure does not ap-pear in Drilliens monograph, but was constructed fromdata presented in tables in that volume. It shows the im-pact of the quality of mother-infant interaction at age 2on the number of observed problem behaviors at age 4 asa joint function of social class and three levels of lowbirthweightthose underweight by a pound or more, notmore than one pound, and those of normal birthweight.Measures of maternal responsiveness were based on

800 The Bioecological Model of Human Development

observations in the home and interviews with the mother.The investigators measure of social class was a compos-ite index that took into account not only parental incomeand education but also the socioeconomic level of theneighborhood in which the family lived. The quality ofinteraction was assessed by extent to which the motherwas responsive to changes in the state and behavior of theinfant. The measure of the developmental outcome wasthe frequency of reported behavior disturbances such ashyperactivity, overdependence, timidity, and negativism.

Our primary interest is not in the research findings,but in the extent to which the structure of the research de-sign corresponds with the defining properties of the bioe-cological theoretical model. The first point to be noted inthis regard is that Proposition I defines ProximalProcesses as bidirectional. Drilliens measure of process,however, was based only on the mothers responsivenessto changes in the state and behavior of the infant, and nodata are reported that would permit calculating a comple-mentary measure of the infants responsiveness tochanges in the state and behavior of the mother. Thismeans that the operational measure available in Drilliensresearch taps only one side of the theoretical definition ofproximal process. For that reason, it appears likely that,to the extent the infants contribution to reciprocal inter-action carries any weight, the obtained results may un-derestimate the true magnitude of the observed effects.

Nevertheless, as revealed in Figure 14.1, maternal re-sponsiveness across time, a one-sided measure of proxi-

1 Synergism refers to cooperative action of discrete agenciessuch that the total effect is greater than the sum of the two ormore effectstaken independently (Websters Third NewInternational Dictionary).

mal process, still emerges as the most powerful predictorof developmental outcome. In all instances, responsivematernal treatment reduces substantially the degree ofbehavioral disturbance exhibited by the child.

Herein lies the main justification for distinguishingbetween proximal process on the one hand, and, onthe other, the environments in which the processesoccur; namely, in accord with Proposition I, proximalprocesses turn out to be the most potent force influenc-ing the developmental outcome (in this case, the fre-quency of problem behaviors at 4 years of age).Furthermore, as stipulated in Proposition II, the powerof the Process varies systematically as a function of theenvironmental Context (i.e., social class) and of thecharacteristics of the Person (i.e., weight at birth). Theprocess appears to have made its greatest impact onyoung children growing up in the most disadvantagedenvironment (i.e., the lowest socioeconomic level), butin that environment, it is those who at birth were of nor-mal weight who benefited most. Moreover, it was in thissame disadvantaged Context that, under high levels ofmaternal responsiveness, birthweight showed its mostconsistent effect, with the number of behavior problemssteadily rising as birthweight fell. Finally, across theboard, maternal responsiveness had the general result ofdecreasing or buffering against environmental differ-ences in developmental outcome. Thus, at high levels ofmother-child interaction, social class differences inproblem behavior became much smaller.

From the perspective of developmental science, whatis most noteworthy about these findings is not their spe-cific content but that their simultaneous discovery wasmade possible by a research design based on a theoreticalmodel that allowed for the emergence of patterns of thisform. Not only are the four key components of Process,Person, Context, and Time all represented but the designalso provides for the detection of the kinds of synergis-tic1 interdependencies among these components that areposited in the bioecological model as a dynamic theoret-ical system. Two specific examples of such interdepen-dencies are revealed in the analysis of Drilliens data:

1. Proposition II stipulates that the developmental ef-fects of proximal processes vary as a joint function of

Figure 14.1 Effect of mothers responsiveness on problembehavior of child at age 4 by birthweight and social class.

High SES Middle SES Low SESNrm Low Vlow Nrm Low Vlow Nrm Low Vlow

Aver

age

Scor

e on

Beh

avio

r Pro

blem

s

0

2

4

6

8

10

12

Nrm = Normal BirthweightLow = Between Normal and 5.5 lbs.Vlow = 5.5 lbs. or Less

Poor ProcessGood Process

From Theory to Research Design: Operationalizing the Bioecological Model 801

Figure 14.2 Effect of mothers responsiveness on problembehavior of child at ages 2 and 4 by social class.

0

2

4

6

8

10

Age 2High SES Mid SES Low SES

Aver

age

Scor

e on

Beh

avio

r Pro

blem

s

Age 4High SES Mid SES Low SES

Poor ProcessUnresponsive MothersGood ProcessResponsive Mothers

Person and Context; that is, the indirect effects ofPerson and Context on the relation of Process to out-come are not to be conceived as simply additive.Consistent with this expectation is the finding thatproximal processes had their greatest impact in themost disadvantaged environment but on the healthiestinfant. The combination of Person and Context ex-hibit a mutually reinforcing, multiplicative, indirecteffect on the power of proximal processes as the en-gines of development.

2. In Drilliens research, the frequency of problem be-haviors was assessed at two points in timefirstwhen the infants were 2-years-old, and then again at4-years-old. If one makes the not unreasonable as-sumption that mothers continued to interact withtheir children over the intervening period, then theresults shown in Figure 14.2 provide evidence for theeffect of proximal processes that have taken placeover an extended period of time. Youngsters experi-encing low levels of interaction with their mothersexhibited an accelerating increase in the number ofproblem behaviors from 2 to 4 years of age, whereasthose exposed to substantially higher levels of thisproximal process showed only a modest rise.

