Examining the Iatrogenic Effects of the Cambridge-Somerville Youth Study: Existing Explanations and New Appraisals

Abstract

Criminology has paid increasing attention to the prospect that prevention programmes can cause harm. The Cambridge-Somerville
Youth Study, a delinquency prevention experiment of 506 boys that began in 1939, provides some of the earliest evidence of
programmatic iatrogenic effects. A series of hypotheses were advanced by Joan McCord and other scholars to explain these unintended
effects. Drawing upon this scholarship, related research and developmental theory, this article examines the leading explanations
and offers new appraisals of iatrogenic effects of crime prevention programmes. The research suggests that there is not a
grand explanation, and we encourage a more nuanced perspective for understanding iatrogenic effects of crime prevention programmes.
Implications for policy and practice are discussed.

Full text

© The Author 2015. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD).
All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
EXAMINING THE IATROGENIC EFFECTS OF THE CAMBRIDGE-
SOMERVILLE YOUTH STUDY: EXISTING EXPLANATIONS AND
NEW APPRAISALS
S N.Z, B C.W*  G M.Z
Criminology has paid increasing attention to the prospect that prevention programmes can cause
harm. The Cambridge-Somerville Youth Study, a delinquency prevention experiment of 506 boys
that began in 1939, provides some of the earliest evidence of programmatic iatrogenic effects.
Aseries of hypotheses were advanced by Joan McCord and other scholars to explain these unin-
tended effects. Drawing upon this scholarship, related research and developmental theory, this
article examines the leading explanations and offers new appraisals of iatrogenic effects of crime
prevention programmes. The research suggests that there is not a grand explanation, and we
encourage a more nuanced perspective for understanding iatrogenic effects of crime prevention
programmes. Implications for policy and practice are discussed.
Keywords: Cambridge-Somerville Youth Study, delinquency prevention, iatrogenic
effects, developmental theory
Introduction
The foundation of this article is rmly grounded in the premise that interventions
attempting ‘to modify human behavior … have the power to bring about unintended,
harmful consequences’ (Rhule 2005: 621). This knowledge has the ability to move us
beyond the ‘effective’ versus ‘not effective’ distinction in the ‘what works’ literature, and
toward a third possibility—that a prevention programme may cause harm. While many
scholars and practitioners have observed that crime prevention programmes must rst
‘do no harm’ (MacKenzie 2013), it is unlikely that this Hippocratic Oath can be upheld
without an understanding of why some programmes causeharm.
In the spirit of this view, this article sets out to examine the leading explanations and
offer new appraisals of iatrogenic effects of crime prevention programmes. We seek to
accomplish this with reference to Joan McCord’s work on the Cambridge-Somerville
Youth Study (heretofore the CSYS), a longitudinal-experimental study that began in
1939 in neighbouring towns of Boston, Massachusetts. Situating an examination of
iatrogenic effects with reference to the CSYS is appropriate, both because of the long-
term prospective, longitudinal-experimental design of the study and the empirically
examined explanations for the study’s iatrogenic effects.
McCord’s inuential article on the CSYS, published in 1978, reported on the results
of the 30-year follow-up (mean age=45years) of the delinquency prevention experi-
ment. Findings indicated that the programme produced harmful effects. Her research
Steven N.Zane, Sc hool of Crim inology and C rimina l Justice, Nort heastern Un iversity, Chur chill Hal l, 360 Hunti ngton Avenue,
Boston, M A 02115, USA; *Brandon C.Welsh, School of Criminolog y and Criminal Justice, Northeastern University, Churchill
Hall, 360 Huntington Avenue, Boston, M A 02115, USA; Netherlands Institute for the Study of Crime and Law Enforcement, De
Boelelaan 1077a, 1081 HV A msterdam, the Netherlands; b.welsh@neu.edu; Gregory M.Zimmerman, School of Criminology
and Criminal Justice, Northeastern University, Churchill Hall, 360 Huntington Avenue, Boston, MA 02115, USA.
doi:10.1093/bjc/azv 033 BRIT. J. CRIMINOL. (2016) 56, 141–160
Advance Access publication 14 June 2015
141
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

was comprehensive, investigating and detecting iatrogenic effects for a wide range of
important life-course outcomes, including criminal offending, physical and mental
health, family stability, employment and alcohol abuse. Of the 15 criminal outcome
comparisons between the treatment and control groups, none favoured the treatment
group and one favoured the control group; of the 15 health outcome comparisons,
none favoured the treatment group and 4 favoured the controls and of the 13 com-
parisons of family, work and leisure time, none favoured the treatment group and 2
favoured the controls (McCord 1978).
To compare outcomes across a common metric, McCord (1978) grouped subjects by
whether or not they experienced any ‘undesirable outcomes’, which included FBI index
crimes (all of which were felonies), treated alcoholism, serious mental illness and death.
The results indicated that 42 per cent of treatment group participants experienced
undesirable outcomes compared to 32 per cent of controls, a statistically signicant dif-
ference (p=0.02). McCord (1980) also discovered that the iatrogenic effects were more
pronounced when the intervention was more frequent, longer in duration and involved
more than one counsellor. While these differences were often small in magnitude, they
were all statistically signicant (McCord 1978).
Based on these ndings, McCord became a pioneering force for increasing attention
to, and disarming hostility toward, the possibility of harmful effects arising from crime
prevention programmes. The following best captures her position:
Researchers typically fail to consider whether social programs have adverse effects, looking only for
favorable results of treatment. . . . Yet providers of social services do not have a right to harm their
clients. Nor do most providers wish to do so. But the social climate that buries evidence of harm is powerful.
That social climate must be changed. (McCord 2003: 27–8, emphasis added)
Surprisingly, since McCord’s inuential article, there has been no published research
that has attempted to bring together the rich body of scholarship on the CSYS as well
as research on iatrogenic effects more generally. We are well aware of the concerns that
criminology is becoming far too concerned with present day events and is overlooking
the insight that can be gleaned from classic studies, including those with longitudinal
designs like the CSYS (see Laub 2004). It is our position that the CSYS still holds rel-
evance to understanding how and why crime prevention programmes causeharm.
We begin with a brief discussion of the background of the CSYS and McCord’s early
hypotheses for the observed iatrogenic effects of the delinquency prevention experi-
ment. This is followed by a broader discussion of the implications of McCord’s work for
iatrogenic effects. Specically, we highlight recent explanations for iatrogenic effects
of crime prevention programmes—most notably peer deviancy training, theory failure,
implementation failure and heterogeneous treatment effects. We conclude by discuss-
ing concrete implications for research, policy and practice.
Background
The CSYS began in 1939 at a time of tremendous innovation to prevent delinquency
outside of the formal justice system. The programme was developed by Richard
Clarke Cabot, a medical doctor and professor of social ethics and clinical medicine
at Harvard University (McCord 1992). Inspired by studies of juvenile delinquency
ZANE ETAL.
142
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

