Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
351 The deliberations and decisions by the committee about which casesto select for intensive study and how to structure the collection andanalysis of the data were undertaken in a pragmatic manner by a working group of researchers confronting an unusual and important task under a tight deadline. The workings and outcomes of that process are described in the report. This appendix connects that pragmatic process to formal methodological writings dealing with case studies and with the study of human behavior, including deviant and psychopathological be- havior, in ecological context. The nature of the phenomena to be studied necessitated a compara- tive case study approach by virtue of both the extreme rarity of the phe- nomena, in terms of numbers of occurrences, as well as the extreme sever- ity of the behavior involved, whether that behavior is seen on any of a number of possible continua, including general violence, youth violence, or school violence. The small number of cases precluded gathering a large sample, while the severity of the violence seemed to demand in- depth scrutiny of such extraordinary events. While single case studies and comparative studies of small numbers of cases are not suited to the same ends as more conventional statistical analyses of large samples, this approach is well suited to situations like the present one. The underlying assumptions and appropriate ends of case study approaches have been laid out with increasing sophistication in recent years, particularly in such fields as historical sociology and com- munity studies, in which dealing with small numbers of cases is the usual Appendix A Case Study Methodology and the Study of Rare Events of Extreme Youth Violence: A Multilevel Framework for Discovery Mercer L. Sullivan and Mindy Thompson Fullilove
352 APPENDIX A state of affairs. This appendix reviews briefly some of those assumptions and that methodological literature as they relate to the current investiga- tion of extreme violence in American schools. Following that examination, the appendix then connects the aims and strengths of case study methodology to the study of extreme, and there- fore puzzling, human behavior. The same extremity of deviance that makes such cases rare demands a multilevel conceptualization of the prob- lem to be investigated. The emphasis on discovery of relevant factors through naturalistic methods of inquiry, which is the hallmark of case studies, leads directly to the conceptualization of human behavior as tak- ing place in a set of hierarchically nested systems that underlies several well-established schools of thought and research methodology in psy- chology and psychiatry. It is precisely when things occur that seem ex- treme, unprecedented, and difficult to explain that one needs to cast the widest possible net to identify potentially important factors and processes. Viewing human behavior as occurring in a set of hierarchically nested systems is of great heuristic value in this situation, certainly as a way of casting a wide net and potentially as a way of actually catching something in an uncharted sea. APPROPRIATE USES OF CASE STUDY METHODS IN SCIENTIFIC RESEARCH Case studies have a variety of practical uses, including widespread application as teaching texts in such fields as law and business. In recent years, however, the particular and appropriate uses of cases studies as tools for scientific research have been delineated. Yin (1989) has noted that case studies, far from being suitable merely as illustrative material for quantitative analyses using larger samples, can be employed for three distinct and legitimate types of scientific ends: descriptive, exploratory, and explanatory. As an example of how even a single case study can serve explanatory ends, he cites Allisonâs study of the Cuban missile cri- sis, in which the author tests three rival hypotheses against the data (Allison, 1971). In introducing Yinâs exposition of case methods, Donald Campbell, among the most distinguished of all social science methodologists, con- trasts this kind of operation with the highly touted ability of randomized experiments to exclude all rival hypotheses, without, he notes with a large dose of irony, âspecifying what any of them areâ (Campbell, 1989:8). Campbell invokes the effort to exclude rival hypotheses as the âcore of scientific methodâ (p. 7). In the case of extreme school violence, for ex- ample, there may be popular theories, such as âthe perpetrators are all the victims of bullyingâ that might be excluded on the basis of even a single
