Roth y Metha 2002

Positivist and interpretivist analytical approaches are frequently believed to be incom-
patible as research strategies and ways of understanding the world. This article argues
that not only may versions of positivism and interpretivism be combined in the analysis
of contested events, but this combination can further the goals of both approaches by
contributing information that may have been missed by adopting only one perspective.
The authors illustrate this using two case studies of lethal school shootings near Pad-
ucah, Kentucky, and Jonesboro, Arkansas, and introduce methodological strategies to
manage potential biases that may lead to contradictory testimony. However, these same
contradictions act as distinct data points from the interpretivist perspective, offering
insight into the cultural understandings of a community. The authors develop new forms
of triangulation that are tailored to these research goals and illustrate how, just as
positivist analysis may be used to aid interpretivism, an interpretive understanding of
a community may be necessary to develop causal theories of contested events such as
school shootings.
The Rashomon Effect

Combining Positivist and Interpretivist Approaches
in the Analysis of Contested Events
WENDY D. ROTH and JAL D. MEHTA

Harvard University
INTRODUCTION
Akira Kurosawa’s 1950 film Rashomon presents four different accounts

of a contested event—the murder of a Japanese nobleman and the rape
of his wife. As the events are retold from four different points of view,
the viewer is left wondering which of the four witnesses was telling
AUTHORS’ NOTE: This research was funded by grants from the National Academy
of Sciences and the William T. Grant Foundation (Katherine Newman, principal inves-
tigator), by a National Science Foundation Graduate Research Fellowship, and by a
National Science Foundation Integrative Graduate Education and Research Traineeship
Grant (98070661). We thank Katherine Newman, David Harding, Cybelle Fox, and two
anonymous reviewers for helpful discussions and comments. The authors can be con-
tacted at the Department of Sociology, Harvard University, 33 Kirkland St, Cambridge,
MA 02138; e-mail: wroth@wjh.harvard.edu or jmehta@fas.harvard.edu.
SOCIOLOGICAL METHODS & RESEARCH, Vol. 31 No. 2, November 2002 131-173

DOI: 10.1177/004912402237292
© 2002 Sage Publications
131
132 SOCIOLOGICAL METHODS & RESEARCH
the truth and whether a single “truth” really exists. The film makes
clear that there are different truths for these characters, for they are
not simply lying to protect themselves (in fact, the version of each
main party to the crime implicates the teller for the murder); rather,
they have deceived themselves into believing the version they have
told. These same questions about truth might be asked about contested
events in social research. When multiple sources relate different and
sometimes conflicting accounts of an episode, how do we decide who
is “right”? Is it possible that they all are right?
These questions may appear fundamentally incompatible, for they
imply different perspectives about the meaning of truth and objective
reality. These two perspectives, and the concordant forms of analysis
they imply, have traditionally divided social scientists in several fields.
The positivist approach maintains that a true explanation or cause of an
event or social pattern can be found and tested by scientific standards
of verification. The interpretivist approach does not seek an objective
truth so much as to unravel patterns of subjective understanding. The
latter assumes that all versions of the truth are shaped by the view-
ers’ perceptions and understanding of their world. In the case of a
contested event with many different versions, the analyst’s job is to
uncover what these versions reveal about the people who tell them,
their positions in the social structure of their communities, and their
cultural understandings.
In this article, we argue that the positivist and interpretivist appro-
aches, as we define them, are not fundamentally at odds with one
another but simply require different analytical lenses for the same
data. Questions about the causes of events may be answered using
research methods that permit replicability and allow causal theories to
be tested with additional cases. We may also use the same data but treat
the multiple and conflicting explanations of the event as data points,
evaluating what the nature of the dispute over an event reveals about
underlying social forces in a community.
We illustrate the combination of positivist and interpretivist research
goals using two case studies of lethal school shootings near Paducah,
Kentucky, and Jonesboro, Arkansas. The analysis is shaped by two
distinct research questions: (1) Why do school shootings occur? and
(2) How do people in these communities understand these school
shootings, and what do their interpretations tell us about them? These
Roth, Mehta / THE RASHOMON EFFECT 133
two approaches to understanding the events that took place outside of

Paducah and Jonesboro are equally valuable to social researchers in
their own right. Yet there is added value in asking both questions of
the same set of data, for answering one of these questions may help us
answer the other. Highly contentious events such as school shootings
can serve as a lens through which to understand deeper webs of mean-
ing within a community; thus, an inquiry into the causes of such an
event that uses a positivist framework may further interpretivist goals.
Similarly, interpretivist research, while not intended to provide a causal
explanation, can reveal hidden aspects of the culture and worldview
of community members; as members of these communities, school
shooters are themselves affected by this culture and perspective. To the
extent that interpretivist analysis of the community helps us understand
the actions of its members, it may enable us to use a positivist approach
to answer questions about the causes of contested events.
POSITIVIST AND INTERPRETIVIST APPROACHES

IN SOCIAL SCIENCE
The expressions positivism and interpretivism are used to varying de-

grees in the different social sciences, yet the distinction made between
the two concepts is common to most social science fields. The positivist
approach is modeled on the methods of the natural sciences. It seeks
knowledge based on systematic observation and experiment, with the
goal of discovering social laws analogous to the natural laws uncovered
by the methods of natural science (Angus 1986; Marshall 1994). Posi-
tivist analysis seeks to hypothesize and then evaluate causal inferences
about social phenomena that will be generalizable beyond the specific
data analyzed (Shankman 1984a; Lin 1998). This research approach is
not limited to particular methods, and indeed both qualitative and quan-
titative techniques can further positivist goals when they share a unified
logic of causal inference (King, Keohane, & Verba 1994, 1995). A fun-
damental assumption of positivist research is the existence of objective
reality and facts, which can be known or approximated through these
research methods. Analyses must be both replicable and testable across
cases, and the validity of the analysis will be evaluated accordingly.
Generally, hypotheses are generated and compared to other hypothe-
ses, with an eye toward validity, explanatory power, and parsimony.
The interpretivist approach has received more varied treatment by

different social science fields. It is perhaps most developed within
anthropology, where it is most strongly associated with the work of
Clifford Geertz.1 Geertz proposed a shift away from positivist analyses
toward the study of communally defined subjective understanding.
Defining culture as the “webs of significance” that man himself has
spun, Geertz (1973) stated that “the analysis of [culture is] therefore
not an experimental science in search of law but an interpretive one in
search of meaning” (p. 5). He claimed that what we think of as facts,
our data, cannot be truly objective because they are really “our own
constructions of other people’s constructions of what they and their
compatriots are up to” (p. 9).
On this account, interpretive analysis of subjective meaning cannot
be held to empirical tests of validity across cases because it is by nature
tied to particular cultural systems. The construction of causal laws
and patterns that are generalizable across cases necessarily divorces
the interpretation from what has happened in any particular case.
Because interpretive analysis cannot be systematically theorized and
assessed, according to Geertz, it must be self-validating. The legit-
imacy of the analysis should be measured by how well the “thick
description” holds up within the case and is supported by the evidence
put forth.
This approach toward interpretivist research has influenced other
fields in the social sciences. An interpretivist tradition in sociology
follows the lead of Geertz, typically by starting out with a situation—
often a changing one—and looking at how a certain group of people
understand and make sense of it. One preeminent example is Jonathan
Rieder’s (1985) study of a lower-middle-class community in Canarsie
during the 1970s as it experienced racial transition and school deseg-
regation. Rieder’s task was to understand why a largely Jewish and
Italian area that had been Democratic since the New Deal voted in
heavy numbers for Reagan in 1980. He wrote,
To make sense of middle-class discontent, writers have invoked white

racism, “embourgeoisement,” labor market rivalry, apple-pie authori-
tarianism, Lockean individualism, neopopulist retaliation, right-wing
protectionism, postindustrial society and political ethnicity. If each of
these concepts explain a piece of the puzzle, the danger is that the
pattern as a whole may vanish. (p. 3)
Rather than attempting to adjudicate between these competing hypo-

theses, Rieder (1985) abandoned these neat analytic categories and
plunged into trying to understand the meaning of social life as it is
experienced and constructed by Canarsie residents. Here, to take an
almost trivial example, he found working-class Italians appalled by
public profanity of young Black males; cursing at the factory in the
company of other men was acceptable or even expected, but swearing
in the streets within the earshot of women and children was most
certainly not. Out of hundreds of observations of this sort, Rieder
pieced together a complex and nuanced understanding of how real
and perceived economic, social, and cultural threats led to a backlash
against liberal politicians that—in the eyes of local residents—favored
the interests of the morally lax and irresponsible (Black) poor over the
upstanding and duty-fulfilling (White) lower-middle class.2
In the work of Rieder (1985) and others who share elements of
this interpretivist tradition (see, for example, Hochschild 1997, 1989;
Newman 1988, 1993), the analyst takes a phenomenon that is important
in its own right and seeks to understand it through in-depth interviews
and observation. But the author is not simply a recorder; the job of
the analyst is to put forward a coherent account that makes sense of
the often ambivalent, changing, or even contradictory beliefs that are
held by the respondents. The primary object of the analysis is social
meaning, but this in turn is an important constitutive element of social
action; understanding social meaning therefore provides one important
set of building blocks for understanding important social phenomena,
such as why Reagan was able to break apart what had previously been
a solid Democratic coalition.
Our approach to interpretivism takes its inspiration from the
Geertzian perspective and its successors in the sociological tradition
for which the primary focus of interpretivist research is the subjective
meaning that events hold within a particular culture. Interpretivist anal-
ysis should not be conflated with qualitative research, as interpretivism,
like positivism, is distinguished by its analytical approach and the goals
of the researcher, not by its methodology. These differing approaches,
as they are used in this article, are summarized in Table 1 below. As
employed here, the interpretivist approach is that which adopts the
broad goal of illuminating a set of social meanings that reflect cultural
beliefs and values. The positivist approach, by contrast, is that which
TABLE 1: Summary of Positivist and Interpretivist Approaches
Positivism Interpretivism
Causation—Seeks to understand the causal Interpretation—Seeks to understand how

explanation for a phenomenon or event people interpret a phenomenon or event
Objective reality—Presumes the “existence Subjective reality—Recognizes the
of facts” “construction of facts”; facts are seen as
interpreted and subjective
Generality—Analysis seeks a “law” that Specificity—Analysis is context specific and
extends beyond specific instances studied based only on the subjective understanding
of individuals within a specific context
Replicability—Analyses can be tested and Self-validation—Analyses can only be self-
verified empirically against other cases validating, through the consistency and
coherence of “thick description”
seeks causal laws to explain objectively viewed phenomena, whether

it uses survey methods or—as is the case here—detailed case studies
to do so.3
COMBINING POSITIVIST AND INTERPRETIVIST APPROACHES
Positivist and interpretivist approaches are sometimes posed as diamet-