Developmental Science in the Discovery Mode

What about the possibility that the preceding results arechance findings? Some of them are statistically signifi-cant, yet others could not be tested because the variancesneeded for calculating error estimates were not reported.But that is not the principal issue at stake. With concrete

examples of the relation between theoretical and opera-tional models now before us, we can address what turnsout to be a complex and consequential question: What isthe function of research design in the bioecologicalmodel? The first point to be made in this regard is that themain function is not the usual one of testing for statisticalsignificance. Rather, the research design must provide forcarrying out an equally essential and necessarily priorstage of the scientific process: that of developing hypothe-ses of suf ficient explanatory power and precision to war-rant being subjected to empirical test. We are dealing withscience in the discovery mode rather than in the mode ofverification. In this earlier phase, theory plays an evenmore critical role. From its very beginnings, the bioeco-logical model, through its successive reformulations, rep-resents a sustained effort to meet this scientific need.

What are the appropriate characteristics of researchdesigns for developmental science in the discoverymode? Finding an answer to this question is complicatedby the fact that, compared with the physical and naturalsciences, developmental science is admittedly still in anearlier stage of development. Furthermore, because itsscope falls between the natural and the social sciences,the discovery process must to some extent be adapted tothe requirements of both. Perhaps in part for these rea-sons, we were unable to find any discussion of the issuein the developmental literature. Under these circum-stances, we concluded that the best we could do was totry to make explicit the characteristics of the researchdesigns that had been employed over the past severalyears to arrive at successively more differentiated for-mulations of the bioecological model.

These design characteristics depend on the con-structs, and the possible relations between them, thatare posited in the theoretical model at its present stageof development. Both the constructs, and the possible in-terrelationships, have been indicated in Propositions Iand II, but as yet they appear in a relatively undifferen-tiated form. For example, the directions of the expectedeffects of Person and Context on proximal processes fordifferent types of outcomes are not specified. The rea-son for such lack of specificity is that a more preciseformulation could not be deduced either from the theoryin its present, still evolving state, or induced from anyalready available data (at least, to our knowledge).Given these limitations, we concluded that an appropri-ate design strategy at this point in the discovery processcould be one that involves a series of progressively moredifferentiated formulations and corresponding data

802 The Bioecological Model of Human Development

analyses, with the results at each successive step settingthe stage for the next round. The research designs em-ployed must be primarily generative rather than confir-matory versus disconfirming.

In this generative process, implications derived fromthe theoretical model play a more prominent role thanthose drawn from research findings, but the latter arealso critical. Their importance is best conveyed by spec-ifying a key feature of the corresponding research de-sign: It must provide a structured framework fordisplaying the emergent research findings in a way thatreveals more precisely the pattern of the interdependen-cies that are obtained in the data available. Of primaryscientific interest are not those aspects of the observedpattern already anticipated in the existing theoreticalmodel, but those features that point to more differenti-ated and precise theoretical formulations. These canthen be evaluated in the light of new evidence, and, ifdeemed scientifically promising, can be incorporated inthe research design for a next step. The proposed strat-egy for developmental investigations in the discoverymode involves an iterative process of successive con-frontations between theory and data leading toward theultimate goal of being able to formulate hypotheses thatboth merit and are susceptible to scientific assessmentin the verification mode.

In presenting this definition of the discovery mode,we acknowledge that, in actual scientific practice, it ishardly likely to be a discovery. The process we have de-scribed, or something like it, is what scientists have al-ways done. Our primary reason for seeking to make thatprocess explicit was the belief that doing so could fur-ther the discovery process. But we also hope that the ex-plication and examples of the discovery mode presentedin this chapter will have broader utility in developmen-tal research.

To return to the task at hand, the proposed criteriahave more specific implications for the critical role inresearch design played by statistical analysis. First, inthe discovery phase, Type I errors can entail an evengreater risk than errors of Type II. To state the issuemore broadly, dismissing as invalid a finding that pointsthe way to a fuller and more precise explanation for thephenomenon under investigation may result in a greaterloss than that produced by accepting a finding that ishighly significant because of as yet undifferentiated andthereby confounded factors producing the phenomenonin question (e.g., the failure to distinguish Process fromContext). The greater risk in the discovery process of

dismissing findings as Type I errors is further com-pounded by the phenomenon of magnification of earlyenvironmental differences over time. Thus, as illus-trated by the escalating effects of proximal processesshown in Figure 14.2, changes in outcome associatedwith a proximal process at Time 1 can be quite small andnonsignificant statistically. Yet, as shown, they can bepowerful predictors of a marked increase in develop-mental outcome several years later (in the likely eventthat the process continued to be maintained over the in-tervening period).

At this point, a methodological note is in order. Sta-tistical models widely used for the purpose of hypothe-sis testing are often ill-suited as operational models fordevelopmental investigations in the discovery mode.This is particularly true for models that control statis-tically solely for linear relationships among the factorsin the research design to obtain an estimate of the inde-pendent contribution of each factor in the statisticalmodel to the outcome under investigation. The validityof such analyses rests on what in mathematical statis-tics is referred to as the assumption of homogeneity ofregression. To illustrate the assumption in its simplestgeneral case: given a dependent variable y and two in-dependent variables x1 and x2, then the relation be-tween x1 and y must be the same at all levels of x2. Thisassumption is often not met in developmental data. Forexample, when applied to the analysis shown in Figure14.2, it would require that the relation between proxi-mal process and frequency of problem behaviors be thesame at every social class level, which is not the case.Nor is this requirement likely to hold with respect toany combination of the four defining properties of thebioecological model. As Bronfenbrenner stated in his1979 monograph, In ecological research, the principalmain ef fects are likely to be interactions (p. 38, italicsin original).