353CASE STUDY METHODOLOGY case and that look even more untenable if they are inapplicable to mul- tiple cases from a small universe. Yin (1989:23) offers a working formal definition of a case study as: âan empirical inquiry that: â¢ investigates an empirical phenomenon within its real-life context; when â¢ the boundaries between phenomenon and context are not clearly evident; and in which â¢ multiple sources of evidence are used.â The second of these criteria is of particular interest in the present circumstance. In cases of extreme school violence, the question of the relevant context is both urgent and extremely problematic. Is it the psy- chopathology of the individual who offends? The borders of the school (and are those merely physical or to be thought of as the social field of school-related activities)? The community in which it occurs (and is that a local neighborhood or an incorporated municipality or county)? The national culture, drenched in violent media and grounded in a history of violence unique among wealthy, âcivilizedâ nations (or subnational cul- tural traditions, defined by race, ethnicity or region, that tend more to legitimize violence)? With so few cases and the frighteningly suggestive but statistically unverifiable appearance of a recent upward trend, we cannot rule out any of these possibilities at the beginning of an inquiry. The inquiry at hand, then, is a textbook example of a research situa- tion in which case studies are not merely a default position but the proper tool for the job at hand. The boundaries between these phenomena and their contexts are about as murky as they get. The exploratory functions of the case study method are at the forefront here, and a central task is to probe for the relevant contexts and their interrelationships in these rare and extreme occurrences. Many well-trained social scientists may be tempted at this point to raise two objections. First, how can even a meticulously researched case study prove the relevance of a factor or process to the occurrence of an outcome? Second, even if some such relationships can be established in a given case, how does one know that they have anything relevant to any other case? Both points, appropriate as they are in dealing with many common research situations, are of limited relevance here. A case study may not prove beyond all doubt that a certain relation- ship exists in that particular case, but it is very likely the most trustworthy way to go about it, absent the ability to isolate the relationship in a labora- tory, outside and perhaps irrelevant to the real world in which something
354 APPENDIX A occurs. The use of multiple sources of evidence, as opposed to strictly predefined data elements, helps to ensure construct and context validity and provides checks and balances that protect the inquiry from contami- nation by methodological artifacts related to reliance on single sources of evidence. One of the most common failures in attempting to traverse the peril- ous gulf between correlation and causation is that of ignoring a previous causal factor that explains the association of interest. Because case studies treat the boundary between phenomenon and context as inherently prob- lematic and consult multiple sources of evidence in order to discover as many potentially important factors as possible, the danger of ignoring important factors in this way is reduced. Inferring relationships in a single case from aggregate patterns in a large number of cases sharing some but not all the features of the case in point is not inherently superior. Inference downward from the aggregate to the single case is not illegitimate either, but it runs the well-known risk of ecological fallacy (Robinson, 1950). The danger of generalizing inappropri- ately from the aggregate to the individual is no less than that of moving in the other direction (Sullivan, 1998). And yet, even a single well-docu- mented case can refute absolutely a potentially overly broad theory such as âthey are all victims of bullying.â The second objection, that a single case cannot be generalized to oth- ers without additional scientific operations, is not incorrect. Rather, it misses the point in a situation in which there are as yet few viable hypoth- eses to generalize. Here we return to Campbellâs assertion of the central- ity to the scientific method of examining rival hypotheses. Case studies are excellent tools for pruning extraneous hypotheses and generating po- tentially viable ones. Objecting to case studies in the name of generaliza- tion and science ignores the variety of important functions they play in scientific inquiry along with the complexity of the process of scientific generalization. CASE STUDY, NATURALISTIC INQUIRY, AND SYSTEMATIC DISCOVERY Case studies are inherently naturalistic, conducted by means of gath- ering data from within the naturally existing social fields in which the phenomena of interest are located (Lincoln and Gruba, 1985). This natu- ralistic perspective is well suited to the systematic discovery of both im- portant factors, âvariablesâ in other terminology, as well as the processes that connect these factors, or varying conditions, in the unfolding of social action (Abbott, 1992). Considering the potentially infinite number of fac- tors that might be related to any given social phenomenon, a systematic