rically opposed ways of conducting research. One form of the interpre-
tivist approach—that adopted by Geertz himself—assumes that objec-
tive truth cannot be known since all attempts to understand “facts” are
viewed through various subjective lenses, including the researchers’.
This position sees people’s understandings, and therefore subjectively
constructed social reality, as fleeting, dynamic, and constantly chang-
ing. Clearly from such a perspective, no combination of positivist and
interpretivist research is possible. However, there is another viable
current of interpretivism that flows from Geertz’s approach but is
more compatible with positivist analysis. A number of later writers
maintained that the view of objective truth as unknowable need not
prevent researchers from approaching that truth and should not be
considered incompatible with interpretivist goals (Rappaport 1979;
Rosaldo 1982; Shankman 1984a, 1984b; Farrer 1984). This position
acknowledges Geertz’s concerns about the problems of uncovering an
objective reality, yet argues that complete rejection of such knowledge

is counterproductive. Shankman (1984b) wrote,
I believe that complete objectivity is impossible; but this does not

mean abandoning the search for objective knowledge or conceding that
all versions of reality are equally true. What is necessary, according
to interpretivist Rosaldo (1982: 198) is “ways of moving back and
forth between an actor’s subjective interpretation and a set of objective
determinants” (p. 277).
Farrer (1984, p.274) agreed that data can be understood from a “prod-
uct perspective based upon cause and effect or stimulus and response”
or from a “processual orientation” that focuses on context, interpre-
tation, and the process of developing meaning but claimed that we
can “allow both product and process orientations” simultaneously in
research. This perspective maintains that subjective interpretations of
reality and objective phenomena may be simultaneously sought and
that it is furthermore important to understand the connection between
the two.4 The very fact that subjective understanding differs depending
on the context reveals that objective contextual conditions may influ-
ence subjective meanings (Shankman 1984b). Thus, compatibility can
be found between positivism and this understanding of interpretivism,
which focuses on why and how individuals come to understand events
as they do, yet recognizes that those understandings may be influ-
enced by an objective reality that, while difficult to discern, is poten-
tially knowable. Under this framework, it is possible to simultaneously
accept that there is both a single objective truth of factual events
and multiple subjective views of the truth that reveal much about the
worldviews and perspectives of those who hold them. The objective for
positivist analysis, as we discuss below, is therefore to come as close to
this objective truth about the causes of events as possible by managing
various forms of bias or distortion that may arise from the context or
the respondent’s interpretive understanding of events. The objective
for interpretive research, by contrast, is to understand how individuals
understand that context and how the multiple subjective “truths” they
construct provide insight into their cultural understandings. We argue
that each approach is important in its own right and that combining
the two has even greater analytic value.
Our approach to combining positivist and interpretivist analy-

ses to studying a contested event involves the following three
components:
1. Using a positivist approach to address questions of causation. What

causes school shootings? Here, a researcher needs to uncover the
factual precursors of an event to develop and evaluate theories about
its causes. If, for example, one explanation of the shootings is revenge
because of school bullying, we need to ascertain whether the shooters
were bullied and by whom. Did the shooters target those who bullied
them? Examination of the factual evidence can lead to a tentative
hypothesis within a case study. By incorporating other cases, we can
develop an explanatory theory, which will either be borne out or not
by future school shootings.
2. Providing an interpretive analysis. What does this tragedy tell us about
the culture5 of these small towns? Multiple subjective truths or ways
of understanding the shooting offer valuable insight into the social
structures, group tensions, and conflicting values in these communi-
ties. Rarely do these different accounts represent random variation;
more often they reveal systematic differences in perspective based
on the social status and position of the various community members,
as well as their relationships to the shooter(s) and the victims. They
may reveal the different stakes that community members have in the
shooting and the way it is interpreted by the wider society. These
divisions become a central object of analysis and help us understand
larger cleavages in the community. Here, the object of our inquiry is
not objective fact but the subjective interpretation that individuals in
different social positions bring to objective facts. We do not attempt
to posit a causal argument here.
3. Using interpretive analysis and positivist analysis to inform one
another. Interpretivist and positivist analyses, as we define them, are
not only compatible within the same set of data; they may each help
achieve the goals of the other. Interpretive understanding of communi-
ties is always informed by a number of hard, objective facts, which help
the analyst make sense of the subjective viewpoints of their respon-
dents. Basic background information about the economic climate,
racial or ethnic change, or the rate of turnover in a community is neces-
sary to provide an informed account of the respondents’views on those
same matters (see Newman 1988 for a good example of this interplay).
Using a positivist approach to answer questions about a contested event
that “breaches” social norms can also serve as a window into otherwise
hidden viewpoints. Deeper cultural understandings or social cleavages
may reveal themselves in those explanations about the events’ causes.
Conversely, interpretivist analysis can inform positivist analysis by
providing a more complete understanding of the social milieu in which
the shooters lived. Subjective perspectives of actors can point to direct
causes. For instance, how the shooters understood their plight at school
and why they saw the shooting as a means of resolving it may best be
revealed through an interpretive approach. More indirect causes, such
as why other kids thought it was appropriate to bully or why teachers
did not intervene, are also conducive to an interpretive analysis. These
are actions that are guided by socially constructed understandings of
meaning, and it is these systems of meaning that interpretivist analysis
helps us to clarify.
The remainder of this article illustrates these three objectives, look-

ing initially at positivist and then interpretivist analysis as each might
be conducted independently of the other and then looking at the advan-
tages of combining the two. First, we describe the fieldwork conducted
outside of Paducah, Kentucky, and Jonesboro, Arkansas, and give
a brief overview of the school shootings. Next, we take the posi-
tivist approach of seeking “objective truth” and discuss concerns over
the reliability of data from this perspective. In this light, the objec-
tive “quality” of data must be considered; we therefore demonstrate
methodological strategies to manage potential biases in the data that
may affect a positivist analysis of the causes of school shootings. We
pay particular attention to dealing with the conflicting or contradictory
responses that stem from this highly contested event.6 Then in the
following section, we illustrate the interpretivist approach of viewing
such contradictions not as problematic bias but as data revealing those
webs of significance. In particular, we focus on how “contested events”
can illustrate how social divisions, normally buried in the patterns
of everyday life, become revealed as people project these underlying
biases onto their understandings of the shooting. Finally, we illustrate
how both positivist and interpretivist analysis may inform one another,
by providing examples of how research questions of each type either
support each other or provide information that might have otherwise
been missed.
FIELDWORK ON TWO SCHOOL SHOOTINGS
We conducted in-depth qualitative case studies of the 1997 school

shooting at Heath High School in West Paducah, Kentucky, and the
1998 shooting at Westside Middle School just outside Jonesboro,
Arkansas. These studies were part of a National Academy of Sci-
ences report on lethal school violence (National Research Council and
Institute of Medicine 2002). Both events sent profound shock waves
to the very core of these communities, both because of the heinous
nature of the acts and because of the locations in which they occurred.
Both were in largely rural, White, low-crime communities that many
residents described as the last place they would ever have expected
such a brutal crime to occur. Furthermore, the assailants were students
at the schools they attacked, with no significant record of problem
behavior. The shootings were planned in advance and appeared to be
random killings rather than attacks against specific individuals. It was
partly the senseless and unexpected nature of these events that led to
disagreement and contestation, as community members struggled to
make sense of what had happened and what it meant.
In May and June of 2001, we conducted participant observation
and qualitative interviews with almost 200 people in the two com-
munities, including family members and close friends of some of the
shooters; school faculty and administrators; students and parents; civic,
community, and religious leaders; legal and police officials; and other
community members. These data were supplemented by police and
investigative materials, court records and depositions, psychological
evaluations of one of the shooters, the shooters’ own writings, school
district materials, and media reports. We briefly summarize these two
shootings below, but more details can be found in Harding, Mehta, and
Newman (2002); Fox, Roth, and Newman (2002); and Newman et al.
(forthcoming).
As students returned to Heath High School on December 1, 1997, the
morning after their Thanksgiving break, 14-year-old freshman Michael
Carneal pulled a handgun from his backpack and shot eight bullets
into a group of students gathered for a morning prayer in the school
lobby. He killed three students and wounded five and then put the gun
he had stolen from a neighbor’s garage on the floor and surrendered
to the school principal. Carneal, later diagnosed with a mild form of
depression and the beginning stages of schizophrenia, complained that

he had been teased and picked on by other students relentlessly and
said after the shooting, “People will respect me now.” While he was
not a social outcast and had friends and even had once had a girlfriend,
he was not central to any social group at the school. Carneal had joked
about bringing guns to school and taking over the school with a small
group of peers and had brought a gun to school and showed it to
classmates on two occasions prior to the shooting. The week before
the shooting, the boy whom classmates described as a jokester and
prankster warned some classmates to stay out of the lobby on Monday
morning and that “Something big is going to happen on Monday.”
Carneal’s parents, a respected lawyer and a homemaker, were heavily
involved in school and church events, and his sister was of one of
the school’s valedictorians. The Carneal family had a long history in
the community, going back several generations. Carneal pled guilty
to three counts of murder, five counts of attempted murder, and one
count of burglary and was sentenced to life in prison without parole
for 25 years.
In Bono, Arkansas, outside of Jonesboro, students and teachers
at Westside Middle School had just entered their fifth-period class-
rooms when 11-year-old Andrew Golden, clad in camouflage clothing,
entered the building and pulled the fire alarm. Golden then joined 13-
year-old Mitchell Johnson, already in position on a wooded hill almost
100 yards from the school, as the students and teachers filed out of the
school through their assigned exit routes. The 87 students and nine
teachers who exited the west entrance of the building were met with
a hail of gunfire that killed 4 students and a teacher and wounded
10 others. Earlier that morning, Johnson had stolen his mother’s van
and picked up Golden, and the pair stole guns from Golden’s father
and grandfather. They had an elaborate escape plan but were appre-
hended by police 200 yards from the school about 10 minutes after the
shooting. The day before the shooting, Johnson told peers he “had a
lot of killing to do” and that “tomorrow you will find out if you live
or die.”
Johnson had grown up in Minnesota, and his family had recently
moved to Arkansas. His mother, a former corrections officer, was
divorced and remarried, and the family was quite poor. Johnson had
been sexually abused by a neighbor as a child and had an explosive
temper that landed him in some trouble at school. Despite his occa-
sional outbursts, Johnson was generally known by the adults at his
school as an exceptionally polite boy, who did whatever he could
to please adults. Golden’s family had lived in the Westside area for
several generations, and his parents worked for the post office. He
was not considered a discipline problem at school but was known to
neighbors as a menace, a boy who rode around on his bicycle with a
sheathed hunting knife strapped to his leg. He was taught to hunt at
a young age and was an expert marksman. Some residents claimed
he tortured and killed cats. Too young for adult court, both boys were
sentenced to juvenile detention until they turn 21, at which point they
will be released.
POSITIVIST ANALYSIS OF CONTESTED EVENTS
There are a number of methodological difficulties related to trying to