during the 1930s, including disconcerting recidivism results published by S. Glueck
and E.T. Glueck (1934), Cabot set out to study the development of criminal behav-
iour (McCord 1992). The CSYS was conceived as a treatment programme to prevent
delinquency (Powers and Witmer 1951), but also, in the spirit of Cabot’s medical back-
ground, to test the efcacy of the treatment by using a rigorous experimental design.
As noted by Cabot (1940: 143), ‘treatment and research objectives [were] closely
intertwined.’
The approach to the treatment itself was rmly grounded in developmental insights.
For Cabot, the family unit was the foundation of society, and social problems such as
juvenile delinquency invariably had roots in this foundation (Cabot 1940). The treat-
ment programme paired youth with adults who could provide positive inuences in
their lives and ‘supplement but not replace what would normally be a satisfactory par-
ent-child relationship’ (Cabot 1940: 143). The intent of the treatment was to ‘determine
whether boys would model themselves after a positive gure who was simply warmly
and humanly interested in them’ (Bergin 1963:245).
The study began with 650 boys, aged 5–13years (median age=10.5years), from the
two Boston area towns of Cambridge and Somerville. To prevent stigmatization—a
chief concern when designing the intervention—both youth with and without behav-
ioural problems (rated as ‘difcult’ and ‘average’, respectively) were recommended by
local schools, welfare agencies, churches and police (McCord 1978). Following physical
examinations, interviews (also of family members) and detailed case histories, the boys
were placed into 325 matched pairs and one member of each pair was chosen at ran-
dom (on the toss of a coin) to be in the treatment group (Powers and Witmer 1951: 7).
Boys were matched on a wide range of variables (e.g. grade placement, physical health,
neighbourhood, occupational status of father) with the aim of producing ‘diagnos-
tic twins’ (Cabot 1940: 146). Ultimately, the investigators were interested in achieving
greater comparability between the treatment and control groups—in addition to that
provided by random assignment.
From 12 November 1937 to 13 May 1939, treatment youth were assigned to counsel-
lors who would support character and personality development through mentoring and
provide a ‘friendly interest in the boys’ problems’ (Cabot 1940: 143). Nineteen counsel-
lors (15 men and 4 women) were employed by the study, and the average counsellor had
a caseload of 34 boys (Powers and Witmer 1951). While eight counsellors were social
workers, there were no training qualications beyond a ‘warm, outgoing’ disposition
(Powers and Witmer 1951:92).
The treatment group received individual counselling and home visits. Keeping in
close contact with the police, counsellors talked to the boys, took them on trips, engaged
them in recreational activities, tutored them in reading and arithmetic, encouraged
them to participate in the YMCA and in summer camps, played games with them at
the project’s centre, encouraged them to attend church and gave advice and general
support to the boys’ families (W. McCord and J.McCord 1959). While the treatment
generally followed Cabot’s vision of ‘directed friendship’, the manner in which this was
accomplished varied: ‘[e]ach counselor was left largely to his own resources’ in deter-
mining the treatment approach for a particular youth (Powers 1949: 82). The control
group received no services.
In 1942, shortly following the United States’ entry into World War II, the loss of
nancial resources (owing to gas shortages) as well as some of the young men who were
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
143
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

counsellors (owing to military service) caused the programme to be scaled back to 253
matched pairs of boys. The programme, ending on 31 December 1945 (Powers 1949),
lasted an average of 5years, during which time treatment youth received an average of
two visits per month (McCord 1978).
Follow-ups, carried out in 1948 (Powers and Witmer 1951) and 1956 (W. McCord
and J. McCord 1959), indicated that the programme had a null effect on ofcial
offending. The rst follow-up suggested that the treatment boys committed more
offences up to 1948, but the signicance of this result was not tested (Powers and
Witmer 1951: 326). The later follow-up suggested that the control boys committed
more crimes, but the difference was not statistically signicant (W. McCord and
J.McCord 1959:92).
The next follow-up was carried out by Joan McCord in 1978, 30years post-interven-
tion (mean age=45years). It is important to note that this follow-up was inuenced
by major prospective longitudinal studies of the day, including West and Farrington’s
(1977) Cambridge Study in Delinquent Development, which involves 411 London
boys born in the 1950s. The CSYS ts squarely in the global tradition of longitudinal
research on offending over the life-course (see also Moftt et al. 2001; Vitaro et al.
2013). On the matter of the 30-year follow-up, records were located for 94.9 per cent
or 480 of the 506 participants (of whom 48 had died) and interviews were conducted
with (or questionnaires distributed to) 347 of them. Comparisons between the treat-
ment and control groups indicated that the treatment group had not fared better
on any measured outcome, and actually fared worse on seven outcomes. Compared
to the controls, treatment group men were more likely to (1) commit more than one
crime (among those who committed at least one crime); (2) suffer symptoms of alco-
holism; (3) manifest signs of mental illness; (4) die at a younger age; (5) suffer from
at least one stress-related disorder, especially high-blood pressure or heart trouble;
(6) have occupations with lower prestige and (7) report their work as unsatisfying
(McCord 1978).
Various critiques of the study’s ndings have been advanced over the years. For
example, as Short (1954) suggested, it is possible that the observed iatrogenic effects
were artefacts of measurement bias. That is, measurement failure could have jeopard-
ized internal, construct or statistical conclusion validity, thereby biasing the strength
and signicance of estimated coefcients (see Welsh and Rocque 2014). Additionally,
Vosburgh and Alexander (1980) were concerned that the study involved implicit self-
selection in the treatment design. Counsellors made a series of treatment decisions
(such as the frequency of treatment and the services provided) after random assignment,
thereby threatening internal validity for certain comparisons. Weiss and colleagues
(2005) suggested that construct validity was problematic. Specically, they objected
to the inclusion of schizophrenia and bipolar disorder in the measure of undesirable
outcomes. While criminal offending and alcoholism may be valid measures of negative
outcomes affected by the CSYS, Weiss etal. (2005) argued that mental illness could not
be inuenced by a social intervention programme and therefore should not have been
included in the outcome construct.
McCord had her own ideas about why the programme produced harmful results,
proposing four hypotheses in her 1978 article. In later works (McCord 1980; 1981), she
investigated if the data supported any of these hypotheses. This is the focus of the next
section.
ZANE ETAL.
144
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

McCord’s Hypotheses
Value conict hypothesis
“Interaction with adults whose values are different from those of the family milieu may produce
later internal conicts that manifest themselves in disease and/or dissatisfaction.” (McCord 1978:
289)
The value conict hypothesis drew upon research on ‘hard core unemployables’.
Padeld and Williams (1973) suggested that when certain ‘unemployables’ were nally
employed, it exacerbated their problems rather than solving them because it increased
their conict with middle-class American values. Based on this research, McCord
(1981) suggested that the CSYS might have harmed participants by creating a conict
between the values of the counsellors and those of the participants. In this case, the
programme’s goal of character development would have been at odds with the values
in the treatment boys’ working-class neighbourhoods of Cambridge and Somerville,
making it difcult for the treatment group boys to incorporate middle-classvalues into
theirlives.
To assess this hypothesis, McCord (1981: 399)compared various measures of aca-
demic achievement, assuming that higher achievement would reect ‘middle-class
achievement values’. The only observed statistically signicant difference involved high
school graduation (the treatment group was less likely to graduate), and there were no
signicant differences in how often participants mentioned education when discuss-
ing their children. Men were also asked more generally about qualities they admired,
but again did not differ signicantly. Overall, McCord (1981: 400)concluded: ‘none of
the comparisons indicates that treatment had increased the achievement orientation
attributed to holding middle-classvalues.’
Dependency hypothesis
“Agency intervention may create dependency upon outside assistance. When this assistance is no
longer available, the individual may experience symptoms of dependency and resentment.” (McCord
1978: 289)
According to this hypothesis, the treatment group may have become dependent on the
programme, which resulted in maladjustment when they no longer had access to pro-
gramme resources. For example, the treatment group boys may have become attached
to the positive adult relationships provided by the programme and suffered when these
relationships ended at programme termination. In support of this hypothesis, some
have argued that welfare programmes may harm recipients by creating a dependency
on handouts and preventing the development of self-efcacy (K ane 1987; Kalil and
Kunz 1999).
To assess this hypothesis, McCord used measures of ‘dependency’ constructed from
questionnaires collected at the 30-year follow-up. These included how often men asked
others for help, how often men visited their parents, whether men were active in clubs
and how men spent their leisure time. No signicant differences across treatment and
control group men emerged for any of these measures.
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
145
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