Any research design based on a bioecological modelmust allow for the possibility of such interactions. How-ever, it is also essential, especially in the discoveryphase, that the particular interactions to be examined betheoretically based, and thatif possibletheir antici-pated direction and form be specified in advance so thatdiscrepancies between theoretical expectation and ob-served reality can be readily recognized and thus pro-vide the basis for a next step in the typically slow,iterative process of seeking more differentiated formu-lations that merit further exploration both on theoreticaland empirical grounds. In each case, the new formula-

From Theory to Research Design: Operationalizing the Bioecological Model 803

tion should be consistent with the existing theoreticalspecifications of the bioecological model, but it alsomust take into account any old or new research findingsbearing on the issue.

The foregoing criteria for research in the discoverymode do not imply neglect of the traditional issues of re-liability and validity. These are honored in a somewhatdifferent, theoretically guided way. Essentially, the pro-cess is one of cross-validation at two levels. First, in agiven study, the results at each successive stage of analy-sis are validated in the next, more differentiated formula-tion. Second, the generalizations emerging from a giveninvestigation are cross-validated against findings fromother studies of theoretically related phenomena but witha specific focus on the defining components of the bioe-cological model.

Before we proceed with concrete examples, it is im-portant to emphasize that the criteria we have proposedand applied for conducting developmental science in thediscovery mode represent a first attempt to construct aworking model. Moreover, the working model is subjectto the curious qualification that it is itself the product ofthe same sequential design that it proposes. The criteriawere developed by examining the changes introduced ateach successive stage in the evolution of the bioecologicalmodel to identify the theoretical and operational proper-ties leading to improvement in the models predictivepower. The example that follows illustrates these concur-rent processes.

Different Paths to Different Outcomes:Dysfunction versus Competence

In this instance, our exploratory effort took as its point ofdeparture the stipulation in Proposition II that the effectsof proximal processes vary systematically depending onthe developmental outcome. Once again, rather than tak-ing time to retrace our steps, we begin with where weended up; namely, with the following initial formulation:

The greater developmental impact of proximal processeson children growing up in disadvantaged or disorganizedenvironments is to be expected to occur mainly for out-comes ref lecting developmental dysfunction. By contrast,for outcomes indicating developmental competence, proxi-mal processes are posited as likely to have greater impactin more advantaged and stable environments.

The term dysfunction refers to the recurrent manifes-tation of difficulties on the part of the developing per-son in maintaining control and integration of behavior

across situations, whereas competence is defined as thedemonstrated acquisition and further development ofknowledge and skillswhether intellectual, physical,socioemotional, or a combination of them (e.g., learninghow to care for a young infant involves all three).

The preceding emergent formulation is based on thefollowing considerations. Most parents have the capac-ity and the motivation to respond to manifestationsof physical or psychological distress on the part oftheir children. In deprived or disorganized environ-ments, such manifestations of dysfunction have beenshown to be both more frequent and more severe (e.g.,in Drilliens research), thus drawing on more of par-ents available time and energy. Accordingly, to theextent that, in disadvantaged settings, parents are ableto engage in proximal processes, these are likely tohave greater impact in reducing dysfunction ratherthan in enhancing their childrens knowledge aboutand skill in dealing with the external environment.With respect to problems of dysfunction, in deprivedenvironments there is usually a match between youngchildrens needs and their parents capacity to meetthose needs. This does not mean, however, that chil-dren in such environments will end up functioning aswell as their age-mates growing up in more favorablecircumstance, but rather that, over similar periods oftime, they will show greater improvement in controlover their own problem behaviors as a function ofparental responsiveness.

The situation in advantaged and stable environmentsis rather different. Manifestations of dysfunction arelikely to occur less often and to be less intense. Underthese circumstances, parents are more apt to be at-tracted by and respond to the more frequent and moregratifying signs of their childrens growing competence,with the result that proximal processes may to be fo-cused mainly in this latter sphere. In addition, parentsliving in a middle-class world are themselves more aptto possess and exhibit the knowledge and skills theywish their children to acquire. They also have greateraccess to resources and opportunities outside the familythat can provide needed experiences for their children.Taken together, the foregoing considerations led to theformulation of the previously stated proto-hypothesis.

Because Drilliens study of the influence of mother-in-fant interaction dealt with only one developmental out-come, one has to look elsewhere for evidence that effectsof such processes vary depending on the nature of the out-come under consideration. A rich data archive generously

804 The Bioecological Model of Human Development

Figure 14.3 Effect of parental monitoring on grades inhigh school by family structure and mothers level of educa-tion. Analyses and graph based on data archives generouslyprovided by Professors Stephen A. Small (University of Wis-consin) and Tom Luster (Michigan State University).

Levels of Parental Monitoring Ranging from 012

20 4 6 8 10 1231 5 7 9 11

M 20 4 6 8 10 1231 5 7 9 11

M 20 4 6 8 10 1231 5 7 9 11

M2.9

3.9

4.9

5.9

6.9

Gra

de P

oint

Aver

age

(GPA

) Mother andFatherSingleParent Mother

Mother andStepfather

Mothers with some education beyond high schoolMothers with no education beyond high schoolGPA Scale: 2 = Mostly Ds or less 6 = Mostly Bs3 = 1/2 Cs, 1/2 Ds 7 = 1/2 As, 1/2 Bs4 = Mostly Cs 8 = Mostly As5 = 1/2 Bs, 1/2 Cs