develop a positivist causal theory of school shootings, many of which
stem from the fact that school shootings are such rare events. Included
among these many difficulties is the lack of data points to use in adjudi-
cating between many causal theories and the difficulty of generalizing
from information gathered in a single case or cases (Harding, Fox, and
Mehta 2002 [this issue]). But in addition to the problems created by
the rarity of events like school shootings, another set of complications
ensues if the rare event is also one that is contested. Positivist analysis
must be rooted in accurate data, the “building blocks” on which any
causal theory must be based. If the data take the form of a case study,
then the building blocks are an accurate and comprehensive account
of what happened during the event, the relevant acts that preceded
it, as well as details about the shooters’ past experiences and the
reasons for their actions. Gathering this information is complicated by
a number of problems that are common to most retrospective research
but are exaggerated when the highly charged nature of an event per-
petuates divergent, and often contradictory, accounts of events. The
controversial and highly publicized nature of the school shootings
outside of Paducah and Jonesboro magnified these problems; as a
result, these case studies can elucidate processes affecting data quality
that may be less apparent in other forms of causal analysis of past
events.
CONCERNS OVER DATA QUALITY IN CONTESTED EVENTS
We identified three categories of problems that plagued our attempts

to gather complete and accurate data.7 First, there were the problems
of memory, where respondents simply could not remember crucial
details about the time before the shooting. Second were the problems
of vested interests, where respondents who were seeking to protect the
professional status or personal reputation of themselves or others, or
were attempting to sway the outcome of the research because of their
personal or political beliefs, consciously filtered what they said. If the
issue in the case of memory problems was respondents’ current lack
of knowledge, in the case of vested interests, respondents know but
are deliberately lying, or, more commonly, filtering what they say in
an effort to present what they perceive as a more favorable version
of events. Third, and potentially most problematic for the researcher,
are the cases that fall between the two: where respondents honestly
think that they know something but are mistaken for any of a variety of
reasons. We will call this the problem of mistaken judgments. While it is
often the case that mistaken judgments are influenced by respondents’
vested interests, for our purposes, vested interests problems refer more
narrowly to the conscious filtering of information, whereas mistaken
judgments include all unconscious errors that respondents can make,
only some of which are the result of vested interests. Individuals
may also reach mistaken judgements by employing “heuristics,” or
mental shortcuts, to reach what are often erroneous conclusions (Fiske
and Taylor 1984; Bazerman 1998), yet this can be unaffected by the
respondents’ vested interests.
In practice, these categories of data problems may be present simul-
taneously, and it may be difficult for researchers to determine which
type of problem is most prominent.8 But analytically, it is helpful to
distinguish between the three types of problems because each directs
the researcher to distinct strategies to try to counter the bias and
improve the collection of information. We will consider each of these
in turn and then propose some strategies to deal with these potential
sources of bias in the following section.
Memory Problems
Perhaps the most obvious challenge to understanding an event that

happened more than three years before the data collection is the simple
problem of inadequate or inconsistent memory. Many respondents

simply could not remember relevant information. For example, school
personnel had trouble remembering whether programs put into place
to help freshman adjust to high school appeared before or after the
shooting, a crucial difference. The shootings were defining moments
that brought about dramatic changes in school policy, student climate,
and town image. It was difficult for respondents to remember clearly
the different world that prevailed before this life-changing event. To
take one obvious example, students from Heath in Michael Carneal’s
9th-grade class, who were in 12th grade when we interviewed them,
found it hard to remember the specifics of 9th-grade peer groupings.
Many of the students reported that there had been a conscious effort in
the wake of the shootings to become less petty and nicer to one another,
and thus while students could tell us generally that social exclusivity
was higher before the shootings, the ins and outs of 9th-grade social
hierarchies before the shooting were somewhat forgotten.
The ability of interviewees to recall information was also affected
by the traumatic nature of the event. Many who were closely connected
to the shooting—often those who should know the most about it—have
been diagnosed with post-traumatic stress disorder, and it is likely that
many more suffer from this condition who have not been diagnosed.
The stress and emotional difficulty of reliving past traumatic events
such as a school shooting can cause respondents to block information
about the events from their own memories. In the Jonesboro case, for
example, we interviewed one of shooter Mitchell Johnson’s friends
who had also been among the students to exit the building during the
fire alarm and walk into the line of fire. He explained how badly upset
he had been by the shooting and how it affected his behavior for a long
time afterward:
Respondent: [After the shooting,] I would not sleep in my room. I would

have nightmares in my room, so I slept on the couch, my couch
for about six months. I would sleep on my couch in the living
room.
Interviewer: Was it better if you slept on the couch because you didn’t have
nightmares?
Respondent: I didn’t have nightmares. There was a couple times I had
nightmares, but it wasn’t serious nightmares. . . . I would just
lay in my bed. See, my first time laying in bed, oh my gosh, I

was freaking out.
From this interview, it appears that the respondent was badly upset by
the shooting in general, but it is not clear why he preferred not to sleep
in his room. Later, after talking to an adult who was close to him, we
found out that this respondent had been diagnosed with post-traumatic
stress disorder and had repressed painful and difficult memories. Only
after a long period of therapy did it come out that Mitchell Johnson
had been over at his house a few days before the shooting and had
told him of his plan while sitting on his bed. This was the reason he
had felt unsafe sleeping there and had had nightmares, but it took him
many months to remember this fact, and even after remembering, its
painful recollection was enough to prevent him from relating it again to
researchers. Knowing that one of the shooters had leaked his intentions
a few days before may be an important puzzle piece in adjudicating
between theories that account for the shooting. But the traumatic nature
of the event and memories associated with it may hinder the research
process.
Vested Interests
Even when respondents could accurately remember past events, some-

times they clearly had vested interests that affected the content of what
they said. The most obvious form of vested interests was in which
respondents were protecting themselves or the institutions that they
represented from accusations of blame. Given that there were past
and pending lawsuits, those who were in some way implicated in the
events were concerned with defending their own actions from future
accusations of blame. At the extreme, some teachers, administrators,
and fellow students simply refused to talk with us because of the
potential implications for legal proceedings. Among those who did
talk to us, their own practical interest in a certain account of events
must be a consideration. Even setting the lawsuits aside, the purpose
of the research—in our case, that it was part of a congressional report
that had implications for future policies—may influence the content of
the responses. One school official sought to direct the interview away
from a discussion of school as the root of the problem and instead
put responsibility on the family. At the same time, this informant was
careful to emphasize the view that the family was not an appropriate
sphere for governmental intervention:
If [something] happens at home, how is the school going to know?
They’re model students at school, but in the community they’re not. . . .
School does not mean to be intrusive into the homes unless we’re
asked to. We’re not big brother. The home life is home life. It isn’t, the
government should not be sitting in the living room with them. . . . How
are you going to assess the needs in schools, the needs of the student,
when . . . this all started with . . . the home life.
Here, the interviewer’s attempts to discuss ways that schools could

identify potentially problematic children are rebuffed by a school
administrator whose personal view is that this is not an appropriate
role for the school.
A second type of vested interest is when the respondents are not
protecting themselves or deferring blame but rather are using the
interview to promote their own agendas. For some respondents, the
opportunity to become part of a congressional study led them to filter
information so as to influence policy. Asked for her explanation of the
shooting, one educator gave a long description of how she felt school
testing was creating undue pressures on teachers, which distracted
from their ability to connect with their students:
There has been [sentiment] brewing that public schools are no good or
not doing a good job and we are going to fix that by upping the ante and
making them more accountable. They have to prove and they have to
show us that they are really doing what they are supposed to do and we
are going to punish them if they don’t. We are going to put their names
in the papers. We are going to take away their money. We’re going to
make them fire teachers, all that kind of stuff. It creates an attitude of
fear within schools about making kids learn . . . . So teachers are under
a lot of pressure and they are scared of these tests. And kids are too.
Kids throw up. Kids get anxious. Kids don’t want to go to school. They
feel the stress of these changes, too.
There is very little evidence that this is a relevant issue in these specific
schools or for these specific students. However, knowing the potential
audience of the report, this respondent ties testing to school violence
without any real knowledge of whether this is empirically supportable
in an effort to mobilize public sentiment against testing.
Mistaken Judgments
More difficult still were cases where respondents thought that they
knew something but were mistaken. One way that judgments were
mistaken was when respondents were repeating second- or third-hand
information that stemmed from an initially erroneous source. On a
number of occasions, respondents substituted what they had read in
the media or heard around town for what they actually knew, often
without acknowledging as much. For instance, in Jonesboro, some of
our respondents told us that there was a third shooter involved in the
shooting, a fact that was not supported by any police or media account
but that had spread quickly through the rumor mill. The shootings were
focal events in the lives of these small communities, and stories and
speculation about them circulated widely. Differentiating what people
actually knew from what they had heard or read became a central
problem in the research.
A second form of mistaken judgment occurred when respondents
made faulty inferences based on incomplete information. Psycholo-
gists have identified a variety of heuristics, or mental shortcuts, that
individuals use to make judgments under conditions of uncertainty
(Tversky and Kahneman 1974). While heuristics can be an efficient
way of simplifying a complex world, they can also lead to cognitive
errors when individuals attempt to reach conclusions about what they
do not know on the basis of what they do know (Bazerman 1998).9 For
example, in our research, some respondents attributed a greater causal
role to the factors or individuals they were knowledgeable about. In
Jonesboro, where there were two shooters, some respondents ascribed
the primary role to the shooter that they personally knew better. One
student who knew Mitchell Johnson well but did not know Andrew
Golden at all claimed that Johnson was the instigator and he got
Golden involved because of his access to guns. To accurately assess the
probability of who was the more likely instigator, information about
one boy needs to be balanced against information about the other. But
in this case, the respondent’s knowledge of one of the shooters but
not the other may partly weight his judgment of the probability of the
shooter he knew taking the more prominent role.10
Another example of how the use of heuristics can lead to erroneous
conclusions was when one student at Westside claimed that Andrew
Golden was taking the drug Ritalin, and that this probably affected his
behavior during the shooting. It became clear during the course of the
interview, however, that alerting the public about the negative effects
of this drug was the cause célèbre of the student’s mother, as well
as some of their family friends. The students’ willingness to assume
that Andrew the school shooter must have also been a Ritalin taker
is shaped by the students’ prior beliefs that Ritalin usage is highly
associated with negative outcomes.
MANAGING DATA BIAS IN POSITIVIST RESEARCH
There are no magical solutions to the problems raised here. The first
and most crucial step is simply awareness of the various reasons
why respondents may give incomplete, inaccurate, or contradictory
information. If this awareness infuses the research design, the field-
work itself, and the analysis of the data, some of these problems
can be minimized, and more accurate information can be obtained.
Of course, part of this process is to construct factual questions that
limit people’s abilities to pontificate or state opinions instead of facts.
But since subjective views and vested interests will inevitably shape
some of those responses, we need to consider how to manage the
bias that does occur. We focus here on two complementary strategies
to improve the quality of factual data: (1) contextualizing individual
respondents’ comments in light of everything known about the source
of their knowledge, their personal or political agendas, and their social
position; and (2) triangulating among various respondents and sources
of data based on this specific contextual knowledge (see Table 2). These
strategies can be combined or tailored to the specific nature of the data
problem.
Contextualizing Responses
The process of contextualizing responses is that of framing the infor-