Labelling hypothesis
“[T]hrough receiving the services of a ‘welfare project,’ those in the treatment program may have
justied the help they received by perceiving themselves as needing help.” (McCord 1978: 289)
McCord hypothesized that the CSYS harmed participants by creating a self-perception
that they ‘needed treatment’ because they had been selected for help. Placing youth
in the treatment group may have created a negative self-image that undermined any
positive effects of the treatment. This possibility—‘labelling’—was an explicit concern
of CSYS founders, who made it a point to include ‘average’ as well as ‘difcult’ boys in
the treatment and control groups (Cabot 1940). Despite the precautions taken, McCord
compared measures of self-condence, feelings of competence and psychosomatic dis-
orders across the treatment and control groups to assess the potential for labelling
effects. The two groups did not differ signicantly in responses, and McCord concluded
that the data did not support a labelling explanation (McCord 1981: 4 01).
Failed expectations hypothesis
“The treatment program may have generated such high expectations that subsequent experiences
tended to produce symptoms of deprivation.” (McCord 1978: 289)
Lastly, McCord (1981) reasoned that a social intervention could harm participants by
creating unrealistic expectations regarding life possibilities. Failed expectations, in
turn, would create a sense of disillusionment and diminish satisfaction with everyday
life events, thereby precipitating subsequent negative outcomes.
McCord investigated this hypothesis by comparing measures of life satisfaction.
Pertaining to marriage, McCord (1981: 401)found that a higher proportion of treat-
ment group men had been separated or divorced than control group men (p<0.05).
Additionally, for current marriages, only 47 per cent of treatment participants, com-
pared to 65 per cent of controls, were coded as exhibiting ‘warmth’ toward their spouses
during interviews, a statistically signicant difference. The treatment group was also
signicantly less likely to report being satised with work and with life. Ultimately,
McCord (1981: 405)concluded: ‘The Cambridge-Somerville Youth Study seems to have
raised the expectations for its clients without also providing the means of increasing
satisfactions. The resulting disillusionment seems to have contributed to the probability
of having an undesirable outcome.’ Despite this conclusion, relatively little theoretical
attention has been paid to whether failed expectations and subsequent disillusionment
explain why other crime prevention programmes cause harm (but see Fabricatore etal.
2007, in the context of obesity treatment).
Peer Deviancy Training
Contagion as an explanation of the CSYS’s iatrogenic effects
Almost two decades after she presented and investigated these initial hypotheses,
McCord arrived at a new explanation: the iatrogenic effects of the CSYS were due, in
part, to peer contagion effects among a subgroup of treatment youth (Dishion etal.
1999). Peer contagion involves a mutual-inuence relationship between peers, wherein
ZANE ETAL.
146
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

the behaviour and emotions of each peer inuences the other in ways that potentially
cause harm or otherwise undermine the developmental process (Dishion and Tipsord
2011). ‘Contagion’ is used to describe this process due to its resemblance to the conta-
gious transmission of disease (Loftin 1986).
Peer contagion can occur by a variety of causal mechanisms (Cécile and Born 2009).
In the context of the CSYS, McCord proposed the ‘peer deviancy training’ hypothesis
(Dishion etal. 1999; McCord 2003). This mechanism of peer contagion involves ‘the inter-
personal dynamic of mutual inuence during which youth respond positively to deviant
talk and behavior . . . characterized by give-and-take exchanges between friends that
promote deviant actions (e.g., past stories of deviant acts, suggestions for future behavior,
what ifs) and elicit positive responses, such as laughter’ (Dishion and Tipsord 2011: 190).
In other words, peers who communicate about deviancy—real or imagined—reinforce
antisocial behaviours and emotions and provide motivation for future deviant behaviour.
Notably, however, peer contagion is not only a matter of normative inuence but
also a function of opportunities for deviance in unstructured environments (see, e.g.
Osgood etal. 1996; Osgood and Bridell 2006). That is, socialization may interact with
opportunity via ‘unstructured socializing’ (Haynie and Osgood 2005). This idea com-
ports with recent work by Weerman and colleagues (2013), which suggests that peer
deviancy training may be more salient in conjunction with socializing (e.g. hanging
out), being in public (i.e. engaging in unstructured activity) and being unsupervised
(see also Anderson and Hughes 2009; Bagwell and Schmidt 2011).
To assess whether deviancy training could explain the iatrogenic effects of the CSYS,
McCord considered summer camp participation by the treatment group boys (Dishion
etal. 1999). Roughly half were sent to summer camp (125/253), and approximately half
of these boys went to camp more than once (66/125). In theory, these camps allowed
for a great deal of unstructured socializing, representing an ideal environment for devi-
ancy training to takeplace.
Re-analyzing the 30-year follow-up data, McCord (2003) found that for boys who
were sent to summer camp only once (n=59), the odds ratio predicting undesirable
outcomes was 1.33, which was signicantly higher than the odds ratio of 1.12 among
treatment group boys who did not attend summer camp. The truly stark result, how-
ever, involved treatment group boys who attended camp more than once (n=66). For
this group, the odds ratio for undesirable outcomes was 10.0, meaning that partici-
pants were ten times as likely to experience undesirable outcomes as their matched
pairs. McCord (2002: 235)concluded: ‘I strongly suspect that the boys from the Youth
Study tended to bond together, encouraging one another’s deviant values.’ McCord’s
construct theory of motivation—which argues that youth respond to situational cues
in constructing motivation based on the way they perceive the choices of others—pro-
vided a theoretical explanation for why deviancy training takes place among high-risk
youth in unsupervised settings (McCord 2003; 2004).
Since McCord’s research, a robust literature on peer deviancy training has emerged
(Poulin etal. 2001; Dishion 2000; Dishion and Dodge 2005; Dodge etal. 2006; Dishion
and Tipsord 2011). As a whole, the empirical literature suggests that unrestricted inter-
action between low- and high-risk youth contributes to higher rates of deviant behaviour
among low-risk youth (Haynie and Osgood 2005; Dishion and Tipsord 2011). These
effects can be observed as early as the rst grade of elementary school (Gifford-Smith
etal. 2005) and persist over time (Dishion and Tipsord 2011).
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
147
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