355CASE STUDY METHODOLOGY approach to discovery of a more limited set of potentially important fac- tors is essential (Glaser and Strauss, 1967). Naturalistic field methods allow researchers a practical method of casting nets wide enough to minimize the chance of missing important contributing factors and yet focused enough to make it possible to do useful and timely work. This process unfolds in practice through an iterative process of conceptualization and data-gathering in which re- search questions and operations are continually refined as initial data analyses suggest subsequent data-gathering, ideally until the data them- selves appear to reach a point of saturation, at which point subsequent data add very little to the overall picture (Strauss and Corbin, 1998). The naturalistic approach allows researchers to test any theoretical notions they have brought to the field against the perceptions of social actors in that field as well as to invite the social actors to nominate other potentially important factors. This systematic approach to discovery, while it does not eliminate the possibility that researchers may ignore important fac- tors or impose biases as a result of preexisting theoretical predispositions, helps to minimize those risks. While the working group of researchers on this project engaged in a minimum of discussion of these fundamental aspects of case study re- search, it is probably fair to say they shared many of these methodological assumptions as a result of past research experience. Prior to entering the field to collect data, they did engage in a concerted collective effort to identify a common set of theoretical categories for which data would be sought. For the most part, this effort was successful. (See the introduc- tion to Part I for a list of the categories used. The work of Katherine Newman and her research team in generating the initial set of categories was particularly helpful for the entire working group.) A notable aspect of the working groupâs ongoing effort to identify and gather data for a common set of categories across the cases is the hierarchical, nested structure of the categories, including community- level, situational, and individual-level factors (see Figure A-1). The emer- gence of this hierarchical, nested structure was the outcome of a prag- matic group process of defining the task at hand. It was not the result of a preexisting set of theoretical orientations shared among a group of re- searchers from diverse disciplinary backgrounds, including sociology, anthropology, education, and psychiatry. Yet there is ample precedent for this kind of theoretical structure in a number of fields, including the study of violence. This appendix next discusses some of the intellectual roots of this approach, not because the particular scholarly traditions to be discussed were explicitly invoked during the process, but because the theories to be discussed help make explicit key issues that this enterprise
356 APPENDIX A had to confront in trying to gather and analyze data about these rare and extreme cases of violence. HIERARCHICALLY NESTED SYSTEMS AND THE STUDY OF VIOLENCE Previous syntheses of the research literature on violence have consis- tently suggested a hierarchically nested explanatory framework. For ex- ample, a comprehensive review of the literature commissioned by the National Research Council a few years ago was organized around the nested levels of community, situation, and individual (National Research Council, 1993). Violence clearly varies in frequency and type in ways correlated with each of these levels. FIGURE A-1 Two hierarchical models. Two Hierarchical Models Engel Bronfenbrenner Macro-System Exo-System Community Meso-System Family Micro-System Two-Person Person Person Nervous System Organ/Organ System Tissue Cell Molecule
357CASE STUDY METHODOLOGY A great deal of recent research in the human sciences has also em- ployed this kind of framework, guided by explicit theoretical orientations that focus attention on processes of person-environment interaction as underlying human behavior and development (Sampson and Lauritsen, 1990). The brief discussion of these traditions presented here delineates some of the key assumptions and discusses their heuristic value in guid- ing the present inquiry. The conceptualization of hierarchically nested systems has under- girded important scientific breakthroughs over the past 50 years in a variety of fields, including engineering, computer science, administra- tion, as well as developmental psychology and psychiatry (March and Simon, 1993; Simon, 1996). Basic principles of systems theory have be- come invaluable for understanding and designing a wide range of sys- tems, both mechanical and social. Among these principles are such ideas as the following: â¢ Each level of a hierarchically nested system operates according to principles unique to that level. â¢ Lower levels are constrained by higher levels. â¢ Levels otherwise operate independently, unless occurrences at one level upset operations at another level. In Herbert Simonâs (1996) well-known illustration of these principles, for example, it does not matter for the navigation of a ship whether the compass used is mechanical or electronic, so long as the compass supplies sufficiently accurate results. The internal workings of the two types of compass, in contrast, are completely dissimilar. At a certain point, how- ever, a demand for greater accuracy at the level of the ship, as a result of more precise steering systems, may demand greater accuracy than the me- chanical compass can provide. At this point, the internal workings of the compass may