mation given in the context of where it comes from and therefore how
reliable it can be considered. This is a process of gathering supplemen-
tary data that give us additional information about the primary factual
data that are our main concern. This data-gathering process, focusing
on relevant background information to identify potential biases, then
TABLE 2: Biases in the Data and Suggested Approaches in a Positivist Framework
Proposed Approaches
Contextually Informed
Contextualizing Responses Triangulation
Biases Found in the Field (Data Gathering) (Data Analysis)
Memory problems
• Inadequate memory of Collect multiple sources of Discover whether memory
past event data, particularly written problems exist through
• Memory altered by impact documentation or accounts comparing interview and
of contested event given immediately after the noninterview data. Give
• Memory affected by post- fact (e.g., police greater weight to non-
traumatic stress disorder investigation reports, media interview data in areas
accounts). where memory is likely an
issue.
Vested interest problems
• Importance of protecting Probe respondents to Using contextual informa-
reputation of self and discover personal beliefs, tion about respondents’place
others interests, and relationship to in the social structure, rela-
• Attempts to influence topic of research. Stratify tionship to event at hand,
content of study to reflect sample not demographically and other salient motivations
political beliefs but theoretically so that it or interests that may affect
• Psychological interests in contains respondents with data, give lesser weight to
maintaining self-esteem, different vested interests. accounts that are motivated
self-importance, and so by vested interests.
forth.
Mistaken judgment problems
• Secondhand or thirdhand Do not accept information Look for similarities bet-
information stemming or inferences at face value. ween different interviews
from erroneous initial Probe to ascertain how and between interview and
source respondents know what they media accounts. Privilege
• Incorrect inferences know. Trace secondhand data not by its frequency but
through use of various information to its source. by reliability of its source.
heuristics
helps the researcher determine the best course of action to analyze the
data collected.
An important task for assessing the quality of positivist data is to
find out not only what the respondent knows but also how he or she
knows it. Often, this may be accomplished by including questions
about the source of the knowledge in the in-depth interview.11 For
instance, in the case of the respondent who ascribed part of Andrew’s
problems to the taking of Ritalin, further questions demonstrated that
the respondent did not really know whether Andrew was using the
drug:
Interviewer: How did you hear that Andrew was on [Ritalin]?
Respondent: I just knew because I was in his class. And I knew that he was
really hyper. . . .
Interviewer: Was he always hyperactive?
Respondent: Always, yeah. I’m pretty sure he was on it because I know
they would call him down to the office and he would take
some medicine. I’m pretty sure that’s what it was, but I’m not
totally sure.
This knowledge helps the researchers recognize whether this version
of events is based on circumstantial evidence rather than concrete
knowledge and determine how much weight to give it until it can be
confirmed or rejected.
When nothing in the interview itself reveals the source of knowledge
for a given fact or set of facts, the researcher should also be aware
of how closely the respondent’s comments correspond to accounts
in the media or from those of others in the community. On numer-
ous occasions, respondents related details that were almost verbatim
from media accounts, and therefore the fact that these accounts were
reported by multiple responses did not make a stronger case for this
version of events. This pattern may also occur as a consequence of
rumors, especially when numerous respondents had talked to one key
respondent, and thus details that seemed to be supported by multiple
“observations” are more properly considered multiple manifestations
of a single observation. One example was the story that there was a third
shooter involved in the Jonesboro case; although not reported by the
media, this story spread quickly throughout the community. We traced
its source to the police investigation immediately after the shooting,
where numerous children reported this account. This questioning took
place after the traumatized students had spent several hours huddled
together in the school gymnasium. During this time, some students
revealed to teachers and each other that Mitchell Johnson had pre-
viously made threats that “something big was going to happen” and
that he, Andrew Golden, and a third person were planning something.
Those who were specifically asked if they had heard this report of a
third accomplice directly from one of the shooters mostly revealed that
they had not. But many people in the community remain convinced that
a third shooter was involved. A teacher who was with the children in the
school gymnasium immediately after the shooting stated emphatically,
The boys hadn’t been arrested at that time. [All the kids in the gym]
were crying and they knew [who had done it]. Actually they said it
was Mitchell Johnson, another person, and Andrew Golden. They said
there was a third person who to this day and [as long as I live], I will
believe there was a third person.
Because this account was so widespread among the student body

immediately afterward, it took on credibility in respondents’ minds,
even though no evidence of a third shooter was ever uncovered in the
police investigation.
In addition to trying to assess the source of respondents’knowledge,
a respondent’s comments should also be evaluated in light of every-
thing else that can be discerned about them. This vital but peripheral
information might include the respondents’ political beliefs, occupa-
tional position, position in the social structure of the community and
relationships with other social groups, and their relationship to the
shooters or victims. Most obviously, this is a caution not to take at
face value the football player’s reports about bullying or the principal’s
reports about school climate before the shooting. In these cases, the
respondent likely has a personal or institutional agenda that affects
what they will and will not say. But this attention to context can
also explain subtler biases such as why, for example, a student whose
mother was involved in speaking out about the dangers of Ritalin might
infer that Andrew Golden was taking Ritalin. On some occasions,
this information will come not from the interview situation but from
participant observation and immersion in the setting.
In some cases, the source of bias in an individual’s account will lie
not in the respondents’ political beliefs or their occupational position
but rather in their relationship to the shooter or victim. Questions
that seek to probe these relationships may be important in evaluating
the importance of the data. Consider, for example, the student who
presented Mitchell as more responsible than Andrew for the crime. A
later question revealed that this presentation of information was shaped
by his greater knowledge of Mitchell:
Adult12 : You make it sound like it was more Mitchell. I always thought
that Drew was the main instigator.
Respondent: Well, I always thought that too. Drew may be, but I never talk
about Drew because I didn’t know Drew.
Adult : You didn’t know him. OK.
Respondent: So I shouldn’t talk about him.
Understanding that this respondent’s perspective is shaped by his social

position in relation to the shooters is important for assessing the claims
he makes about the role played by each. This is especially the case
because this makes it clear that, despite his focus on the leading role
played by Mitchell, he is not himself convinced that Mitchell was the
instigator.
The example of this respondent also illustrates how respondents’
claims about one topic need to be contextualized in a deeper under-
standing of their motivations and interests. Why would the fact that this
student knew Mitchell better than Andrew make him more likely to
portray Mitchell as the shooter? Is it not more reasonable to suppose
that he would blame the shooting on the student who was not his
close friend? The respondent’s actions throughout the interview have
a bearing on this question. At various points during the interview,
the other adult present would begin to provide information, and the
respondent would jump in to correct her (sometimes correcting her
falsely), as if to assert his authority in this interview. He seemed to
view the interview as a chance to show off his unique knowledge about
one of the shooters, as seen by his answer to an unrelated question that
opened the interview:
Interviewer: Part of what I wanted to ask you is just about what life was
like at school, at Westside . . . what it was like before all this
happened, if you can remember back then?
Respondent: Okay. I remember life was good. Mitchell, one of the people
that did this, was one of my best friends. We talked all the
time. It was . . . OK, Spring break, me and him and a few of
my other friends, we were out at the railroad tracks at [Bono],
we carried this little BB gun pistol with us . . . and it was like,
“Well. . . I love this gun.” And [Mitchell] pointed [the gun]
at us and stuff, messing around. It didn’t, it wasn’t filled with
nothing anyway, but still, we were just messing around. And he
was like, he had this death list he wrote out of who he wanted
to kill. That’s where like messing around with me and stuff,
made me mad.
While we should not automatically conclude that this account is inaccu-

rate, the very fact that the respondent volunteers this information
without being asked about the shooter reveals how he interprets the
interview setting—as a way to increase his own status by illustrating
his access to important information that he assumes the interviewer will
want to hear.13 Understanding his approach to the interview is crucial
to interpreting his tendency to portray Mitchell as the instigator of the
shooting despite his own suspicions to the contrary.
Contextually Informed Triangulation
The best way to avoid being misled by a single inaccurate or biased

source of data is to include as many viewpoints and as many sources of
data as time, money,and convenience permit.Using multiple interviews
and multiple sources of data (such as police interviews, psychological
reports, primary documents, etc.) to triangulate the problem—that is,
to see if different sources agree on the same set of facts—is perhaps
the most widely used approach in the social sciences, quantitative as
well as qualitative (Zetterberg 1966; Phillips 1976; MacKay 1978;
Jick 1979; Kirk & Miller 1986; Yin 1989; Miles & Huberman 1994;
Denzin and Lincoln 2000). In this sense, triangulation is similar to the
common rule in journalism not to report anything that is ambiguous
or controversial without at least two and preferably three sources to
give it additional support.14
However, when researchers (or journalists, although using different
terminology) discuss the use of triangulation to reconcile inconsistent
data, they often use an implicit assumption that research should privi-
lege the side that has “more” evidence. There is perhaps some positive
weighting for official sources like newspapers or police reports and a
slight discounting of interview data that are secondhand or thirdhand.
Yet positivist analysis that is informed by a contextualized understand-
ing of the positions and interests of respondents can triangulate in a
way that is more consistent with the goals of uncovering causation.
As we have seen, a piece of evidence need not be more true because
more people relate it; the greater availability of supporting testimony
may derive from their all having heard it from the same source, be it
media or word of mouth, or because many respondents have personal
or political reasons for contributing such an account.
Contextually informed triangulation takes this understanding into