Mean

MeanMean

into a closely corresponding research design can produce reli-able findings even when there are relatively few cases insome, or even all, the cells of the model. This comes about be-cause, in effect, the bioecological model requires, in its dis-covery phase, advance specification primarily not only ofmain effects but also in the form and direction of their mostplausible interactions in the light of both the evolving theoret-ical model and the then available research evidence. This isespecially true for well-designed experiments. For examples,see pp. 808809.

made available by Small and Luster (1990) from theirstatewide studies of youth at risk in Wisconsin met thisneed.2 Figure 14.3 depicts the results from an analysis ofthe differential effects of parental monitoring on the aca-demic achievement of high school students living in one ofthe three most common family structures found in thetotal sample of over 2,500 cases.3 The students were be-tween 14 and 16 years of age. It was also possible to strat-ify the sample by two levels of mothers education, withcompletion of high school as the dividing line. Parentalmonitoring refers to the effort by parents to keep in-

formed about and set limits on their childrens activitiesoutside the home. In the present study, it was assessed bya series of items in a questionnaire administered to ado-lescents in their school classes. All items referred to par-ents in the plural, with no distinction as to whether themother or the father was doing the monitoring. Levels ofparental monitoring, ranging from 0 to 12, are shown onthe horizontal axis, and grade point average (GPA) isshown on the vertical. The markers to the right of eachcurve record the mean GPA for each of the six groups.

Once again, the results reveal that the effects of prox-imal processes are more powerful than those of theenvironmental Contexts in which they occur. In this in-stance, however, the impact of the Process was greatestin what emerges as the most advantaged ecologicalnichefamilies with two biological parents in which themother had some education beyond high school. More-over, the developmental effect of the proximal process onschool gradesa measure of competencewas strongerfor families living in more advantaged socioeconomiccircumstances. This finding is directly opposite to thatrevealed by the analysis of Drilliens data, where the out-come was one of psychological dysfunction (i.e., the fre-quency of problem behaviors). At the same time, theprincipal finding from both studies documents the power-ful effect of proximal processes on human development,a result consistent with the first defining property of thebioecological model stipulated in Proposition I.

The reader may well ask why the data in each scatter-gram were fitted to a curve with a declining slope ratherthan simply with a straight line. In accord with the crite-ria for research in the discovery mode, the introduction ofthe quadratic term was based on theoretical considera-tions. Higher levels of academic performance requiremastery of more complex tasks, and hence are more diffi-cult to achieve. As a result, at each successive step, thesame degree of active effort would be expected to yield asomewhat smaller result. More specifically, for pupilswho are not doing so well in school, parental monitoring

2 The analyses of data from the Wisconsin archive reported inthis chapter were carried out in collaboration with Stephen A.Small (University of Wisconsin) and Tom Luster (MichiganState University) who designed and conducted the surveyfrom which the data are drawn. We are deeply indebted tothem for the theoretical thinking that underlies the construc-tion of the survey. It is an excellent example of developmentalscience in the discovery mode. We are also grateful to ReginaCannon (then a graduate student at Cornell University) whocarried out the statistical analyses with care and dispatch.3 The large number of cases in this study should not be takento imply that the bioecological model can be applied only insamples with a large N. As illustrated here, precision in theformulation of the theoretical model and in its translation

From Theory to Research Design: Operationalizing the Bioecological Model 805

can have a substantial effect by ensuring more stability ofTime and place so that some learning can occur. But forsuperior school achievement, students would require inaddition high levels of motivation, focused attention,prior knowledge, andespeciallyactually workingwith the material to be learned. These are all qualitiesthat stability of Time and place by itself cannot provide.

As can be seen in Figure 14.3, the relation betweenparental monitoring and school grades shows a curvilin-ear trend. Moreover, in accord with criteria for researchin the discovery mode (see pp. 801803), both in its di-rection and form the trend corresponds with theoreticalexpectations in being more pronounced when the motherhas some education beyond high school, especially in atwo-parent family structure. A test for heterogeneity ofregression confirms visual inspection. The differences inslopes between the two educational levels are highly sig-nificant (p .01), with the quadratic component emerg-ing as reliable only in the higher educational group.4 Alsostatistically significant are differences in school achieve-ment by family structure in each level of mothers educa-tion, with students growing up in two-parent familiesgetting the highest grades, and those from single-parentfamilies the lowest, a rank order corresponding to thepower of the proximal process in each group as measuredby the slopes of the associated regression coefficients.

Finally, a result not shown on the graph provides addi-tional evidence pointing to another tentative generaliza-tion. The first indication appeared in the analysis ofDrilliens data, which, among other findings, revealedthat maternal responsiveness had the general effect of de-creasing or buffering against environmental differences indevelopmental outcome. Thus, at high levels of mother-child interaction, social class differences in problem be-havior became smaller. A similar pattern emerges for theeffects of parental monitoring on school grades. Acrossthe six groups shown in Figure 14.3, stronger parentalmonitoring was associated not only with a higher meanon school performance but also with a lower standard de-viation. These differences, too, were statistically reliable.Hence the following working hypothesis:

For outcomes of competence, proximal processes not onlylead to higher levels of developmental functioning but also

serve to reduce and act as a buffer against effects of dis-advantaged and disruptive environments.

To turn from substance to method, the foregoingfindings also demonstrate that tests of significance havea place in research in the discovery mode, but, as withhypothesis verification, only after a specific theoreticalexpectation has been formulated in advance.