need to be changed completely, along with the articulation of operations at the level of the compass with those at the level of the ship (Simon, 1996). Two of the case study authors, the authors also of this appendix, independently found it useful to make use of hierarchical theories of this kind in order to analyze the data gathered in their case studies. One of these theories is drawn from clinical psychiatry (Engel, 1980), the other from developmental psychology (Bronfenbrenner, 1979, 1992). The two theories share some common assumptions about how hu- man behavior unfolds within a set of hierarchically nested ecological lev- els. Neither is a prescriptive theory; both were designed as guides for practice, clinical treatment in the case of Engelâs (1980) work and psycho- logical research in that of Bronfenbrenner (1979, 1992). The main overlap-
358 APPENDIX A ping points of these theories are discussed here, not because they were explicitly employed across the case studiesâwhich they were notâbut because they highlight some of the intellectual assumptions reflected in the template of factors arrived at by the working group as a guide for collecting and analyzing data across the case studies. Figure A-1 is an interpretive comparative representation of the theo- ries of Engel and Bronfenbrenner. The left-hand column lists the hierar- chical levels delineated by Engel, from molecules at the bottom to com- munity at the top. The right-hand column lists Bronfenbrennerâs levels, from person at the bottom to what he calls âmacro-systemâ at the top. In the middle is the pivot of interpretive comparison, showing âpersonâ as the common element of the two theories. As the figure illustrates, Bronfenbrennerâs theory is purely social and psychological and does not deal with biological levels of organization underneath the level of the person. Engelâs theory is, as he names it, biopsychosocial theory. Engelâs theory was developed for clinical application and is widely used in the training of psychiatrists and other medical practitioners. The theory emphasizes the embeddedness of biological functioning in social context and the continuous interaction of psychological and organic func- tions at the level of the person. Engel notably emphasizes both that one knows enough to know that this embeddedness exists, while, at the same time, one knows that one does not know very much about how it works. His theory is therefore fully grounded in empirical science and in no way a misty invocation of alternative medical systems based on anything other than empirical scientific premises. He therefore stresses the need to con- sider all potential factors at various levels of hierarchical embeddedness (within the practical limits of time and available information) when deal- ing with a case of medical need. As is also true of the case studies under- taken here, the great stimulus for casting as wide a net as is practical under the given circumstances is the perplexing acknowledgment that one does not know what it is that one needs to know. Bronfenbrennerâs work has been dedicated to expanding knowledge of the interaction of person and environment in the process of human development. Building on the foundational work in this area by Vygotsky (1978) and Lewin (1935), especially Lewinâs fundamental theorem that development is a function of the interaction of person and environment, Bronfenbrenner proposes a hierarchical conceptualization of the social environment, divided into the formally defined levels listed in the right- hand column of Figure A-1. This appendix is not the place to rehearse Bronfenbrennerâs schema in detail, but, briefly, his conception of micro- system is of the immediate environment of the developing child. The meso-system is the set of relationships of the different social systems directly experienced by the child; the exo-system is the larger community
359CASE STUDY METHODOLOGY embracing meso-system components; and the macro-system is the most embracing system of all, that of the national culture in which communi- ties are nested. Figure A-1 presents an interpretive comparison of Engelâs and Bron- fenbrennerâs models. Bronfenbrennerâs micro-system is staggered in between Engelâs two-person and family levels, to indicate that Bron- fenbrennerâs category of micro-system probably embraces the more fine- grained distinction between dyadic and familial relationships made by Engel. Similarly, Engelâs community level seems to include both the meso- and exo-system levels of Bronfenbrenner. Finally, Engel stops at the level of community, while Bronfenbrenner goes up one additional level, to that of national culture. Discussing what of these systems is more correct would be futile, both for present purposes and almost certainly from the perspectives of their originators. They were always intended for heuristic rather than prescriptive purposes, and they are remarkably similar in theoretical as- sumptions, structure, and intent. They share two fundamental assump- tions: the embeddedness of behavior in the social environment and the hierarchically nested structure of that social environment. They are frame- works for solving problems by gathering appropriate data in