Several meta-analyses and systematic reviews also provide support for the peer con-
tagion hypothesis. For example, Lipsey and colleagues (2000) conducted a review of
juvenile delinquency intervention programmes and found that the most successful
programmes were those in which delinquent youth did not have the opportunity for
unstructured interaction with deviant peers. In addition, a recent review of (both juve-
nile and adult) programmes with iatrogenic effects indicated that almost three-quarters
(14/19) of programmes with harmful effects occurred in group settings, as compared
to individual settings (Welsh and Rocque 2014). Despite these ndings, whether devi-
ancy training actually undermines group interventions and causes harm remains the
subject of much controversy (Rhule 2005).
Challenges to establishing peer deviancy effects
Some argue that group-based interventions are a cost-effective alternative to individu-
alized treatment (Kaminer 2005; Greenwood 2006)and only cause harm under cer-
tain conditions, e.g., in the absence of parental involvement and adequate supervision
(Dishion et al. 1999; Kaminer 2005; Rhule 2005; Dishion and Tipsord 2011). Others
question the causal relationship between group treatment and iatrogenic effects (see
Weiss etal. 2005). The contention, which is the most notable objection to peer deviancy
effects, is that negative group treatment effects are not actually treatment effects but
rather represent selection effects under which deviant peers associate due to homoph-
ily (Arnold and Hughes 1999; Weiss etal. 2005). Failing to account for selection effects
may lead one to conclude that deviant peers are driving personal delinquency (i.e.
socialization), when in reality personal delinquency precedes deviant peer associations
(see Steglich etal. 2010). This is especially true if high-risk youth are selected for group
treatment, since these youth are likely to afliate with deviant peers both in and out of
the treatment context.
In the case of the CSYS, the selection of treatment group boys for summer camp
was based on counsellor discretion rather than on random assignment (Arnold and
Hughes 1999; Gottfredson 2010), making it impossible to distinguish selection from
socialization effects. To be fair, McCord appeared eminently aware of this poten-
tial objection: ‘One caution of [the peer-deviancy] interpretation is that youth self-
selected into summer camp experiences; because their matched controls did not
make a similar selection, the intervention group may be biased toward the deviance
in an unknown way’ (Gifford-Smith et al. 2005: 261). But, this quandary need not
plague all studies of group interventions, since research designs (e.g. focusing on
the transmission of specic types of crime; see Bayer etal. 2004; Mennis and Harris
2011) and statistical techniques (see Weerman 2011; Young 2011) can separate selec-
tion and socialization effects.
These arguments aside, even if peer aggregation does lead to iatrogenic effects in
the group treatment context, it may not follow that deviancy training is the responsible
causal mechanism (Weiss etal. 2005). Critics argue that while deviancy training may be
correct as a more general explanation for delinquency—that is, outside of treatment—
peer inuence during group treatment is minimized via restricted exposure time. This
leaves room for theorizing about other causal mechanisms responsible for iatrogenic
effects in group-based crime prevention programmes.
ZANE ETAL.
148
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

Theory, Implementation and Heterogeneous Treatment Effects
While the explanations for the CSYS ndings discussed above focus on aspects of the
treatment that may have had unintended, harmful consequences, we should not over-
look potential explanations that focus on the treatment programme itself: the underly-
ing programme theory (theory failure), the way the programme was designed (design
failure), and the way it was implemented (implementation failure). Each of these rep-
resents a commonly considered explanation for null effects, but they also provide an
important perspective from which we can interpret iatrogenic effects.
Theory failure
The rst step to planning an effective treatment programme involves a solid theoretical
foundation on which to base the treatment (Rossi etal. 2004). Most importantly, this
should include a theory of programme change: how will the treatment elicit change in
the treatment group? The importance of programme theory has led to developments
in the evaluation literature such as logic models, which describe how a programme
works (or is supposed to work) in terms of inputs, resources, activities, outputs, out-
comes and long-term impact (Mertens and Wilson 2012). In the context of crime pre-
vention, theory failure is likely responsible for the ineffectiveness of Scared Straight
programmes, boot camps and intensive supervision probation (Welsh and Rocque
2014; see also Sherman 2007).
In the case of the CSYS, the question is whether it was based on a coherent theoreti-
cal approach to preventing delinquency. And, if a sound theory was absent, what was
the implicit logic of the programme? As others have noted, interventions tend to have
some theoretical basis—whether it is articulated or not (Pawson and Tilley 1997; Mears
2007). For crime prevention efforts to be successful, it is necessary for this theoretical
basis to be rationally related to the development of criminal offending. It is thus impor-
tant to assess the rationale underlying the CSYS and examine whether partial theory
failure could be responsible for its long-term iatrogenic effects.
Friendly mentoring is not a therapeutic treatment
Cabot’s vision for the CSYS was to use friendly mentoring from positive role models to
supplement the parent-child relationship and provide positive guidance for character
and personality development (Cabot 1940). Cabot believed that ‘the thing that “keeps
any of us straight” is “the contagion of the highest personalities whom we have known”’,
and that providing an ‘ego-ideal’ to troubled youth would effectively prevent delin-
quency (Short 1954). While this may seem intuitively plausible, others have since sug-
gested that it does not present an effective therapeutic model to combat delinquency
(Tremblay 2005). As MacKenzie (2013: 3)observed, ‘since the McCord study, we have
learned a great deal about what is effective in reducing criminal activities of delin-
quents and offenders. “Friendly understanding” programmes do not address the crimi-
nogenic decits of these individuals and, thus, are doomed to failure.’
Elsewhere, the CSYS has been criticized for not presenting a clear therapy model.
Weiss et al. (2005: 1038)suggested that the CSYS treatment programme—consisting
mostly of case management and mentoring—would ‘not be considered a potentially
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
149
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

efcacious treatment’ by current standards in clinical psychology. Even in its day, it was
observed that the CSYS could not ‘be truly said to have been what most of us would
term psychotherapy’ (Bergin 1963: 245). In the end, friendly mentoring may not have
represented a therapeutically sound treatment model, echoing an early suggestion that
the lesson of the CSYS ndings is that ‘good will is not sufcient’ (Wrenn 1952: 215).
The compensatory model is a criticalerror
Despite the important observation that the CSYS did not have a clear therapeutic
approach, the treatment is perhaps best described as individual-directed—because it
focussed explicitly on the individual child as opposed to the child’s social environ-
ment—and compensatory—because it sought to compensate for perceived decits in
the child’s development. However, both of these aspects of the treatment approach may
lead to partial theory failure.
First, individual-level treatment may not present an effective approach to preventing
delinquency without being grounded in a broader ecological approach (Short 1954).
As Lundm an (2001) suggests, such an individual treatment model is premised on the
theoretical assumption that delinquency is an individualized phenomenon as opposed
to the result of social forces. Yet this assumption may be too narrow to deal effectively
with the complex social forces (school, peer and neighbourhood factors) in a child’s
life (Shirk etal. 2000).
Second, the compensatory approach to reducing delinquency may itself be ‘a critical
error’ (McCord 1992: 203). The underlying theory of the CSYS seemed to be that at-risk
youths needed positive attention from adults because such attention had been largely
missing from their lives. Although parental attention (such as supervision and monitor-
ing) may reduce the likelihood of delinquency for a young child, it may have the oppo-
site effect for an at-risk or delinquent child who has already experienced these decits:
“Why did the treatment have harmful effects? Part of the reason, it seems to me, has been the com-
pensatory model on which the treatment was based. Cabot—and many others—have assumed that
an appropriate treatment would undo decits in backgrounds of people at high risk for developing
problems. This can be a critical error. Achild rejected by parents may not be best served by someone
else who tries to take the role of a parent. Such a strategy might result in an exaggerated sense of
loss; it might produce expectations for or dependence on assistance. . . . [A] child who has not been
supervised may become more antisocial if he is placed under close supervision.” (McCord 1992: 203)
Because the compensatory model was grounded in assumptions similar to Hirschi’s
(1969) control theory (e.g. providing conventional bonds), the failure of the compen-
satory model provided ‘grounds for doubting the adequacy of control theory as an
explanation for crime’ (McCord 1992: 204). Consequently, it may not be surprising
that McCord (2004) later developed the construct theory of motivation to explain
delinquency more generally—a social learning rather than control theory of criminal
offending.
Implementation failure
Even if the theoretical foundations of the CSYS were not responsible for the pro-
gramme’s failure, it is possible that the implementation of Cabot’s vision was awed.
ZANE ETAL.
150
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