In a discovery context, however, the aim is not toclaim empirical validity for a particular theoretical for-mulation but to indicate its plausibility for inclusion inthe research design at subsequent stages of exploratorywork. To be sure, doing so may result in a failure ofreplication. But not doing so risks missing potentiallyimportant, theoretically guided research opportunitiesnot yet recognized. Garmezy and Rutter (1983), in theirlandmark studies of stress and coping in childrens development, did not differentiate between those protec-tive or disruptive forces emanating from the environ-ment, and those inherent in the biopsychologicalcharacteristics of the person. As evidenced from theanalysis of Drilliens data shown in Figure 14.1, thesevectors do not always operate in the same direction.Nevertheless, Garmezy and Rutters formulations andfindings played a significant role in the early stages ofthe process through which the bioecological modelreached its present, still-evolving form.

The still-evolving form imposes the obligation to takeadvantage of existing opportunities for continued explo-ration. With respect to the present inquiry, the next step inthat process was once again to pose the question about theextent to which the research design meets the definingproperties of the bioecological model. At first glance, weappear to be confronted with the same problem that we en-countered with Drilliens study. Proposition I definesproximal processes as bidirectional. As previously noted,Small and Luster (1990) defined parental monitoring asthe effort by parents to keep informed about and set limitson their childrens activities outside the home. As stated,such behavior implies influence from one side onlythatof the parents. An examination of the actual items used intheir questionnaire, however, revealed that they were oftwo kinds. Some were cast in the language of parental ex-pectation and prescription (e.g., If I am going to be homelate, I am expected to call my parent(s) to let them know;When I go out, my parent(s) ask me where Im going).By contrast, other items implied that the desired expecta-tions or prescriptions were being met (e.g., My parent(s)know where I am after school; I tell my parent(s) who

4 The degree of curvilinearity is measured by the correspon-ding regression coefficients and not by difference in thelength of each curve from top to bottom. The latter is deter-mined by empty cells in the scatter plot below or above whichentries for both monitoring level and GPA were available.

806 The Bioecological Model of Human Development

Figure 14.4 Parental monitoring and high school grades bygender: mothers with more than a high school education.Analyses and graph based on data archives generously pro-vided by Professors Stephen A. Small (University of Wiscon-sin) and Tom Luster (Michigan State University).

2-ParentFamily

MeanMean

2

3

4

5

6

7

Aver

age

Gra

de in

Hig

h Sc

hool

Sub

jects

BoysMean

Mean

Mean

Mother andStepfather

Single-ParentMother

Girls

20 4

6810

122

0 46

810

122

0 46

810

122

0 46

810

1220 4

6810

122

0 46

810

12Level of Parental Monitoring

Mean

Girls

BoysBoys Girls

Mean

Mean

High school students are between 14 and 16 years old.

5 Within each pair, both means and regression coefficientswere statistically significant, the latter confirming reliabledifferences in slope.

Im going to be with before I go out). Although the firsttype of item is unidirectional, the second entails some de-gree of reciprocity to the extent that the adolescent isproviding the information desired by the parents. Accord-ingly, we hypothesized that items of the second typewould show stronger relationships to developmental out-comes than those that described only the parents expecta-tions of how they wished their children to behave.

Separate analyses of scales based on each type ofitem provided substantial support for our working hy-pothesis. Although responses to both types of questionsshowed reliable effects on school performance, the rela-tionships for the reciprocity scale were significantlystronger and were much more likely to show curvilineareffects. Accordingly, the latter was the scale used in an-alyzing the results presented in Figure 14.3.

From the perspective of the biological model, the re-search design producing the results shown in that figure ismissing an important Person component. It is a generalfinding in educational research that at the high schoollevel female students score higher on measures of aca-demic performance than do males. The question thereforearises: To what extent is this gender difference attributa-ble to variations in proximal process? Figure 14.4 pro-vides a tentative answer to this question for studentswhose mothers had more than a high school education. Ineach family structure, parental monitoring exerted a morepowerful effect on the school achievement of girls than of

boys, a result that is paralleled by corresponding differ-ences in average GPA for the two sexes.5 In each of thethree family structures, girls received higher grades thanboys, with the difference being most pronounced in two-parent households and lowest in single-mother families.

As seen in Figure 14.4, however, a distinctive featureof the pattern for girls is a marked flatteningof the curve, especially for daughters of single-parentmothers. This result suggests that, in each of thethree family structures, mothers may be pushing theiralready successful daughters to the point where con-formity to maternal control no longer brings educationalreturns, particularly when the mother is the only parent.

An analysis of data on students whose mothers hadno more than a high school education showed a similargeneral pattern, but the effects were less pronounced.The influence of monitoring was appreciably weaker,and its greater benefit to girls was also reduced. Never-theless, girls with less-educated mothers both in single-parent and in stepfamilies still had higher GPA scoresthan boys. This means that some other factor not yetidentified must account for this difference.

Although a number of possibilities come to mind re-garding this unknown, regrettably the Wisconsin archivedoes not contain any data on the principal suspects. Whatis available is information about another trail of discoverythat we have already begun to explore. Our successivelymore differentiated working models, both conceptual andoperational, for assessing the effects of parental monitor-ing on school achievement have provided increasing sup-port for the tentative hypothesis that, for outcomesreflecting developmental competence, proximal processesare likely to have greatest impact in the most advantagedenvironments. But what about the other half of the originalformulation: the complementary postulate that the greaterdevelopmental impact of proximal processes growing upin poor environments is to be expected to occur mainly foroutcomes reflecting developmental dysfunction?

Data from Small and Lusters archive also providethe opportunity for cross-validating this provisionalclaim. In addition to measures of academic achievement,the Wisconsin study also included information onteenagers sexual activity. The decision to analyze thisoutcome in the context of a bioecological model wasprompted by Small and Lusters (1990) finding that such

From Theory to Research Design: Operationalizing the Bioecological Model 807

Figure 14.5 Effect of monitoring on girls sexual activity(high school students between 14 and 16 years of age).