an efficient manner, acknowledging the challenge of the unknown, refusing to accept inappropriate limits of possibly important information, and setting prob- lems in their real-world contexts. For all these reasons, these frameworks inform this set of case studies in an appropriate and useful manner. It is hardly surprising that the theoretical premises of this work are highly congruent with the research template that emerged from the work of the committee and the case study researchers to coordinate collection and analysis of data. The practical needs of the research group were shaped by the extraordinary nature of the phenomena to be studied, the lack of an extensive base of previous research on these phenomena, and the problematic nature of the bound- aries between phenomena and their contexts. The research design that emerged was one that approached the problem as one involving multiple levels of context requiring field inquiry that could investigate different levels of context and their interrelationships in the emergence of these events of extraordinary violence. The fact that these are case studies on highly unusual phenomena makes this kind of multilevel, comparative case study approach particu- larly appropriate. Both case study methodology and multilevel ecologi- cal frameworks are potentially useful for a wide range of research prob- lems. Bronfenbrennerâs theory, for example, was designed for the study of normal child development, across societies and developmental stages, while case study methods have long been used to study both ordinary
360 APPENDIX A and extraordinary phenomena (Hamel, 1993; Rabin and Becker, 1992; Yin, 1989). The lack of knowledge about these kinds of rare events, however, makes the multilevel, case study approach particularly useful. Systematic, multilevel, comparative case studies, in this instance, are far more than simply a fallback methodological option, undertaken only because of the frustrating lack (from a scientific point of view) of a suffi- ciently large number of cases for statistical analysis. The challenge here is more daunting than that. Not only do we lack a large-N database; we also do not know what kinds of questions we should be asking when con- fronted with such rare, puzzling, and ostensibly irrational behavior. Given this need to define questions, even if, horrifically, we suddenly had large numbers of cases, it would be prudent to undertake these kinds of case studies at this stage of inquiry. WHAT KINDS OF CONCLUSIONS ARE APPROPRIATE? Given the methodological issues discussed to this point, what kinds of conclusions would it be reasonable to expect might be drawn from a group of case studies such as those found here? No attempt is made here to frame those conclusions. Rather the question here is what kinds of conclusions might be appropriate, whether from this volume itself, on the part of readers of the case studies, or in terms of directions for future work in this area. First, at least in the opinion of the authors of this appendix, expecta- tions should not be confined solely to the framing of hypotheses. As discussed earlier, case studies as instruments of scientific research can be descriptive, exploratory, or explanatory. The descriptive and exploratory functions should certainly come to pass, but the explanatory is not out of the question. The most likely form of that would probably be negative case refutation of overly broad or simplistic generalizations, such as âthey were all victims of bullying.â Only one case is needed for that. Beyond that, the small universe of these extremely serious incidents presents the possibility that finding one or more negative cases might cast serious doubt on a number of potential generalizations of interest. Nega- tive explanatory findings can be highly useful in times of sudden and emotionally charged attention to a problem, during which many explana- tions are likely to be and should be tried out. Unfortunately, some theo- ries can become highly popular in the absence of systematic testing of any kind. The range of data assembled for these case studies is unique at the present time, making these studies a distinctive and important resource. For these reasons, the potential explanatory use of these studies should not be ruled out.
361CASE STUDY METHODOLOGY Negative case analysis also plays a central role in exploratory, hy- pothesis-generating activity of case study analysis and of qualitative re- search generally (Lindesmith, 1968; Becker, 1998). The process involves the iterative trying out of hypotheses on data, followed by successive refinements of the hypotheses in an effort to make them fit more and more cases. These cases should provide ample grist for that kind of effort, both here and in the future. The multilevel structure of the data gathered across these cases pre- sents unique opportunities for generating hypotheses, for it invites think- ing and perhaps future research about the interactions across levels. Are some ecological levels more salient than others, for example? Do condi- tions at one level seem to constrain processes at a lower level, as systems theory would suggest? If so, how does that seem to happen? These kinds of questions