account during the process of triangulation and the subsequent weigh-
ing of evidence. Rather than adjudicating a particular question on
the basis of which interpretation has more evidence, the goal is to
differentiate between sources of evidence, privileging data that are
likely to be less subject to the known sources of bias, and holding less
reliable evidence to higher standards of support from triangulation. In
the analysis phase, this means understanding each piece of information
within its context and tailoring the triangulation to the nature of the
data, sometimes at the level of specific questions. Contextualizing
responses is therefore a necessary first step for contextually informed
triangulation, but the two processes are iterative. Contextual informa-
tion about the type of bias present influences the type of supplementary
data that is sought. New data should be collected that are likely to be
free of this form of bias and that would receive greater credibility as
different sources of data are weighed.15
To give one example, when bias results from memory problems,
either stemming from the fact that the event was long ago or that it
altered the way many see and therefore remember the world, particular
effort should be taken to seek support from written documents of the
period. The claim that a program to help freshmen adjust to high school
was in place before the shooting should be supported by written doc-
umentation relating to the policy, particularly when this information
comes from a respondent who seems unsure of the timing of events.
Students’ uncertainty about the structure of peer relations before the
shooting because the shooting altered those relations may be checked
against photographs from yearbooks showing candid groupings of
students and membership in different school activities at the time; such
a source would likely be most effective if shown to the respondents
to trigger their memories or enable them to interpret the documented
information (which may be misleading to the uninformed outsider).
Other biases may call for triangulation with other forms of data. The
question of Andrew’s alleged use of Ritalin should be somewhat dis-
counted if the source of the information has a stake in this possibility.
To determine Andrew’s Ritalin use, a researcher might consider reports
from both official sources such as a school administrator who would
have had to administer the drug to Andrew in the school office, close
friends or relatives who might be expected to know about his medical
history, the accounts of other students and teachers who do not have
similar personal stakes in the issue, and written documentation. There
are clearly practical limitations to these approaches: Some of these
avenues may be confidential or unavailable. But contextually informed
triangulation allows for an appropriate weighting of the information
that is available from different sources.
Developing a contextualized understanding of respondents and the
information they give and using that knowledge to shape triangulation
in a way that balances potential sources of bias are important tech-
niques for managing poor data “quality” from the positivist perspective
of seeking objective facts. In the case of a highly contested rare event
such as a school shooting, considerations such as these are necessary
before researchers can go on to distinguish causal patterns and develop
explanatory theories that, held against additional cases, may gain status
as hypotheses.
INTERPRETIVIST ANALYSIS OF CONTESTED EVENTS
While hypothesis testing and causal theorizing provide a specified

goal for the positivist researcher, the interpretive analyst has a no
less important but much less clearly defined task. Interpretive analysis
seeks to uncover the often-competing sets of social meanings held
by respondents. A lengthy substantive rendering of our interpretivist
analysis of these cases can be found in Newman et al. (forthcoming),
but in this space we illustrate how interpretivist analysis might work
with a specific example that ties the analysis to the contested event.
While interpretivist analysis is not restricted to contested events, we
focus here on the special value of studying such events for interpretive
analysis. Contested events create a series of methodological prob-
lems for positivist researchers because of contradictory and incon-
sistent responses; yet for interpretive analysts, these same conflicting
responses provide key data points that help us understand social mean-
ings and underlying divisions within these communities.
THE CONTESTED EVENT AS A CULTURAL BREACH
An atypical event that is controversial can serve as a breach of the social

norms of the community (Feldman 1995). Harold Garfinkel’s (1967)
work illustrated the value of breaches in identifying the norms and

values of an individual or community. Researchers using Garfinkel’s
approach would deliberately violate simple social norms to observe
the reaction of their subjects. These responses, often irritated or angry,
exposed the underlying socially agreed-on understandings that govern
much of daily life.16
An extremely traumatic event such as a school shooting acts as a
breach of the accepted understandings of how people usually behave in
these communities. Children, particularly White children in low-crime,
rural communities, are not supposed to shoot their classmates. How
were our respondents to make sense of this anomaly that challenged
some of their most strongly held assumptions about their communities?
Given the shocking nature of the shooting and lack of real information
about its causes, residents were free to project their own interpretations
on the events. Here, we focus on one example to illustrate how reactions
to the shooting can be used to identify important aspects of the cultures
of these communities.
One of the most important divisions that emerged in our discuss-
ions with respondents was not race, religion, or class, but that bet-
ween long-standing community residents (“insiders”) and newcomers
(“outsiders”). While many communities develop a distinction between
insiders and outsiders, how respondents defined these categories and
drew boundaries between them is an important interpretive goal
(Gieryn 1983; Lamont 1992, 2001). Our interviews revealed that in
these communities, great salience is given to one’s “generational
status,” which served as one of the most important stratifiers of social
life.
In both communities, “insider” was not a title that could be earned
by longtime membership or conformity to local cultural norms but
was attributed only to those with multiple generations of family born
into the community. For example, one teacher at Heath High School
claimed that she would never be a “Heatheren,” even though she has
been teaching at Heath for 27 years, because she grew up in “the city”
(i.e., Paducah). Although in the national media, Heath and Paducah
are often conflated (the tragedy is often described in the press as
the Paducah shooting), to locals there is a stark demarcation between
the city of Paducah (population 25,000) and the rural town of Heath
20 miles away.
“Insiders” with multigenerational family histories had higher social

status in the communities and were perceived to enjoy better treatment
as a consequence. These hierarchies surfaced as an issue after the
shootings, framed in terms of the local response to the shooters’ fami-
lies. The families that had been part of the community for a long time
were seen as receiving preferential treatment. In Paducah, one victim’s
family objected to how positively the Carneals were portrayed within
the community after Michael had shot their daughter, including a “big
campaign” by their church for the community to embrace them. Asked
whether he believed that the Carneals received favorable treatment by
the community or the legal system because of his class status as a
lawyer, one member of the Carneals’ church claimed that, “I don’t
think it’s because his father was a lawyer as much as the fact that
they were well-liked people and highly regarded.” However, he later
explained that this high regard in which they were held in the church
community at least stems from the fact that
[John Carneal’s] folks were raised here. He was baptized here. His
parents still go to church here, his brother and his family still goes to
church here. And they started going to church after they got married.
They were even married at the church.
The extent of the “preferential” treatment they received is in dispute,

but certainly the status of the Carneals and others in the Heath com-
munity is related to their history and length of connection to that
community. The victim’s family, however, was extremely upset by
this treatment and chose to frame the issue in terms of their insider
status.
In Westside as well, Andrew Golden’s family was well established
and went back many generations. Some felt their status and connec-
tions led to special protections. Well before the shooting, several dogs
in the neighborhood were found hurt or shot, and Andrew’s grandfather
was suspected of being responsible. A neighbor reported this to the
sheriff but claimed that nothing was ever done about it. The neighbor
felt that the sheriff turned his back on the complaints because the
grandfather was well known and well established in the community,
especially in law enforcement circles. After the shooting, Mitchell
Johnson’s mother also stated publicly that the Goldens received pref-
erential treatment in the local media coverage of the shooting because
of their higher status in the community. In a newspaper article, she

claimed that the local media coverage made Mitchell look worse to
detract attention fromAndrew because his family was more established
and she was seen as an outsider. The local newspaper, The Jonesboro
Sun, and others in the community denied this allegation.17 But for the
purposes of interpretivist analysis, what is important is not whether
the community really did give more favorable coverage to the Golden
family but that it was perceived to be especially friendly to the Goldens
because of the family’s long-standing community roots. That this
perception is held by a relative newcomer to the community only
underscores the level of distrust between insiders and outsiders.
One teacher at the Westside Middle School outside of Jonesboro
also felt that this insider/outsider dichotomy surfaced in the treatment
given to the shooting victims. She noted that one girl, who had always
gone to school at Westside and whose family were longtime residents,
received so many gifts and stuffed animals from community members
after the shooting that they had to clear her hospital room out every
day. Another victim, whose family had only just moved to the school
that year, received little attention. This differential probably developed
partly because the newcomer knew fewer people in the community,
yet the fact that the teacher brought it up in these terms reveals her
concern over the unequal status given to members of the community.
In these examples, an understanding of the fault line in the commu-
nity between longtime residents and newcomers was revealed explic-
itly through their reactions to the shooting and its aftermath. But this
interpretation was also buttressed by answers to a series of questions
that had no relation to the shooting. For example, when we asked what
made students popular at school, one Heath student explained that this
was related to family status:
Popularity, first thing, at Heath is based on how long your family has
been in the community. That’s pretty—and how much money you
have—but it really was an issue [of], like how long did your great-
grandparents go to Heath? And then you’re popular. And if they didn’t,
then you’re probably not.
At Westside as well, students who had been in the community for many
generations were well known within the school and were more often
popular.
Thus, respondents could vocalize this wide-reaching stratification

between insiders and outsiders without reference to the shooting. How-
ever, from the interpretivist perspective, the shooting is a breach that
forces people to confront social divisions that are already present
but are sometimes submerged in the course of their everyday lives.
That interpretations of the community’s response to the shooting were
framed in terms of this hierarchy of generational status is but one
example of how interpretivist research can uncover the presence and
importance of particular value systems to community life. It may be
possible to uncover these themes without an intervening contested
event, but using the shooting as a window into the communities’
cultures gives the researcher particular analytical leverage in under-
standing how much weight to give these themes.
POSITIVIST AND INTERPRETIVIST ANALYSIS

AS COMPLEMENTARY PROCESSES
Up to this point, we have considered positivist and interpretivist anal-

ysis on their own merits, as two separate but equally important goals
that a researcher can pursue using a single set of data. Here, we argue
that these approaches can additionally inform one another in ways
that further the goals of each. In the following section, we offer a
nonexhaustive discussion of some of the ways that these approaches
contribute to one another.
USING POSITIVIST ANALYSIS IN THE AID OF INTERPRETIVISM
All interpretivist analysis is informed to some degree by relevant fac-

tual information. Understanding the interplay of social and economic
change and individuals’ lived experiences of these changes is one
of the primary tasks of sociology, and thus most good interpretivist
work is informed by a factual understanding of the broader context. In
Rieder’s (1985) study, for example, patterns of Black encroachment
onto formerly White areas as well as the growth of social policies that
favored the poor during the Great Society period are two crucial pieces
of positivist data that inform his interpretivist conclusions.
Yet our research showed a more specific way that positivism can
inform interpretivism: Using conflicting responses to factual ques-
tions can help uncover previously hidden social structures and value
systems. Thus far, we have treated inconsistencies in the accounts
of our various respondents as underbrush that needs to be cleared
away to get onto the real work of answering important questions
in a positivist framework. But from the interpretivist vantage point,
many of these inconsistencies reflect differences in the community
that are themselves of central analytic interest. What were problems
of data quality become data points of their own right when we don the
interpretivist hat. What is important is not whether the information is
objectively true but rather whether there is a way to make meaning of
the subjective truths that respondents present.18
Consider the rather specific question that we asked of most of
our respondents in Jonesboro: Who was primarily responsible for the
shooting—Mitchell Johnson or Andrew Golden—and what evidence
would support that conclusion? In response to what we had intended
as a factual query,19 we often received some of the most revealing
statements about how our respondents understood their communities.
Consider this lengthy response from a third-generation Westsider and
teacher at the school about which boy was more responsible:
I’ll tell you what my perspective of it is. You know the Oklahoma
bombing? . . . You know, of course that woke up America and shocked
America.And I’m just like anybody else. When I first heard it, I thought,
“Okay. This has got to be somebody from another country. It couldn’t be
an American. It couldn’t be.” And when they started saying, you know,
that day, the next day that they were looking at Americans, charging,
I was just in shock. I thought there is no way. I couldn’t accept that
someone from America, where we’re from, [could do] that.
And that’s kind of the way it is with the shooting because when
it happened, it was like, “Okay. Someone didn’t like Westside.” You
know, “There’s something, some foreign person from another school
or whatever must have done this.” And then when they started talking
about someone from Westside doing it, I was, it’s like the Oklahoma
thing. I just thought this can’t be. You know, I just thought this can’t be
somebody from Westside because we’re all like a family.You know, we
all help each other. We’re, everybody from Westside is a person from
Westside, you know.
And I mean, I don’t know, and I don’t want to say anything against
anybody, but that Mitchell Johnson, you know, he, I consider him to be
an outsider, I now say. And maybe I shouldn’t even think of it like that,
but I do. I can’t help it. And Andrew, . . . I mean, I went to school with
his [uncle], you know, thought they were good people and everything.
And I, in my mind, I guess, I feel like that Mitchell Johnson was the
one that thought it up and maybe influenced the other one. I mean, I
know he’s got a troubled background with his family and everything.
And I just think he’s the instigator. I mean, not that Andrew, I mean of
course he did some of the shooting too, but I still can’t rationalize that
someone from Westside, you know, thought of doing that.
As this teacher and member of the Westside “family” tries to make

sense of the shooting, she reveals that she perceives an important
difference between those from Westside and those who are not.
By asking a question about the causes of the shooting, we have
opened a window onto the issue of how respondents interpret social
divisions in the community, a sensitive topic that we might not have
been able to understand by broaching it directly.20 This teacher expres-
ses some reluctance to think of someone as an outsider, either because
of a desire to be inclusive or because of concern over how this may
sound to others, but her true feelings about her community overpower
this concern. Her beliefs about insiders and outsiders, and the differ-
ences implied by these titles, become clear in her interpretation of the
shooting and reveal the presence of this cleavage in the community
life in general.
USING INTERPRETIVIST ANALYSIS IN THE GENERATION