Implementation failure occurs when a sound theory is not properly executed. As oth-
ers have noted, implementation failure is perhaps not as serious a problem as theory
failure because it does not call the basic idea or mechanism behind the programme
into question (Welsh and Rocque 2014). But, the literature suggests that poorly trained
therapists and dosage effects are the most consistent explanations for iatrogenic effects
(Lilienfeld 2007).
Early observers of the CS YS considered the theoretical foundations of the programme
robust but the actual clinical approach—the implementation of the treatment—unfo-
cussed (Lowry 1940). Perhaps most importantly, there was no formal training of CSYS
counsellors to ensure treatment integrity, and, as a result, the treatment delivered by
counsellors was often unclear and inconsistent. As Barlow (2010: 16)explained, the
implementation of the CSYS involved ‘instructing 10 therapists with no formal training
to do whatever they thought best over a minimum of ve sessions per year for up to ve
years with pre delinquent boys.’ Such instructions do not provide programme delity
or yield a clear treatment model that can be easily evaluated (Vosburgh and Alexander
1980; Barlow 2010). Indeed, W. McCord and J.McCord (1959: 32)observed that the
CSYS may have failed to produce positive results because the counsellors did not have
‘intimate, intensive’ relationships with most of the treatment boys as dictated by Cabot’s
treatmentmodel.
In addition, dosage effects may have contributed to implementation failure (Mears
2007). Just as medical treatment requires proper dosage (where too little is ineffective
and too much is harmful), social intervention programmes necessitate a specic treat-
ment amount. As discussed above, McCord (1978) found that treatment frequency var-
ied considerably among youth and that negative outcomes were associated with more
frequent treatment. While this seems contrary to a traditional dosage effect, an alter-
native explanation would be that while a moderate amount of intervention is more
harmful than little-to-no intervention, frequent intervention (not present in the CSYS)
would have been more effective. Given that the most frequent home visits in the CSYS
were twice per month, it does not appear that any CSYS participants received treatment
that could be characterized as intensive (W. McCord and J.McCord 1959:32).
To date, the most effective treatment prevention programmes for delinquent youth
demonstrate a strong programme model (theory) and delity to that model (imple-
mentation) (Rhule 2005). It is possible that the CSYS was lacking in one, or both, of
these elements.
Heterogeneity of treatment effects
While these design and implementation issues warrant critical attention, there may
be an even more plausible explanation for why the outcomes were iatrogenic rather
than null: the variability of treatment and resulting heterogeneity of treatment effects.
Such concerns have been raised in other contexts. In psychotherapy, Bergin (1963)
argued that many (if not most) null results can actually be ‘accounted for by the mutu-
ally cancelling effects of two different kinds of therapists having provided treatment—
one kind apparently promotes positive change and the other kind promotes negative
change’ (Bergin 1963: 247). On closer observation, Bergin observed that therapist and
client characteristics often interact with treatment to produce wide variation in out-
comes. According to Bergin, this variation is greater for treatment groups than for
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
151
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

controls. For example, based on a careful review of seven unrelated studies with null
results, Bergin (1966) found a signicant difference in the variability of change across
treatment and control groups, despite the average null effect. While treatment and con-
trol groups shared ‘no change’ and ‘spontaneous change’ (positive) effects over time,
the treatment group was subject to an additional ‘double-edged effect’, what he called
‘therapeutic change’ and ‘deterioration’ (Bergin 1966: 237). This theory was named
‘the deterioration effect’ for its provocative suggestion that psychotherapy could actu-
ally harm some clients (Barlow 2010).
Importantly, this nding did not call into question psychotherapy itself, as the poten-
tial for a deterioration effect is accompanied by the potential for an improvement
effect. Instead, this nding suggested that therapy is ‘powerful and, like any discovery
in nature, it can have benecial or harmful effects’ (Bergin 1970: 301), implying a need
to understand what makes different treatments effective, ineffective and even harmful
(Bergin 1966).
Outside of psychotherapy, it has been suggested that the very concept of ‘aver-
age treatment effect’ implies systematic variation in treatment effects (Gaines and
Kuklinski 2011). For example, in their initial follow-up of the CSYS, Powers and Witmer
(1951) suggested that the null effect of the programme seemed to be masking varia-
tion in effects across different groups of youth in the study. Specically, the most at-risk
youth appeared to have the most negative outcomes, while youth with strong family
and community support seemed to benet from the programme. It may be that more
heterogeneous samples have the effect of ‘increasing the variance in subject response
to intervention and reducing statistical power to detect truly powerful effects on some
subgroups but not others’ (Sherman 2007: 304). As McCord (1993: 413–4) suggested,
‘Heterogeneity in samples can mask relationships if the direction of relationships
within subsamples are opposite in direction.’
Sample heterogeneity is closely related to theory and implementation failure. For
example, the CSYS may have been ineffective as a whole due to theory or implemen-
tation failure, and the ndings of harm may have represented iatrogenic effects of a
subset of the study. In other words, a small subgroup could have experienced a deterio-
ration effect—perhaps the summer camp youth due to peer deviancy training—that
would push the results of the otherwise ineffective programme—due to theory and/or
implementation failure—into the red, producing an average iatrogenic effect for the
treatmentgroup.
Even with this understanding, the peer deviancy thesis would not tell the whole story.
Following the logic of Bergin (1963), had the CSYS been more effective for other sub-
groups, this should have cancelled out the deterioration effects of the summer camp
group. Given the heterogeneity of the CSYS programme, there are many subgroup
analyses that can be performed, and the matched-pair design allows for this to be done
in a highly rigorous manner. For all discordant pairs, we have grounds to investigate
cases where the treatment men did better (improvement effect) and where the treat-
ment men did worse (deterioration effect).
Another possible implication of heterogeneous effects is that the quantitative, nomo-
thetic approach of seeking generalizable ndings should be complemented with a
more idiographic approach (such as case study) that focuses on how individual cases
are affected by treatment. While a nomothetic approach tends to treat variation as
a function of sampling and focuses on average between-subject effects, qualitative
ZANE ETAL.
152
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

supplementation may increase understanding of the complex variation in outcomes
by focusing on intra-subject variability (Barlow 2010). This speaks directly to Pawson
and Tilley’s (1997) ‘scientic realist’ approach to evaluation, which calls for a greater
understanding of the mechanisms, contexts and conditions associated with observed
effects. In the words of Pawson and Tilley (1994: 292, emphasis in original), ‘we need
to know why and in what circumstances programmes affect potential subjects before we
can begin to say if they ‘work’.’
For the CSYS, treatment may have helped some, hurt others, and had no effect for
many more. It is necessary to examine patterns of variability of effects in the data
through subgroup analysis as well as through the qualitative assessment of outcomes.
The CSYS provides an important lesson for evaluating crime prevention programmes
today: ‘meaningful discussion of iatrogenic effects must consider these sources of varia-
tion (e.g., implementation quality, program focus and strategies, sample and subgroup
differences) in the youths’ responses to intervention’ (Rhule 2005: 621).
Discussion and Conclusions
This article reviewed the proffered explanations for the iatrogenic effects of the
Cambridge-Somerville Youth Study. In addition to McCord’s hypotheses, others have
weighed in on the subject, providing a rich body of scholarship upon which to draw. It
also presented an opportunity for us to draw upon a wider body of research that has
investigated iatrogenic effects of social interventions and apply this knowledge to the
study. The many stand-alone and interdependent explanations for the CSYS’s iatro-
genic effects illustrate the difculty of ascertaining why the programme was harmful.
We proposed that the heterogeneity of treatment and treatment effects might have
resulted in small overall positive or negative effects via subgroups that were especially
helped or harmed. Ultimately, this suggests that there is not a grand explanation for
the programme’s iatrogenic effects, a lesson that may apply to crime prevention efforts
more generally.
The oft-ignored possibility of iatrogenic effects of social interventions carries impor-
tant implications for research, policy and practice. First, while it would be easy enough
to dismiss a study with null or negative effects and simply ‘move on’, studying why a
social intervention did not work—or even caused harm—is a critical component of con-
necting sound developmental theory with prevention science (Poulin etal. 2001). In our
immediate context, it has been observed that ‘crime policy on the whole lacks a rational
foundation’ (Mears 2007: 679). While one important response to this problem has been
a focus on ‘what works’ in preventing crime, this may ignore important lessons that
can be gleaned from programmes that fail to work or even cause harm. The focus on
positive results is too narrow, ‘because it fails to recognize that some treatments cause
harm’, such as increased offending, increased substance abuse or a general inability to
‘cope with life’ (McCord 2003: 17). This can lead to an under-appreciation of what is
truly at stake in designing social interventions.
A second, related implication is that prevention programmes that are not grounded in
sound criminological theory are unlikely to succeed and may do more harm than good.
Determining the theoretical basis for why some programmes cause harm represents an
important consideration when designing future social interventions, both to diminish
risk of harm and to allow for a more nuanced understanding of intervention effects. To
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
153
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