may appear overly ambitious, and perhaps they are. If the kind of theory-building engaged in is the kind that aims to establish necessary and causal conditions for the occurrence of a certain kind of phenomenon, then asking these questions of these data is indeed inappropriately grandiose. Recent writings on case study methodology, however, have argued that the search for causality is not the only possible or useful social scien- tific enterprise. If one presses hard on the conditions needed to establish causality, the number of truly successful causal demonstrations in social science diminishes very rapidly. Becker has contrasted the search for understanding âthe real complexity of historical casesâ with the effort to find ârelationships between variables in a universe of hypothetical cases.â He calls the relationships established in this kind of enterprise âconjunc- turesâ rather than âcausesâ and advocates the search for such conjunc- tures as an eminently worthwhile scientific goal (Becker, 1992:208). In a similar vein, Ragin has noted that finding a relationship in a particular case is interesting and important, even if other possible com- binations of antecedents or independent variables might have produced the same outcome (Ragin and Becker, 1992). Such findings, which take on a triangular form in bivariate scatterplots, have often been dismissed in conventional social science thinking as uninterpretable, because they do not fit the expectations of linear causal modeling. But finding one kind of relationship, albeit not the only possible one, surely represents an advance over not having found any kind of relationship, or over having found every possible kind of relationship, which amounts to the same thing. Extraordinary acts of violence in schools cannot be ignored just be- cause they are too few in number to be subjected to rigorous statistical analysis. Theories about how they happen have arisen and will continue
362 APPENDIX A to do so. Establishing causal processes leading to these rare and heinous outcomes is not the only scientific approach possible in the service of the search for prevention and control. Case studies like those presented here are essential, appropriate, and scientific tools for use in this search. Only by carefully analyzing those patterns that do exist in the unfolding of these occurrences, using the full range of data available, can we make headway. REFERENCES Abbott, A. 1992 âWhat do cases do? Some notes on activity in sociological analyses.â Pp. 53â82 in What Is a Case? Exploring the Foundations of Social Inquiry, C.C. Ragin and H.S. Becker, eds. Cambridge: Cambridge University Press. Allison, G.T. 1971 Essence of Decision-Making: Explaining the Cuban Missile Crisis. Boston: Little, Brown. Becker, H.S. 1992 âCauses, conjunctures, stories, and imageryâ in What Is a Case? Exploring the Foun- dations of Social Inquiry, C.C. Ragin and H. S. Becker, eds. Cambridge: Cambridge University Press. Becker, H.S. 1998 Tricks of the Trade: How to Think About Your Research While You Are Doing It. Chicago: University of Chicago Press. Bronfenbrenner, U. 1979 The Ecology of Human Development. Cambridge, MA: Harvard University Press. Bronfenbrenner, U. 1992 âEcological Systems Theory.â in Six Theories of Child Development, R. Vasta, ed. Philadelphia: Jessica Kingsley Publishers. Campbell, D.T. 1989 âForeword.â in Case Study Research: Design and Methods, Robert K. Yin, ed., Newbury Park, CA: Sage Publications. Engel, G.L. 1980 âThe Clinical Application of the Biopsychosocial Model.â American Journal of Psychiatry 137:535â544. Glaser, B.G., and A. Strauss 1967 The Discovery of Grounded Theory. Chicago: Aldine. Hamel, J. 1993 Case Study Methods. Newbury Park, CA: Sage Publications. Lewin, K. 1935 A Dynamic Theory of Personality. New York: McGraw-Hill. Lincoln, Y.S. and E. G. Guba 1985 Naturalistic Inquiry. Newbury Park: Sage Publications. Lindesmith, A.R. 1968 (1947) Method and Problem. in Addiction and Opiates, A. R. Lindesmith, ed. Chi- cago: Aldine. March, J.G., and H.A. Simon 1993 Organizations. Cambridge, MA: Blackwell.
363CASE STUDY METHODOLOGY Ragin, C.C., and H.S. Becker (Eds.) 1992 What Is a Case? Exploring the Foundations of Social Inquiry. Cambridge: Cambridge University Press. Robinson, W.S. 1950 Ecological Correlation and Behavior of Individuals. American Sociological Review 15:351â357. Sampson, R.J., and J.L. Lauritsen 1990 Violent Victimization and Offending: Individual-, Situational-, and Community- Level Risk Factors.â Pp. 1â115 in Understanding and Preventing Violence: Social Influences, A.J. Reiss and J.A. Roth, eds. Washington, DC.: National Academy Press. Simon, H.A. 1996 The Sciences of the Artificial. Cambridge, MA: MIT Press. Strauss, A., and J. Corbin 1998 Basics of Qualitative Research. Thousand Oaks, CA: Sage Publications. Sullivan, M.L. 1998 Integrating Qualitative and Quantitative Methods in the Study of Developmental Psychopathology in Context. Development and Psychopathology 10:377â393. Vygotsky, L.S. 1978 Mind in Society: The Development of Higher Psychological Processes. Cambridge, MA: Harvard University Press. Yin, R.K. 1989 Case Study Research: Design and Methods. Newbury Park: Sage Publications.