OF CAUSAL THEORIES
Interpretivist analysis can provide crucial information to inform posi-

tivist causal theories that might otherwise have been missed. Here, we
explore two ways in which knowledge of the subjective understandings
of community members can improve causal analysis: (1) by improving
data collection and analysis, using what we call interpretively informed
triangulation; and (2) by allowing for a more informed understanding
of the worldviews of the shooters and those with whom they interacted.
Interpretively Informed Triangulation
Geertz himself noted that our knowledge of objective facts, which

might lead to causal theories, is compromised by subjective inter-
pretation. Yet others argue that we should not let this fact lead us to
abandon our efforts to approximate objective knowledge (Shankman

1984b; Farrer 1984). We should consider, then, what techniques can
help us manage or control for such subjective understandings.
The example of Westside residents’ opinions about which shooter
was the instigator illustrates how answers to factual questions may
inform an interpretivist understanding of those residents’ worldviews.
Alternatively, those worldviews also serve as another form of positivist
bias, albeit one that is particularly difficult to determine without a
simultaneous interpretivist analysis. Using this same example, respon-
dents who believed strongly that insiders were less culpable than
outsiders were likely to bias not only their opinions about who was the
leader but also factual information that portrayed the insider boy in a
more positive light. This may occur unbeknownst to the respondent,
through psychological processes whereby the respondent believes a
certain objective truth because they are in some way invested in that
version of events. In the passage from the Westside teacher quoted
above, the belief that no Westside insider would want to harm the
community because Westside is the kind of place where everyone
helps and cares about one another is precisely the sort of belief system
that this respondent goes to great lengths to defend. From a posi-
tivist perspective, this “worldview bias” is another way that respon-
dents may distort information but one that will not be rectified by the
strategies suggested above. A new strategy—interpretively informed
triangulation—is needed.
Interpretively informed triangulation builds on contextually infor-
med triangulation yet adds to it an explicit effort to try to understand
respondents’ (socially influenced) worldviews and how these world-
views influence their responses to data questions that seek objective
truth. A respondent’s understanding of her world and culture is a
fourth and, for our purposes, most illuminating form of bias that is
not captured by our previous categories of memory, vested interests,
or mistaken judgments.21 This understanding is a lens through which
respondents perceive the world, and they may go to great lengths
to uphold interpretations of events that support this belief system.
Triangulation must therefore be shaped to take this worldview bias
into account, much like the other forms of bias discussed. This strategy
is compared with contextually informed triangulation and traditional
triangulation in Table 3.
TABLE 3: Three Forms of Triangulation
Traditional Contextually Informed Interpretively Informed

Triangulation Triangulation Triangulation
Description Relies on multiple Weighs information Weighs information

sources of data or collected from different collected from different
multiple accounts by sources and individuals individuals by knowledge
different individuals to by knowledge of the of respondents’
give greater credibility information’s context, worldviews or value
to a version of events. its reliability, and systems and how they
potential biases. may shape respondents’
interpretations.
Method Combines information Tailors the collection Stratifies respondents

from as many different of supplementary data based on cultural
sources of information to address suspected perspective or social
as can be obtained; biases; uses this positions derived from
typically privileges the information about the specific social cleavages
side with “more” nature of the data in that interpretive analysis
evidence. weighing of evidence. reveals.
Example: Asks all respondents Weighs information Seeks responses from both
Which for information about from respondents based “insiders” and “outsiders,”
Jones- which boy was the on their knowledge of after this is revealed to
boro instigator and favors the shooters, their be an important dividing
shooter the account with more social position and line in the community that
was the support. Seeks external vested interests in a influences respondents’
leader of sources of data (e.g., particular interpretation, interpretations of events;
the pair? police reports, school how they learned the weighs responses based on
records, the shooters’ information they report knowledge of respondents’
writings) to give an about the shooters and social positions or their
indication of which how reliable it is or how cultural understanding of
individual had greater much it conforms to the positions of insiders
leadership qualities, media reports or rumor, versus outsiders;
history of deviance, or and their stake in the specifically, factual
motivation and uses interview situation. For information from insiders
this information to example, information suggesting the outsider
increase confidence in that Andrew was the shooter was the instigator
an interpretation. leader that comes from would be held to higher
members of Mitchell’s standards of confirmation
church is viewed as from noninsider sources.
potentially subject to a
vested interests bias,
and information to
confirm or deny is
sought from others with
a different relationship
to the shooters.
Interpretively informed triangulation may take the form of explicitly

stratifying the sample into groups of people who possess different
worldviews or it may mean considering interpretive biases in the
weighing of evidence. At its best, the process will be iterative, moving
back and forth between decisions of whom to interview and the emerg-
ing interpretive analysis. A self-aware researcher will have sought out
a cross section of divergent views in her initial list of interviewees, yet
the emerging understanding of the culture of the community allows her
to redefine the categories that comprise the relevant viewpoints. Our
sample might, for example, have to be explicitly stratified on the basis
of insiders and outsiders to gain a wide range of perspectives. Fur-
thermore, recognizing that this is an important (socially constructed)
dividing line within the community enables the researcher to more
easily understand future responses that are filtered through this lens.
A detailed example illustrates both how interpretively informed
triangulation functions and how it builds on the process of contextually
informed triangulation and everything learned from our efforts to
contextualize responses. Consider the contentious issue of whether
Michael Carneal was bullied. This proved to be a model example of
a contested event, with different accounts coming from all sides. The
media provided extensive reports that Carneal was bullied. Community
members reaffirmed these assumptions in our initial interviews, as
those who were at some distance from the case concluded on the basis
of the media reports and Carneal’s small physical stature that he must
have been bullied. Yet when we began to talk with other students in
Carneal’s class, many reported that Carneal picked on other kids as
much as he was picked on himself. We spoke with teachers, thinking
that perhaps their greater age and maturity would give them needed
perspective. However, teachers were largely unaware of the pervasive-
ness and seriousness of bullying in general and were unaware of serious
bullying in Carneal’s case. Finally, we obtained access to Michael
Carneal’s psychological reports, in which Carneal explicitly detailed
harassment going back to elementary school, which had intensified in
the year leading up to the shooting. However, the reports also stipulated
that Carneal suffered from a “schizotypal personality disorder with . . .
paranoid features,” which meant that he sometimes saw things that
were not there, such as snakes coming through the vents in his house.
What were we to conclude?
Our initial approach was to contextualize the information we recei-

ved by determining our various respondents’ social positions, vantage
points, and firsthand knowledge of events. We asked our informants
about their personal knowledge of Carneal, the source of their informa-
tion about him, and their own position within the social structure of the
school. Using contextually informed triangulation, we assigned lesser
weight to the reports of community members on the basis that they
were the ones who were farthest from the situation and had the least
specific knowledge. The teachers, we concluded, were quite unaware
of the high prevalence of bullying in general reported by students
and thus were also unlikely to be good sources of information about
Carneal specifically.
Next, we considered the vantage points of the students who told us
that Carneal was not bullied and how this might affect their accounts.
Were these more popular kids, less able to see the corrosive effects
of bullying and teasing? Using our contextual knowledge of students’
places in the social hierarchy, we endeavored to stratify student respon-
dents with respect to their position. We sought out other students who
were more likely to have been teased themselves, but these students,
perhaps by virtue of having come through the gauntlet of bullying
themselves, were if anything even more dismissive of the notion, argu-
ing that lots of kids were picked on, and Carneal’s torment was nothing
special. We considered whether there was variation in the assessments
of bullying from students who sat in different “seats,” which might,
consciously or unconsciously, affect their view of bullying. But as
we talked to boys and girls, 9th graders and 12th graders, weaklings
and athletes, those who knew him well and those who knew him less
well, what stood out was the constancy of the reports—bullying was
overblown, everyone deals with it, why couldn’t he?
It seemed that the most important divide was between all of the
other students on one hand and Carneal on the other. As we examined
the transcripts more carefully, we realized that these two groups agreed
on the facts—that he was called gay in a school newsletter, was rou-
tinely called “faggot” and “four eyes,” and was occasionally physically
tormented by bigger boys—but that they ascribed different meanings
to these facts. Other students did not see these slights as unusual or
excessive, comparing them instead to the ones that they and other
students had experienced.Yet Carneal, already psychologically fragile
and socially marginal, did not contextualize these perceived affronts

in a similar way. His reaction was to internalize them until they built
up and became one justification for the shooting. Thus, for those who
were equally likely to know, our analysis needed to be informed by an
interpretivist consideration of how Michael and others understood the
events.
Interpretively informed triangulation here involved a balancing of
accounts from these different perspectives, enabling us to call into
question the concept of bullying as an objective phenomenon. What
started as a factual research question resulted in a deeper understanding
of events and their interpretations, leading to the need to ask both
factual and subjective questions. By employing contextually informed
triangulation and interpretively informed triangulation together, we
develop both a factual conclusion in the positivist framework (the
actual events that took place, how Michael was teased, etc.) and multi-
ple subjective conclusions in the interpretivist framework (how
Michael and others made sense of these events).
Interpretively Informed Causal Theorizing
Perhaps the most important contribution that interpretivist analysis can