21 3 4Mother CompletedHigh School Only

Mother Had Some EducationBeyond High School

21 3 4

20

40

60

80

Mean

Mean

Perc

ent S

exual

ly A

ctiv

e

0

Daughter with Own Mother and Father

Daughter with Mother and StepfatherDaughter with a Single-Parent Mother

6 We are also indebted to Kristen Jacobson, now a doctoralstudent at Pennsylvania State University, for her ingenuityand accuracy in translating into a common format dataarchives recorded on different computing systems.

behavior varied systematically by family structure. Sex-ual activity was measured by a single question: Haveyou ever had sexual relations with another person?

This documentation of variations in sexual activityby family structure takes on special significance in thelight of broader social changes taking place in the livesof children, youth, and families in contemporary U.S.society. Today, the United States has the highest rate ofteenage pregnancy of any developed nation, almosttwice as high as that of its nearest competitors (Bron-fenbrenner, McClelland, Wethington, Moen, & Ceci,1996, p. 117). Adolescent sexual activity is also one ofthe prominent elements in the so-called teenage syn-drome, an escalating pattern of co-occurring behaviorsincluding smoking, drinking, early and frequent sexualexperience, adolescent pregnancy, a cynical attitude to-ward education and work, and, in the more extremecases, drugs, suicide, vandalism, violence, and criminalacts (for references and successive summaries of the ev-idence, see Bronfenbrenner, 1970, 1975, 1986a, 1989c,1990, 1992; Bronfenbrenner et al., 1996; Bronfenbren-ner & Neville, 1994).

In anticipating the effects of parental monitoring onteenagers sexual activity, we were again confrontedwith the issue of the possible direction of influence. Inrelation to sexual activity as an outcome, however, someleverage for the resolution of the issue was provided be-cause each direction could be expected to produce oppo-site effects. On the one hand, if parental monitoringfunctions to defer sexual activity, then the more moni-toring the less sexual activity. On the other hand, if theparents begin to monitor only after the fact, the associa-tion would be reversed, with monitoring occurring in re-action to the adolescents behavior; hence, sexuallyactive adolescents would be monitored more.

The results of the analysis are shown in Figures 14.5and 14.6.6 The most salient finding for both sexes is thatparental monitoring does substantially reduce adoles-cents sexual activity. In many other respects, however,the patterns for female and male adolescents are quitedifferent. The results for girls in Figure 14.5 show thatthe effect of parental monitoring is stronger for daugh-ters of mothers with no education beyond high schoola finding consistent with the working hypothesis

that, for outcomes reflecting lack of control, proximalprocesses have greater impact in poorer environments.Tests for heterogeneity of regression confirm that thisfinding holds for each of the three family forms.

Yet, as shown in Figure 14.6, the corresponding analy-sis of the data for boys reveals the very opposite result.Parental monitoring has a more powerful effect on boyswhose mothers have had more education rather than less.Once again, the finding holds in each family structureand is confirmed by tests for heterogeneity of regression.

Figure 14.6 Effect of parental monitoring on boys sexualactivity (high school students between 14 and 16 years of age).

21 3 4 21 3 4

20

40

60

80

Perc

ent S

exual

ly A

ctiv

e

0

Mother CompletedHigh School Only

Mother Had Some EducationBeyond High School

Group %

Sons in Biological Families

Sons in StepfamiliesSons of Single-Parent Mother

Group %

808 The Bioecological Model of Human Development

This is not the only departure from the expecta-tions generated by the most recent working model. Forexample, there was not always correspondence betweenthe developmental power of proximal processes in agiven family structure and the percentage of sexuallyactive adolescents in that structure: In stepfamilies inwhich the mother has only a high school education, ma-ternal monitoring of daughters is as high as it is in two-parent families, but the percentage of sexually activegirls is even greater than that for single-parent mothersat the same educational level. The finding is consistentwith research indicating that living in a family with astepfather entails a special developmental risk for girls(Hetherington & Clingempeel, 1992).

And so, we find ourselves engaged in a next stage of thediscovery process in which we are seeking to develop amore differentiated formulation that, through a corre-sponding research design, will be most effective in reduc-ing the observed empirical departures from expectationsbased on the existing working model. The first step is toask an obvious question: What is most likely to accountfor such discrepancies? Restating the question from theperspective of the bioecological model, which of the fourcomponents is a likely suspect? It has to be somebody whois already on the scene. Parents are already there. Whoelse is around who could exert some influence on the sex-ual activity of high school students? The question answersitselfthe peer group. And if it is indeed true that proxi-mal processes are at least as powerful determinants of de-velopment as either the characteristics of the person or ofthe environment, what might that process be?

A tentative first nomination is progressively more in-tense interaction with peers who are already sexually ac-tive. Among other considerations, this suggestion is guidedby the possibility that peer pressure to engage in sexual ac-tivity and the prestige that such activity brings are likely tobe higher for boys from less educated families with the re-sult that parental monitoring is not as effective. With re-spect to the other components in the model, given thefindings just reported, gender would still be a Person char-acteristic of major importance. The choice of an appropri-ate environmental Context depends on the precise researchquestion being asked. Family structure would also still beappropriate. But from the perspective of the bioecologicalmodel, an option to consider would be the parents beliefsabout the activities they wanted their adolescent son ordaughter to engage in or refrain from, as well as the close-ness of the parent-child relationship.