take a more recent example, Scared Straight was developed on the theoretical premise
that ‘delinquency could be prevented by giving wild youngsters a taste of what it would
be like to be imprisoned’ (McCord 2003: 26). Without any scientically credible evi-
dence that such a programme would actually work, Scared Straight was adopted across
the United States in 38 states—due largely to endorsement by judges and politicians
along with a lm that popularized the ‘get tough’ approach. In their systematic review
of Scared Straight evaluations, however, Petrosino and colleagues (2003) found harm-
ful effects for youth who participated in the programme. As McCord (2003) observed,
if the theory behind a social intervention is erroneous—here, that at-risk youth can
be ‘scared’ into non-deviant behaviour—then there is little reason to expect such pro-
grammes to work. In the case of Scared Straight, exposing youth to adult prisoners
seems to have romanticized the prisoners’ lives and thereby encouraged further offend-
ing, creating a ‘delinquency fullling prophecy’ (Finckenauer 1982:169).
In a similar vein, evaluations of boot camp correctional programmes have shown
them to be generally ineffective in reducing re-offending (Wilson etal. 2005). While
aspects of boot camp programmes may borrow from social learning and social control
theories, the central component is militaristic: strict discipline and rigorous physical
activity are utilized to rehabilitate offenders. This get tough approach may be politi-
cally appealing, but it does not represent a ‘coherent theoretical model’ for reducing
crime and delinquency (MacKenzie 2006: 279). Too often, prevention and intervention
programme designs are based on popular but erroneous notions that ‘draw loosely if at
all on mainstream criminological theories’ (Mears 2007: 671). Yet it is clear that ‘in the
absence of sound theory, social programs stand little chance to bring about social good
and may even cause harm’ (Welsh and Rocque 2014:262).
These concerns are not limited to punitively oriented programmes. Another impor-
tant implication for designing crime prevention programmes is that aggregating
high-risk participants in groups—like the CSYS youth who attended summer camp—
may have unintended consequences. Group interventions are a popular approach in
crime prevention, because they represent a cost-effective alternative to individualized
treatment (Greenwood 2006; Kaminer 2005). However, the possibility of increasing
offending by grouping delinquent peers should not be taken lightly. One example is
the Adolescent Transitions Program, by Poulin and colleagues (2001). Unlike Scared
Straight and correctional boot camps, the intervention appeared to be based on a theo-
retically coherent approach to reducing delinquency and other antisocial behaviours:
to promote self-monitoring, pro-social goal-setting and communication skills while
developing peer environments supportive of such pro-social behaviour. The authors
found that, compared to controls, treatment group youth were signicantly more likely
to have higher rates of self-reported smoking and teacher-reported delinquency three
years later. Moreover, these iatrogenic outcomes were especially stark for treatment
youth with initially low levels of delinquency. The authors concluded that grouping
high-risk youth appeared to undermine the treatment and cause long-term deleterious
effects, leading them to advocate that social interventions should rigorously assess the
risks posed by a proposed design before implementation due to the possibility of unin-
tended, iatrogenic effects.
Yet another key implication is that, in addition to sound theory, incorporating exper-
imental designs into crime prevention programmes may be necessary to evaluate their
true impact (McCord 2003). Despite the failure of the CSYS to prevent delinquency,
ZANE ETAL.
154
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

McCord (1992) insisted that there were several reasons why it should still be consid-
ered a success, including the importance of random assignment to assess the validity
of ndings. After all, the treatment group did fare better than expected. It was only in
comparison to the control group that the treatment group appeared harmed (McCord
1978). As Sherman (2007: 300)observed, if only experimental conditions can reliably
detect (or rule out) iatrogenic effects, then the advancement of experimental criminol-
ogy is tied to the social obligation to preventharm:
The potential for any program to cause harm, no matter how sensible it may appear in theory, remains
the primary ethical justication for experimental criminology … Absent any other widely accepted
means of creating unbiased estimates of treatment effects …, the only way we can be sure that treat-
ments do not harm people is to subject those practices to randomized controlled experiments in eld
settings. One promise of such tests is that they will guide democratic societies in deciding what not to
do, as one way to reduce human suffering.
In addition to these ethical considerations, experiments ‘may be the best way to test the-
ories of the causes of crime’ due to the elimination of various confounding inuences
(Tremblay and Craig 1995: 153; see also McGloin and Thomas 2013). As McCord’s later
research attests (e.g. McCord 1991a; 1991b; 1994), part of the value of the CSYS is that,
as a longitudinal-experimental design, it allows for the testing of causal hypotheses
within the context of the natural history of the development of offending (Farrington
2006; Farrington etal. 2010).
One other key implication for research and policy is the importance of transparency
of evaluation results. Arecent review of 15 Campbell Collaboration systematic reviews
of crime prevention programmes found 22 harmful effects from 22 unique studies out
of a pool of 574 experimental and quasi-experimental studies—approximately 3.8 per
cent. While this may seem small, such a risk of harm is ‘by no means trivial’ (Welsh
and Rocque 2014: 258), particularly if crime prevention must rst ‘do no harm’. More
disturbing is that two-thirds of these harm-causing studies were unpublished, substan-
tiating McCord’s concern that evidence of harm is often ‘buried’ (McCord 2003:28).
The resistance to publishing null or negative ndings is so commonplace today
that unexpected results have been referred to as ‘Type-III’ errors by some statisticians
(Barlow 2010: 17). Such strong resistance should be jettisoned by the research com-
munity and transparency of ndings must become a greater priority. As Rhule (2005:
622)has suggested, ‘the research and clinical community should commend those who
share the negative results of their own interventions, recognizing such a disclosure as a
service to the eld.’ In one of her nal publications, McCord (2003) proposed that eval-
uation results—regardless of the ndings—be made available through a centralized
data bank to combat publication bias and to enhance transparency. This has begun
with the Campbell Collaboration’s library of systematic reviews.
In order for future crime prevention efforts to be successful, we must learn from the
lessons of past programmes. One such lesson is that well-intentioned social interven-
tions can produce iatrogenic effects. While the CSYS is not the only crime prevention
programme to produce iatrogenic effects, it does represent one of the most enduring
examples of how well-intentioned programmes can do harm: ‘No study in the history
of criminology has ever demonstrated such clear, unintended, criminogenic effects
of a program intended to prevent crime. To this day, it is ‘exhibit A’ in discussions
with legislators, students, and others skeptical of the value of evaluating government
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
155
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