bring to a study of the causes of events like school shootings is the
recognition that locally available cultural frames are not only relevant
for issues of interpretation after the fact by all members of a community
but also shape the lived experience of those responsible for the event.
This recognition allows us to explore how the shooters understood
their place in their school or broader community and why going on
a rampage killing seemed like the best solution to the problems that
they perceived.
We illustrate how interpretivist analysis can be central to under-
standing a shooter’s motivation, using the example we have traced
throughout this article—the valuation of generational status. This point
is best illustrated with evidence about Michael Carneal. Evidence from
Carneal’s psychological reports, conducted after the shooting, as well
as an interview with his psychologist shed light on how this cultural
stratification in his community may have shaped his interpretation
of the bullying and social marginality that other students thought
were par for the course and why he believed that he had no other
options to his dilemma, such as telling his parents and getting help.
In a community that prized belonging and insider membership very
highly, Michael was at the fringes of every group he tried to join (see
Harding, Mehta, et al. 2002 for details; Newman et al. forthcoming).
He had attempted to find a niche for himself in several different social
groups, but each attempt was rebuffed. He even tried to buy his way
into membership in a social clique by giving gifts and other favors, with
limited success. Michael had difficulty forming close relationships and
as a result had very contingent social ties when everyone else around
him had incredibly strong ties. To make matters worse, Michael came
from a family of insiders. By right, he should have belonged in this
community, and the constant reminder in every interaction that he was
not accepted despite his family’s position and his own generational
status made his quest for acceptance close to an obsession. Several
psychological reports confirm that Michael saw the shooting as a way
of impressing people and gaining acceptance and admission into a
social group at school.
Michael also believed that telling his parents about his fears, the
psychological reactions that he recognized as abnormal, or the extre-
me anguish he suffered from his treatment at school was not an
option. Michael’s psychologist from the period after the shooting
explained how Michael had recognized he was having problems but
tried to hide them because of his concern for his family’s status in the
community.
Respondent: One of the reasons he didn’t ask for help was the fact that
people thought that they were this perfect family, and if he let
people know that he was struggling and needed help, then that
would blow that away.
Interviewer: I’d like to hear more about that . . . that there’s this public
perfect family, but inside . . .
Respondent: And I don’t think [there was] a lot of turmoil going [on] inside
the family.
Interviewer: Just that Michael is not perfect?
Respondent: Yeah, and Michael was very aware that he wasn’t perfect. But
I think he didn’t want other people to know he was messing
up that perfection.
Michael did not alert others to the fact that something was wrong
because he felt compelled not to ruin the myth of his idyllic family in
this idyllic community. The family’s position at the top of the hierarchy
based on generational status was a large part of that myth.
Without an understanding of the importance of maintaining one’s
reputation within the community, particularly among longtime resi-
dents, it would be very difficult to make sense of why admitting his
problems to his mother, with whom he was very close, did not seem to
be an option. This is simply one example of a more general point, that if
the goal is to understand social action, particularly action that is heav-
ily dependent on individuals’ subjective perceptions, an interpretivist
understanding of these social meanings is central to any understanding
of a causal theory or process. Locating the school shooter in his cultural
or symbolic milieu helps us understand the positivist causes of the
event as well as their interpretivist meaning.
CONCLUSION
The process by which positivist and interpretivist analysis inform one

another is iterative. An understanding of the local webs of signifi-
cance alters how positivist research is conducted; yet the positivist
inquiry can also show us where those webs of significance lie. The
value of moving interactively between these types of analysis, re-
defining research questions and design accordingly, cannot be under-
estimated.
Just as it raises questions about the nature of truth, the film Rashomon
illustrates the benefit of each approach to the idea of “truth.” The viewer
is bound to be left wondering which version of the terrible rape and
murder is “correct,” and the film encourages the viewer to ask this
question. Some argue that the director intends for one version of the
story to be taken as objectively true, suggesting that the film is not about
the relativity of truth at all but about the lies that individuals will tell
themselves to maintain their most important possession—their self-
image.22 Yet had Rashomon been presented as an ordinary whodunnit
tale, it probably would not have ascended to the classic status it now
holds. Instead, it reveals to us that we can know both things: who killed
the nobleman and why these characters believe what they do about this
event. It gives us a glimpse into the culture of this time and place and
how the ideals the characters present in their different accounts of

themselves—shame, honor, masculine bravado, feminine virtue—are
valued in their world, even as it establishes an objective truth of
the events. Here, as in social research about contemporary contested
events, both kinds of questions can be asked, and we benefit more from
knowing both kinds of answers.
NOTES
1. Fabian (1984) pointed out that Geertz is often incorrectly identified as the founding father
of interpretivism. The development of this approach should be traced back at least as far as Franz
Boas.
2. While Geertz (1973) was insistent that an interpretivist analysis should apply only to
its own time and place, many sociological studies tend to choose their cases as microcosms of
broader phenomena. In Rieder’s (1985) case, the broader phenomena was the turn of “Middle
America” against the Democratic party, and while he notes that there are lots of subpieces of
Middle America, his work is in part intended to provide an explanation of this large shift.
3. For example, Lin (1998), a political scientist who explicitly calls for an integration of pos-
itivist and interpretivist approaches, adopted a different understanding of these two frameworks.
While she defined positivist analysis as determining a causal relationship that can be tested or
identified in other cases, she viewed interpretivist analysis as producing “detailed examinations
of the causal mechanisms in the specific case” (p. 163). Using her example, positivist analysis
might be used to discover a positive relationship between previous work experience and current
employment, while interpretivist analysis would be used to discover the causal mechanism in
this relationship—that employers believe those with previous experience will be better workers.
While determining the causal mechanism can indeed be an important aim of qualitative research,
Lin’s concern with how to explain a causal association between two factors need not delve into
the webs of significance that are central to Geertz’s definition of interpretivism. The example she
selected illustrates this point; employers’ logic that applicants with previous work experience
would make better employees could be based on little more than the latest industry reports
about worker efficiency. No deeper cultural perspective is necessarily implied. Research of this
kind that focuses on establishing generalizable causal arguments is still basically positivist;
incorporating mechanisms into the causal theorizing makes for better research (Hedstrom and
Sweberg 1998) but does not necessarily make it any more interpretivist.
4. Many writers in the interpretivist tradition in sociology implicitly take this position
and treat objective facts and events as potential influences on interpretive understanding (e.g.,
Hochschild 1997, 1989; Newman 1988, 1993; Rieder 1985).
5. We do not seek to contribute here to the extensive debate about the meaning of the word
culture or the role that culture plays in social action. See Kuper (1999) for a good introductory
overview of theorizing on culture over the past century. We use culture, subjective understand-
ings, and social meaning interchangeably as a way of talking about how our respondents made
sense of the social lives of their communities.
6. In this article, we focus on methodological problems confronting positivist analysis rather
than the analysis itself, which may be found in Newman et al. (forthcoming).
7. This is not an exhaustive list. However, it does represent the three most common sets of
problems in our fieldwork.
8. This is particularly the case in distinguishing whether a response that is influenced by

one’s vested interests is conscious or unconscious.
9. Social psychologists have identified a variety of specific heuristics (Fiske and Taylor
1984). Our purpose here is not to explore how each of these heuristics may have influenced our
respondents’ testimony but to address the general problem of data inaccuracies due to mistaken
judgment, through the use of heuristics or otherwise.
10. We believe that this example also displays elements of a “vested interests” problem, but
one that is related more to psychological than material interests—in this case, a psychological
desire to be an important source of information about a central event in the town. This example
illustrates how several biasing influences can be present simultaneously in a given response.
11. In a small community, it may be possible to determine which respondents have spoken
with each other about these issues and from whom they may have heard the accounts they
relate. In theory, it would even be possible to conduct a network analysis of which people in the
community have talked with one another about the event (or about the interview itself) and how
this affected their responses.
12. This “adult” is not the interviewer but was present at the interview because the primary
respondent was a minor.
13. The respondent’s assumptions about what the interviewer wanted to hear may have been
influenced by the context of students’ previous experience being interviewed by the media. After
the shooting, numerous media agencies flocked to Jonesboro and aggressively pursued interviews
with students who knew the shooters (Freedom Forum 1998). Students were even known to
compete over the number of interviews they could conduct. Emphasizing his involvement
with and knowledge of one of the shooters therefore allows this student to demonstrate his
authoritative position to the interviewer.
14. The comparison to journalism is instructive because the newspaper’s job is not to discover
truth but rather to present the best truth it can find in limited time, in a way that is responsible
enough to protect the paper from libel lawsuits.
15. Of course, the new data might have different sources of bias and could be discerned
through the process of contextualizing responses, which would in turn direct further data
collection.
16. These experiments were usually performed by Garfinkel’s students on their friends, family
members, and spouses, without the knowledge of the subjects. A typical exchange taken from
Garfinkel (1967: 44) follows, beginning with a friend waving a cheerful hello to the experimenter:
(S) How are you?
(E) How am I in regard to what? My health, my finances, my school work, my peace of
mind . . . ?”
(S) (Red in the face and suddenly out of control.) Look! I was just trying to be polite.
Frankly, I don’t give a damn how you are.
As with much of Garfinkel’s work, the violation of the common understanding of a simple
question lends insight to the importance of unvoiced social conventions.
17. The Jonesboro Sun claimed that it was not focusing attention more on any one party
than another but was simply trying to give thorough coverage to what it saw as the important
stories. Another explanation for any difference in media portrayals of the two shooters may be
that the media focused more on Mitchell Johnson because they had access to more information
about him and because his family was more cooperative. Little information was available about
Andrew Golden because his family refused to talk to the media.
18. An interpretivist analysis can also focus on providing a meaningful account of incon-
sistencies within a single respondent’s worldview (Hochschild 1989), but in this case we are
more interested in providing a coherent account of how a variety of actors understood a single
contested event.
19. In this case, it is not the respondents’ views about which boy was the leader in the duo that
is taken as a factual query (as our discussion of contextually informed triangulation should make
clear) but rather the specific details they could provide about their experiences with each boy or
their firsthand knowledge of the shooters’ past actions that would contribute to an understanding
of what role each played in the shooting.
20. Good interpretivist researchers might ask such factual or causal questions without con-
sciously intending to combine these two analytical approaches. When this occurs, interpretivist
research is implicitly combined with positivist analysis to aid interpretivist goals. Increased
awareness of this process, as well as the benefits to incorporating a positivist approach, should
improve the quality of interpretivist work.
21. This “worldview bias” does have some similarities to the problem of heuristics, discussed
above. The salient difference is that heuristics are more focused on individuals’ limitations in
making accurate inferences, whereas here we are more concerned with how collectively held
understandings can systematically affect inferences.
22. Solomon (1981) argued that the final account of the rape and murder, related by the person
who was not a party to the crime but simply an observer, is meant to indicate the objective truth in
the director’s view. This is indicated by his use of sophisticated film techniques that distinguish
this account from the other versions of the events.
REFERENCES
Angus, Lawrence B. 1986. “Developments in Ethnographic Research in Education: From