We offer these suggestions not for their relevance tothis particular issue but to illustrate two additional emer-gent corollaries of the bioecological model:

1. The specific components of Process, Person, Con-text, and Time to be included in a given investigationshould be those that, from a theoretical perspective,are maximally relevant to the research question underinvestigation and complementary to each other in re-lation to the given developmental outcome.

2. From a theoretical perspective, the power of a PPCTdesign is most effectively enhanced by including morethan one proximal process in the model.

The next section leads to yet another corollary.

The Role of Experiments in theBioecological Model

The examples considered thus far are essentiallyexperiments of nature: They show how development isinfluenced by variations in the components of the bioe-cological model occurring in already existing societies.They tell us nothing about whether, to what extent, orhow these elements and their combinations can bechanged. This limitation applies particularly to the mostconsequential component of the bioecological modelproximal processes. We know of no research bearingspecifically on this question, but some indirect evidencedoes exist. In research findings already presented, im-proving the quality of the environment has been shown toincrease the developmental power of proximal processes.The indirect evidence comes from experiments in whichresearchers have systematically introduced conditionsinto the environment in which people lived that were hy-pothesized to enhance their psychological functioningbeyond existing levels.

Here are two examples at contrasting ages.

Environmental Dynamics in Old Age

The first example is Langer and Rodins oft-cited exper-imental intervention conducted with residents of a NewHaven nursing home for the aged (Langer & Rodin, 1976;Rodin & Langer, 1977). The contextual manipulationemployed in this study is well summarized in the au-thors words:

From Theory to Research Design: Operationalizing the Bioecological Model 809

The hospital administrator gave a talk to residents in theexperimental group emphasizing their responsibility forthemselves, whereas the communication to a second, com-parison group stressed the staff s responsibility for themas patients. To bolster the communication, residents in theexperimental group were offered plants to care for,whereas residents in the comparison group were givenplants that were watered by the staff. (Rodin & Langer,1977, p. 897)Residents were assigned at random to either the exper-

imental or the control group. Data on psychological andhealth characteristics were collected at three time points:(1) just prior to the introduction of the experiment; (2) 3weeks later, when the experiment was formally ended;and (3) in a follow-up study conducted 18 months later.

The substantial effects of intervention found at theend of the experiment (Langer & Rodin, 1976) were stillin evidence in the follow-up assessment. To be sure, be-cause the residents were almost a year-and-a-half older,the added age had taken some toll, but, nevertheless,those in the induced responsibility group not only sig-nificantly surpassed their controls, but were appreciablybetter off, both psychologically and physically, thanthey had been months earlier before the intervention hadbegun. In ratings by observers blind to the experimentalconditions, they were judged to be more alert, sociable,and vigorous. The most striking results were seen in thecomparison of death rates between the two treatmentgroups. Taking the 18 months prior to the original inter-vention as an arbitrary comparison period, in the subse-quent 18 months following the intervention, 15% in theresponsibility-induced group died, compared with30% in the control group.

Environmental Dynamics in InfancyA remarkable, independent cross-validation of Langer andRodins principal hypothesis appears in the findings ofanother intervention experimentthis one almost un-knownthat was carried out at about the same time with asample of 100 9-month-old infants and their mothers inthe Dutch city of Nijmegen (Riksen-Walraven, 1978). Al-though this author, Marianne Riksen-Walraven, appearsnot to have been aware of Langer and Rodins work con-ducted during the same period, one of the two interven-tion strategies she employed with her sample of infantswas similar to that used in the New Haven study of elderlypatients. Mothers, randomly assigned to what Riksen-Walraven called the responsiveness group, were given a

workbook for parents stressing the idea that the infantlearns most from the effects of its own behavior (p. 113):

Caregivers were advised not to direct the childs activitiestoo much, but to give the child opportunity to find outthings for himself, to praise him for his efforts, and to re-spond to his initiations of interaction. (p. 113)

By contrast, mothers of infants in the stimulationgroup received a workbook that emphasized the impor-tance of providing the infant with a great variety of per-ceptual experiences of all kinds, to point to and nameobjects and persons, and to speak a lot to their in-fants (p. 112).

In the follow-up assessment conducted 3 months later,infants of mothers who had been encouraged to be re-sponsive to their babies initiatives exhibited higher lev-els of exploratory behavior and were more likely to prefera novel object to one that was already familiar. The babiesalso learned more quickly in a learning contingency task.

Neither of the preceding investigations included any systematic assessment of the activities in which theparticipants in the experiment subsequently engaged, ofthe balance between unidirectional and bidirectional be-havior in the two groups, or of any other specific featurethat could provide a measure of the extent to which prox-imal processes were operating in each of the two con-trasting experimental conditions.

In both of the preceding experimental studies, elegant as they are, the keystone of the bioecologicalmodela measure of proximal processwas not in-cluded in the research design. In addition, the demonstra-tion (in Figure 14.3) of the joint, indirect effects offamily structure and parents education on the relation ofproximal processes to school grades does only half thejob, for it provides no information on whether differencesin students personal characteristics (such as gender)exert a similar indirect effect. Nevertheless, viewed fromthe theoretical perspective of a bioecological model, allthese findings are impressively consistent with expecta-tions derived from the model; the findings illustrate themodels practicability, andperhaps most promising forthe future of developmental sciencegenerate questionsthat, when answered, provide ways for enhancing themodels scientific power. It is these questions and an-swers that are addressed in the following sections.

Up to this point, our exposition has given primary at-tention to the core concept of proximal process and its

810 The Bioecological Model of Human Development

key position in the bioecological model as a whole. Wenow proceed to a more detailed examination of each ofthe other three defining properties of the modelPer-son, Context, and Time.

HOW DO PERSON CHARACTERISTICSINFLUENCE LATER DEVELOPMENT?