programs of any sort, letalone crime prevention programs’ (Sherman 2005: 124). We
have observed that crime prevention research tends to focus on positive results. While
choosing to study programmes that cause harm may be contrary to this focus, doing so
is necessary for sound policy and practice. We could not present the stakes any better
than did McCord (2003: 17): ‘Unless social programs are evaluated for potential harm
as well as benet, safety as well as efcacy, the choice of which social programs to use
will remain a dangerous guess.’
R
A, A. L. and H, L. A. (2009), ‘Exposure to Situations Conducive to
Delinquent Behavior’, Journal of Research in Crime and Delinquency, 46: 5–34.
A, M. E. and H J. N. (1999). ‘First Do No Harm: Adverse Effects of Grouping
Deviant Youth for Skills Training’, Journal of School Psychology, 37: 99–115.
B, C. L. and S, M. E. (2011), Friendships in Childhood and Adolescence. Guilford
Press.
B, D. H. (2010), ‘Negative Effects from Psychological Treatments: APerspective’,
American Psychologist, 65: 13–20.
B, P., H, R. and P, D. (2004), ‘Building Criminal Capital Behind
Bars’, Quarterly Journal of Economics, 124: 105–47.
B, A. E. (1963), ‘The Effects of Psychotherapy: Negative Results Revisited’, Journal of
Counseling Psychology, 10: 244–9.
——. (1966), ‘Some Implications of Psychotherapy Research for Therapeutic Practice’,
Journal of Abnormal Psychology, 71: 235–46.
——. (1970), ‘The Deterioration Effect: AReply to Braucht’, Journal of Abnormal Psychology
75: 300–2.
C, P. S. deQ. (1940), ‘A Long-Term Study of Children: The Cambridge-Somerville
Youth Study’, Child Development, 11: 143–51.
C, M. and B, M. (2009), ‘Intervention in Juvenile Delinquency: Danger of
Iatrogenic Effects?’, Children and Youth Services Review, 31: 1217–21.
D, T. J. (2000), ‘Cross-Setting Consistency in Early Adolescent Psychopathology:
Deviant Friendships and Problem Behavior Sequelae’, Journal of Personality, 68: 1109–26.
D, T. J. and D, K. A. (2005), ‘Peer Contagion in Interventions for Children
and Adolescents: Moving Towards an Understanding of the Ecology and Dynamics of
Change’, Journal of Abnormal Child Psychology, 33: 395–400.
D, T. J. and T, J. M. (2011), ‘Peer Contagion in Child and Adolescent Social
and Emotional Development’, Annual Review of Psychology, 62: 189–214.
D, T. J., MC, J. and P, F. (1999), ‘When Interventions Harm: Peer Groups
and Problem Behavior’, American Psychologist, 54: 755–64.
D, K. A., D, T. J. and L, J. E. (2006), ‘Deviant Peer Inuences in
Intervention and Public Policy for Youth’, Social Policy Report, 20: 1–20.
F, A. N., W, T. A., W, L. G., S, D. B., B, R. I.,
F, G. D. and B, J. R. (2007), ‘The Role of Patients’ Expectations and Goals in
the Behavioral and Pharmacological Treatment of Obesity’, International Journal of Obesity,
31: 1739 –4 5.
F, D. P. (2006), ‘Key Longitudinal-Experimental Studies in Criminology’, Journal
of Experimental Criminology, 2: 121–41.
ZANE ETAL.
156
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

F, D. P., L, R. and W, B. C. (2010), ‘Longitudinal-Experimental
Studies’, in A. R. Piquero and D. Weisburd, eds., Handbook of Quantitative Criminology,
503–18. Springer.
F, J. O. (1982), Scared Straight and the Panacea Problem. Prentice-Hall.
G, B. J. and K, J. H. (2011), ‘Experimental Estimation of Heterogeneous
Treatment Effects Related to Self-Selection’, American Journal of Political Science, 55: 724–36.
G-S, M., D, K. A., D, T. J. and MC, J. (2005), ‘Peer Inuence
in Children and Adolescents: Crossing the Bridge from Developmental to Intervention
Science’, Journal of Abnormal Child Psychology, 33: 255–65.
G, S. and G, E. T. (1934), One Thousand Juvenile Delinquents. Harvard University
Press.
G, D. C. (2010), ‘Deviancy Training: Understanding How Preventative
Interventions Harm’, Journal of Experimental Criminology, 6: 229–43.
G, P. W. (200 6), Changing Lives: Delinquency Prevention as Crime-Control Policy.
University of Chicago Press.
H, D. L. and O, D. W. (2005), ‘Reconsidering Peers and Delinquency: How Do
Peers Matter?’ Social Forces, 84: 1109–30.
H, T. (1969), Causes of Delinquency. University of California Press.
K, A. and K, J. (1999), ‘Self-Esteem, Self-Efcacy, and Welfare Use’, Social Work
Research, 23: 119–26.
K, Y. (2005), ‘Challenges and Opportunities of Group Therapy for Adolescent
Substance Abuse: ACritical Review’, Addictive Behaviors, 30: 1765–74.
K, T. J. (1987), ‘Giving Back Control: Long-Term Poverty and Motivation’, Social Service
Review, 61: 405–19.
L, J. H. (2004), ‘The Life Course of Criminology in the United States’, Criminology, 42: 1–26.
L, S. O. (2007), ‘Psychological Treatments That Cause Harm’, Perspectives on
Psychological Science, 2: 53–70.
L, M. W., W, D. B. and C, L. (2000), Effective Intervention for Serious Juvenile
Offenders. Ofce of Juvenile Justice and Delinquency Prevention, U.S. Department of
Justice.
L, C. (1986), ‘Assaultive Violence as a Contagious Social Process’, Bulletin of the New
York Academy of Medicine, 62: 550–5.
L, L. G. (1940), ‘Special Comment: The Cambridge Somerville Youth Study’, American
Journal of Orthopsychiatry, 19: 611–2.
L, R. L. (2001), Prevention and Control of Juvenile Delinquency. Oxford University Press.
MK, D. L. (2006), What Works in Corrections: Reducing the Criminal Activities of Offenders
and Delinquents. Cambridge University Press.
——. (2013), ‘First Do No Harm: A Look at Correctional Policies and Programs Today’,
Journal of Experimental Criminology, 9: 1–17.
MC, J. (1978), ‘A Thirty-Year Follow-Up of Treatment Effects’, American Psychologist, 33:
284–9.
——. (1980), ‘The Treatment that Did Not Help’, Social Action and the Law, 5: 85–7.
——. (1981), ‘Consideration of Some Effects of a Counseling Program’, in S. E. Martin, L.
B. Sechrest and R. Redner, eds., New Directions in the Rehabilitation of Criminal Offenders,
394–405. National Academy Press.
——. (1991a), ‘Family Relat ionships, Juvenile Delinquency, and Adult Crimin ality’, Criminology,
29: 397–417.
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
157
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