Interpretive to Critical Ethnography.” Journal of Research and Development in Education
20(1):59-67.
Bazerman, Max. 1998. Judgment in Managerial Decision Making. New York: John Wiley.
Denzin, Norman K. and Yvonna S. Lincoln, eds. 2000. Handbook of Qualitative Research. 2d
ed. Thousand Oaks, CA: Sage.
Fabian, Johannes. 1984. “Comment.” Current Anthropology 25(3):273.
Farrer, Claire R. 1984. “Comment.” Current Anthropology 25(3):273-74.
Feldman, Martha S. 1995. Strategies for Interpreting Qualitative Data. Thousand Oaks, CA:
Sage.
Fiske, Susan T. and Shelley E. Taylor. 1984. Social Cognition. Reading, MA: Addison-Wesley.
Fox, Cybelle, Wendy D. Roth, and Katherine S. Newman. 2002. “A Deadly Partnership: Lethal
Violence in an Arkansas Middle School.” Pp. 91-116 in Deadly Lessons: Understanding
Lethal School Violence, Case Studies of School Violence Committee, edited by Mark H.
Moore, Carol V. Petrie, Anthony A. Braga, and Brenda L. McLaughlin. Washington, DC:
National Academy Press.
Freedom Forum. 1998. “Jonesboro: Were the Media Fair?” The Freedom Forum Report No. 98-
W08, Freedom Forum, Arlington, VA.
Garfinkel, Harold. 1967. Studies in Ethnomethodology. Englewood Cliffs, NJ: Prentice Hall.
Geertz, Clifford. 1973. “Thick Description: Toward an Interpretive Theory of Culture.” Pp. 3-30
in The Interpretation of Cultures: Selected Essays by Clifford Geertz. New York: Basic
Books.
Gieryn, Thomas F. 1983. “Boundary-Work and the Demarcation of Science From Non-Science:
Strains and Interests in Professional Ideologies of Scientists.” American Sociological
Review 48:781-95.
Harding, David J., Cybelle Fox, and Jal D. Mehta. 2002. “Studying Rare Events Through
Qualitative Case Studies: Lessons From a Study of Rampage School Shootings.”
Sociological Methods & Research 31(2):174-217.
Harding, David, Jal Mehta, and Katherine S. Newman. 2002. “No Exit: Mental Illness,
Marginality, and School Violence in West Paducah, Kentucky.” Pp. 117-42 in Deadly
Lessons: Understanding Lethal School Violence, Case Studies of School Violence
Committee, edited by Mark H. Moore, Carol V. Petrie, Anthony A. Braga, and Brenda L.
McLaughlin. Washington, DC: National Academy Press.
Hedstrom, Peter and Richard Swedberg. 1998. Social Mechanisms: An Analytical Approach to
Social Theory. Cambridge, UK: Cambridge University Press.
Hochschild, Arlie Russell. 1989. The Second Shift. New York: Avon Books.
———. 1997. The Time Bind: When Work Becomes Home and Home Becomes Work. New York:
Metropolitan.
Jick, Todd D. 1979. “Mixing Qualitative and Quantitative Methods: Triangulation in Action.”
Administrative Science Quarterly 24:602-11.
King, Gary, Robert O. Keohane, and Sidney Verba. 1994. Designing Social Inquiry: Scientific
Inference in Qualitative Research. Princeton, NJ: Princeton University Press.
———. 1995. “The Importance of Research Design in Political Science.” American Political
Science Review 89(2):475-81.
Kirk, Jerome and Marc L. Miller. 1986. Reliability and Validity in Qualitative Research.
Beverly Hills, CA: Sage.
Kuper, Adam. 1999. Culture: The Anthropologists’ Account. Cambridge, MA: Harvard
University Press.
Lamont, Michele. 1992. Money, Morals, and Manners: The Culture of the French and American
Upper-Middle Class. Chicago: University of Chicago Press.
———. 2000. The Dignity of the Working Men: Morality and the Boundaries of Race, Class
and Immigration. Cambridge, MA: Harvard University Press.
Lin, Ann Chih. 1998. “Bridging Positivist and Interpretivist Approaches to Qualitative
Methods.” Policy Studies Journal 26(1):162-80.
MacKay, Robert. 1978. “How Teachers Know: A Case of Epistemological Conflict.” Sociology
of Education 51:177-87.
Marshall, Gordon, ed. 1994. The Concise Oxford Dictionary of Sociology. Oxford, UK: Oxford
University Press.
Miles, Matthew B. and A. Michael Huberman. 1994. Qualitative Data Analysis: An Expanded
Sourcebook. 2d ed. Thousand Oaks, CA: Sage.
National Research Council and Institute of Medicine. 2002. Deadly Lessons: Understanding
Lethal School Violence, Case Studies of School Violence Committee, edited by Mark H.
Moore, Carol V. Petrie, Anthony A. Braga, and Brenda L. McLaughlin. Washington, DC:
National Academy Press.
Newman, Katherine S. 1988. Falling From Grace: The Experience of Downward Mobility in
the American Middle Class. New York: Free Press.
———. 1993. Declining Fortunes: The Withering of the American Dream. New York: Basic
Books.
Newman, Katherine S., Cybelle Fox, David Harding, Jal Mehta, and Wendy D. Roth.
Forthcoming. The School Shooter: The Causes and Consequences of Rampage Killings in
American Schools. New York: Basic Books.
Phillips, Bernard. 1976. Social Research: Strategy and Tactics. 3d ed. New York: Macmillan.
Rappaport, Roy A. 1979. Ecology, Meaning and Religion. Richmond, VA: North Atlantic.
Rieder, Jonathan. 1985. Canarsie: The Jews and Italians of Brooklyn against Liberalism.
Cambridge, MA: Harvard University Press.
Rosaldo, Renato. 1982. “Review of Geertz and Culture,” by Kenneth A. Rice. American
Ethnologist 9:197-98.
Shankman, Paul. 1984a. “The Thick and the Thin: On the Interpretive Theoretical Program of
Clifford Geertz.” Current Anthropology 25(3):261-80.
———. 1984b. “Reply.” Current Anthropology 25(3):276-78.
Solomon, Stanley J. 1981. “The Film Idea.” Reprinted from Toronto Film Society program
notes, March 2. Available at http://filmsociety.wellington.net.nz/FilmIdea.html.
Tversky, Amos and Daniel Kahneman. 1974. “Judgment Under Uncertainty: Heuristics and
Biases.” Science 185:1124-31.
Yin, Robert K. 1989. Case Study Research: Design and Methods. Newbury Park, CA: Sage.
Zetterberg, Hans L. 1966. On Theory and Verification in Sociology. 3d enlarged ed. Totowa,
NJ: Bedminster.
Wendy D. Roth is a Ph.D. candidate in sociology and social policy at Harvard University
and a doctoral fellow of the Harvard Multidisciplinary Program in Inequality and Social
Policy. She has formerly worked at the National Centre for Social Research in London on
youth offending and welfare-to-work policy evaluations. Her current research focuses
on the racialization of Latino immigrants and the socialization of multiracial children.
Jal D. Mehta is a Ph.D. candidate in sociology and social policy at Harvard University
and a doctoral fellow in the Harvard Multidisciplinary Program in Inequality and
Social Policy. A National Science Foundation graduate fellow, his previous work focused
on the role of social-psychological mechanisms in perpetuating social stratification
and sponsoring social mobility and on racial differences in achievement. His research
interests include combining normative political theory on justice with empirical research
on poverty and inequality.

Roth y Metha 2002

Uploaded by

Copyright:

Available Formats

Roth y Metha 2002

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Roth y Metha 2002

Uploaded by

Copyright:

Available Formats

Positivist and interpretivist analytical approaches are frequently believed to be incom-

The Rashomon Effect

WENDY D. ROTH and JAL D. MEHTA

Akira Kurosawa’s 1950 ﬁlm Rashomon presents four different accounts

SOCIOLOGICAL METHODS & RESEARCH, Vol. 31 No. 2, November 2002 131-173

two approaches to understanding the events that took place outside of

POSITIVIST AND INTERPRETIVIST APPROACHES

The expressions positivism and interpretivism are used to varying de-

The interpretivist approach has received more varied treatment by

To make sense of middle-class discontent, writers have invoked white

Rather than attempting to adjudicate between these competing hypo-

TABLE 1: Summary of Positivist and Interpretivist Approaches

Causation—Seeks to understand the causal Interpretation—Seeks to understand how

seeks causal laws to explain objectively viewed phenomena, whether

COMBINING POSITIVIST AND INTERPRETIVIST APPROACHES

Positivist and interpretivist approaches are sometimes posed as diamet-

objective reality, yet argues that complete rejection of such knowledge

I believe that complete objectivity is impossible; but this does not

Our approach to combining positivist and interpretivist analy-

1. Using a positivist approach to address questions of causation. What

The remainder of this article illustrates these three objectives, look-

FIELDWORK ON TWO SCHOOL SHOOTINGS

We conducted in-depth qualitative case studies of the 1997 school

depression and the beginning stages of schizophrenia, complained that

POSITIVIST ANALYSIS OF CONTESTED EVENTS

There are a number of methodological difﬁculties related to trying to

CONCERNS OVER DATA QUALITY IN CONTESTED EVENTS

We identiﬁed three categories of problems that plagued our attempts

Perhaps the most obvious challenge to understanding an event that

problem of inadequate or inconsistent memory. Many respondents

Respondent: [After the shooting,] I would not sleep in my room. I would

lay in my bed. See, my ﬁrst time laying in bed, oh my gosh, I

Even when respondents could accurately remember past events, some-

Here, the interviewer’s attempts to discuss ways that schools could

MANAGING DATA BIAS IN POSITIVIST RESEARCH

The process of contextualizing responses is that of framing the infor-

TABLE 2: Biases in the Data and Suggested Approaches in a Positivist Framework

Because this account was so widespread among the student body

Understanding that this respondent’s perspective is shaped by his social

While we should not automatically conclude that this account is inaccu-

Contextually Informed Triangulation

The best way to avoid being misled by a single inaccurate or biased

Contextually informed triangulation takes this understanding into

INTERPRETIVIST ANALYSIS OF CONTESTED EVENTS

While hypothesis testing and causal theorizing provide a speciﬁed

THE CONTESTED EVENT AS A CULTURAL BREACH

An atypical event that is controversial can serve as a breach of the social

work illustrated the value of breaches in identifying the norms and

“Insiders” with multigenerational family histories had higher social

The extent of the “preferential” treatment they received is in dispute,

of their higher status in the community. In a newspaper article, she

Thus, respondents could vocalize this wide-reaching stratiﬁcation

POSITIVIST AND INTERPRETIVIST ANALYSIS

Up to this point, we have considered positivist and interpretivist anal-

USING POSITIVIST ANALYSIS IN THE AID OF INTERPRETIVISM

All interpretivist analysis is informed to some degree by relevant fac-

As this teacher and member of the Westside “family” tries to make

USING INTERPRETIVIST ANALYSIS IN THE GENERATION

Interpretivist analysis can provide crucial information to inform posi-