As already indicated, at midstage in the development ofthe bioecological model, an effort was begun to arrive atsome answers to this question, and it has continued up tothe present day. As before, rather than describe the suc-cessive stages in this emergent reconception, we presentit in its most recent, still-evolving form.

Most developmental research treats the cognitive andsocioemotional characteristics of the person as dependentvariables; that is, as measures of developmental outcomes.Far less often are such characteristics examined as pre-cursors and producers of later development From the per-spective of the bioecological model, their effectiveness inthe latter role derives from their capacity to influence theemergence and operation of proximal processes.

Accordingly, in an effort to identify such process-rel-evant Person characteristics, we applied the sequentialdesign strategy described in the preceding section. Be-ginning with implications derived from the theoreticalmodel, which are then related to existing research find-ings, successive applications of this strategy have re-sulted in the conceptualization of three kinds ofprocess-relevant Person characteristics, which, for con-venience of brevity, we have labeled as Person forces, re-sources, and demands.7

Force Characteristics as Shapersof Development

In the bioecological model, the characteristics of thePerson most likely to influence future developmentwould be active behavioral dispositions that can setproximal processes in motion and sustain their opera-tion, orconverselyactively interfere with, retard, oreven prevent their occurrence. It is therefore useful to

distinguish between these two propensities. We refer tothe former as developmentally generative characteristics;to the latter as developmentally disruptive.

Examples of developmentally disruptive dispositionscome more readily to mind. At one pole, they includesuch characteristics as impulsiveness, explosiveness, dis-tractibility, inability to defer gratification, or, in a moreextreme form, ready resort to aggression and violence; inshort, difficulties in maintaining control over emotionsand behavior. At the opposite pole are such Person attri-butes as apathy, inattentiveness, unresponsiveness, lackof interest in the surroundings, feelings of insecurity,shyness, or a general tendency to avoid or withdraw fromactivity.8 Persons exhibiting either of the precedingpropensities would find it difficult to engage in proximalprocesses requiring progressively more complex patternsof reciprocal interaction over extended periods of time.

By contrast, developmentally generative characteris-tics involve such active orientations as curiosity, tendencyto initiate and engage in activity alone or with others, re-sponsiveness to initiatives by others, and readiness todefer immediate gratification to pursue long-term goals.

We have found few investigations that shed light on thedevelopmental effects of either type of dynamic character-istics on proximal processes and their outcomes. A majorreason for this shortcoming is the absence of theoreticalconstructs for conceptualizing their changing nature overthe course of development from early infancy, throughadolescence, into and beyond early adulthood. The follow-ing framework is offered as an initial basis for meeting thisrequirement beginning in the Person domain in greaterneed of conceptual definitionthat of developmentallygenerative characteristics. The corresponding structurefor developmentally disruptive Person qualities can thenbe derived as an inverted mirror image of the former.9

Developmentally Generative Dispositions inLife-Course PerspectiveThe first and earliest manifestation of generative dispo-sitions takes the form of what we call selective respon-siveness. It involves differentiated response to, attractionby, and exploration of aspects of the physical and socialenvironment.

7 As is documented later in this chapter (p. 819), the recentlyrenewed, and far stronger, claims by behavior geneticists forthe predominant role of genetic factors in determining bothindividual and group differences in all forms of human char-acteristics are directly challenged by alternative explanationsand research findings derived from the bioecological model.

8 Depending on the available alternatives, withdrawal may bethe only course left for dealing with an unbearable situation.9 The material that follows represents a further developmentby the present authors of ideas first introduced in Bronfen-brenner (1989).

How Do Person Characteristics Inf luence Later Development? 811

The next generative characteristic to evolve goes be-yond selective responsiveness to include the tendency toengage and persist in progressively more complex activ-ities; for example, to elaborate, restructure, and even tocreate new features in our environmentnot only physi-cal and social but also symbolic. We refer to propensi-ties of this kind as structuring proclivities.

The transition from one to the other of these dynamicforms of orientation during early childhood is illustratedin successive publications from a longitudinal study of in-fants being carried out by Leila Beckwith, Sarale Cohen,Claire Kopp, and Arthur Parmelee at UCLA (Beckwith &Cohen, 1984; Beckwith, Rodning, & Cohen, 1992; Cohen& Beckwith, 1979; Cohen, Beckwith, & Parmelee, 1978;Cohen & Parmelee, 1983; Cohen, Parmelee, Beckwith, &Sigman, 1986). Their imaginative and careful work re-veals a progressive sequence of such environmentally ori-ented dispositions from birth through 7 years of age.Thus, immediately after birth, infants are especially re-sponsive to vestibular stimulation (being picked up andheld in a vertical position close to the body), which hasthe effect of soothing babies so that they begin to engagein mutual gazing; by 3 months, visual exploration extendsbeyond proximal objects, and the mothers voice is mostlikely to elicit responses especially in the form of recipro-cal vocalizations.

From about 6 months on, the infant begins actively tomanipulate objects spontaneously in a purposeful wayand to rearrange the physical environment. By now, bothvocalization and gesture are being used to attract theparents attention and to influence their behavior. In ad-dition, there is a growing readiness, across modalities,to initiate and sustain reciprocal interaction with awidening circle of persons in the childs immediate en-vironment. This is the emergence of what we call struc-turing proclivities.

A number of other investigations have yielded com-parable findings, and have extended them to still otheractivity domains; for example: individual differences inchildrens creativity in play and fantasy behavior (Con-nolly & Doyle, 1984; MacDonald & Parke, 1984) orJean and Jack Blocks longitudinal studies of ego re-s