——. (1991b), ‘The Cycle of Crime and Socialization Practices’, Journal of Criminal Law &
Criminology, 82: 211–28.
——. (1992), ‘The Cambridge-Somerville Youth Study: A Pioneering Longitudinal
Experimental Study of Delinquency Prevention’, in J. McCord and R. E. Tremblay,
eds., Prevention of Antisocial Behavior: Interventions from Birth through Adolescence, 196–206.
Guilford Press.
——. (1993), ‘Descript ions and Pred ictions: Three Problems for the Future of Cr iminological
Research’, Journal of Research in Crime and Delinquency, 30: 413–26.
——. (1994), ‘Family So cialization a nd Antisocial B ehavior: S earching for Causal Relat ionships
in Longitudinal Research’, in E. G.M. Weitekamp and H. -J. Kerner, eds., Cross-National
Longitudinal Research on Human Development and Criminal Behavior, 177–88. Kluwer.
——. (2002), ‘Counterproductive Juvenile Justice’, Australian and New Zealand Journal of
Criminology, 35: 230–7.
——. (2003), ‘Cures that Harm: Unanticipated Outcomes of Crime Prevention Programs’,
Annals of the American Academy of Political and Social Science, 587: 16–30.
——. (2004), ‘Toward a Theory of Criminal Responsibility’, in J. McCord, ed., Beyond
Empiricism: Institutions and Intentions in the Study of Crime, 147–76. Transaction Publishers.
MC, W. and MC, J. (1959), Origins of Crime: ANew Evaluation of the Cambridge-
Somerville Youth Study. Columbia University Press.
MG, J. M. and T, K. J. (2013), ‘Experimental Tests of Criminological Theory’,
in B. C. Welsh, A. A. Braga and G. J.N. Bruinsma, eds., Experimental Criminology: Prospects
for Advancing Science and Public Policy, 15–42. Cambridge University Press.
M, D. P. (2007), ‘Towards Rational and Evidence-Based Policy’, Journal of Criminal
Justice, 35: 667–82.
M, J. and H, P. (2011), ‘Contagion and Repeat Offending Among Urban Juvenile
Delinquent s’, Journal of Adolescence, 34: 951–63.
M, D. M. and W, A. T. (2012), Program Evaluation Theory and Practice:
AComprehensive Guide. Guilford Press.
M, T. E., C, A., R, M. and S, P. A. (2001), Sex Differences in Antisocial
Behavior. Cambridge University Press.
O, D. W. and B, L. (2006), ‘Peer Effects in Juvenile Justice’, in K. A. Dodge, T.
J. Dishion and J. Lansford, eds., Deviant Peer Inuences in Programs for Youth: Problems and
Solutions, 141–61. Guilford Press.
O, D. W., W, J. K., O’M, P. M., B, J. G. and J, L. D. (1996),
‘Routine Activities and Individual Deviant Behavior’, American Sociological Review, 61: 635–55.
P, H. and W, R. (1973), Stay Where You Were: A Study of Unemployables in
Industry. Lippincott.
P, R. and T, N. (1994), ‘What Works in Evaluation Research?’ British Journal of
Criminology, 34: 291–306.
——. (1997), Realistic Evaluation. Sage.
P, A., T-P, C. and B, J. (2003), ‘“Scared Straight” and
Other Juvenile Awareness Programs for Preventing Juvenile Delinquency: ASystematic
Review of Randomized Experimental Evidence’, Annals of the American Academy of Political
and Social Science, 589: 41–62.
P, F., D, T. J. and B, B. (2001), ‘3-Year Iatrogenic Effects Associated
with Aggregating High-Risk Adolescents in Cognitive Behavioral Preventative
Interventions’, Applied Developmental Science, 5: 214–24.
ZANE ETAL.
158
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

P, E. (1949), ‘An Experiment in the Prevention of Delinquency’, Annals of the American
Academy of Political and Social Science, 261: 77–88.
P, E. and W, H. L. (1951), An Experiment in the Prevention of Delinquency: The
Cambridge-Somerville Youth Study. Columbia University Press.
R, D. M. (2005), ‘Take Care to Do No Harm: Harmful Interventions for Youth Problem
Behav ior’, Professional Psychology: Research and Practice, 36: 618–25.
R, P. H., L, M. W. and F, H. E. (2004), Evaluation: ASystematic Approach,
7th edn. Sage.
S, L. W. (2005), ‘The Use and Usefulness of Criminology, 1751–2005: Enlightened
Justice and Its Failures’, Annals of the American Academy of Political and Social Science, 600:
115–35.
——. (2007), ‘The Power Few: Experimental Criminology and the Reduction of Harm’,
Journal of Experimental Criminology, 3: 299–321.
S, S., T, A . and O, D. (2000), ‘A Developmental Psychopathology Perspective
on Child and Adolescent Treatment Policy’, Development and Psychopathology, 12:
835–55.
S, J. F. (1954), ‘An Experiment in the Prevention of Delinquency: The Cambridge-
Somerville Youth Study by Edwin Powers; Helen Witmer’, American Journal of Sociology, 59:
587–8.
S, C., S, T. A.B. and P, M. (2010), ‘Dynamic Networks and Behavior:
Separating Selection from Inuence’, Sociological Methodology, 40: 329–93.
T, R. E. (2005), ‘Towards an Epigenetic Approach to Experimental Criminology’,
Journal of Experimental Criminology, 1: 397–415.
T, R. E. and C, W. M. (1995), ‘Developmental Crime Prevention’, Crime and
Justice, 19: 151–236.
V, F., B, M., G, C. -É. and T, R. E. (2013), ‘Early Prevention
of Life-Course Personal and Property Violence: A19-Year Follow-Up of the Montreal
Longitudinal-Experimental Study (MLES)’, Journal of Experimental Criminology, 9:
411–27.
V, W. W. and A, L. B. (1980), ‘Long-Term Follow-Up as Program
Evaluation: Lessons from McCord’s 30-Year Follow-Up of the Cambridge-Somerville
Youth Study’, American Journal of Orthopsychiatry, 50: 109–24.
W, F. (2011), ‘Delinquent Peers in Context: A Longitudinal Network Analysis of
Selection and Inuence Effects’, Criminology, 49: 253–86.
W, F. M., B, W., B, G. J.N. and P, L. J.R. (2013), ‘When
is Spending Time with Peers Related to Delinquency? The Importance of Where, What,
and With Whom’, Crime & Delinquency. doi:10.1177/0011128713478129.
W, B., C, A., B, S., T, J., J, M. and W, J. R. (2005), ‘Iatrogenic
Effects of Group Treatment for Antisocial Youth’, Journal of Consulting and Clinical
Psychology, 73: 1036–44.
W, B. C. and R, M. (2014), ‘When Crime Prevention Harms: A Review of
Systematic Reviews’, Journal of Experimental Criminology, 10: 245–66.
W, D. J. and F, D. P. (1977), The Delinquent Way of Life. Heinemann.
W, D. B., MK, D. L. and M, F. N. (2005), ‘Effects of Correctional Boot
Camps on Offending’, Campbell Collaboration Library of Systematic Reviews. doi:10.4073/
csr.2005.6.
IATROGENIC EFFECTS OF THE CA MBRIDGE-SOMERVILLE YOUTH STUDY
159
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from

W, C. G. (1952), ‘An Experiment in the Prevention of Delinquency: The Cambridge-
Somerville Youth Study by Edwin Powers; Helen Witmer’, Annals of the American Academy
of Political and Social Science, 279: 214–5.
Y, J. T.N. (2011), ‘How Do They ‘End up Together’? ASocial Network Analysis of Self-
Control, Homophily, and Adolescent Relationships’, Journal of Quantitative Criminology,
27: 251–73.
ZANE ETAL.
160
at Northeastern University Libraries on March 29, 2016http://bjc.oxfordjournals.org/Downloaded from