Kazdin 2007 PDF

ANRV307-CP03-01 ARI 20 February 2007 18:34
Mediators and Mechanisms

of Change in Psychotherapy
Research
Annu. Rev. Clin. Psychol. 2007.3:1-27. Downloaded from arjournals.annualreviews.org
Alan E. Kazdin
by University of Arizona Library on 08/22/09. For personal use only.
Department of Psychology, Yale University, New Haven, Connecticut 06520-8205;

email: alan.kazdin@yale.edu
Annu. Rev. Clin. Psychol. 2007. 3:127 Key Words

First published online as a Review in mediators, moderators, mechanisms of psychotherapy, processes
Advance on October 11, 2006
and outcomes of therapy, randomized controlled trials, mediation
The Annual Review of Clinical Psychology is analyses, treatment evaluation
online at http://clinpsy.annualreviews.org
This articles doi: Abstract

10.1146/annurev.clinpsy.3.022806.091432
There has been enormous progress in psychotherapy research. This
Copyright c 2007 by Annual Reviews. has culminated in recognition of several treatments that have strong
All rights reserved
evidence in their behalf. Even so, after decades of psychotherapy re-
1548-5943/07/0427-0001$20.00 search, we cannot provide an evidence-based explanation for how or
why even our most well studied interventions produce change, that
is, the mechanism(s) through which treatments operate. This chap-
ter presents central requirements for demonstrating mediators and
mechanisms of change and reviews current data-analytic and designs
approaches and why they fall short of meeting these requirements.
The role of the therapeutic alliance in psychotherapy and cognitive
changes in cognitive therapy for depression are highlighted to illus-
trate key issues. Promising lines of work to identify mediators and
mechanisms, ways of bringing to bear multiple types of evidence,
recommendations to make progress in understanding how therapy
works, and conceptual and research challenges in evaluating media-
tors and mechanisms are also presented.
1
Contents
INTRODUCTION . . . . . . . . . . . . . . . . . 2 Meticulous Description . . . . . . . . . . . 12
CONCEPTUAL AND Moderators as a Path to Identifying
DEFINITIONAL ISSUES . . . . . . . 3 Mediators and Mechanisms . . . . 13
REASONS FOR STUDYING Direct Intervention and
MEDIATORS AND Manipulation. . . . . . . . . . . . . . . . . . 14
MECHANISMS . . . . . . . . . . . . . . . . . 3 Converging Lines of Work . . . . . . . . 15
REQUIREMENTS FOR General Comments . . . . . . . . . . . . . . . 16
DEMONSTRATING RECOMMENDATIONS FOR
MEDIATORS AND RESEARCH . . . . . . . . . . . . . . . . . . . . . 16
MECHANISMS OF CHANGE . . 4 Use Theory as a Guide . . . . . . . . . . . 16
Multiple Criteria . . . . . . . . . . . . . . . . . 4 Include Measures of Potential

General Comments . . . . . . . . . . . . . . . 6 Mediators in Treatment
CURRENT STATUS OF Studies . . . . . . . . . . . . . . . . . . . . . . . . 17
RESEARCH ON MEDIATORS Establish the Timeline of the

AND MECHANISMS . . . . . . . . . . . 6 Proposed Mediator or
Examples Where Mediators and Mechanism and Outcome . . . . . . 17
Mechanisms are Discussed But Assess More than One Mediator or
Not Well Established . . . . . . . . . . 6 Mechanism . . . . . . . . . . . . . . . . . . . 17
General Comments . . . . . . . . . . . . . . . 8 Use Designs that Can Evaluate
OVERVIEW OF METHODS FOR Mediators and Mechanisms . . . . 18
STUDYING MEDIATORS AND Examine Consistencies Across
MECHANISMS IN Different Types of Studies . . . . . 19
PSYCHOTHERAPY . . . . . . . . . . . . 8 Intervene to Change the Proposed
Statistical Techniques . . . . . . . . . . . . . 8 Mediator or Mechanism . . . . . . . 20
Design Methods for Studying SPECIAL CHALLENGES AND
Mediators and Mechanisms . . . . 10 OBSTACLES . . . . . . . . . . . . . . . . . . . . 21
General Comments . . . . . . . . . . . . . . . 11 Mechanism-Outcome Relations . . . 21
PATHS TO IDENTIFYING AND Everything in Moderation . . . . . . . . 22
ELABORATING MEDIATORS Measurement Development . . . . . . . 22
AND MECHANISMS . . . . . . . . . . . 11 CONCLUSIONS . . . . . . . . . . . . . . . . . . . 23
INTRODUCTION Discussions and theory about why psy-

chotherapy changes people are plentiful, but
Several forms of psychotherapy for children,
supportive evidence is quite rare. The Amer-
adolescents, and adults produce therapeutic
ican humorist and writer, Mark Twain (1835
change, as demonstrated in scores of con-
1910) noted that, everybody talks about the
trolled treatment studies (Kazdin & Weisz
weather but nobody does anything about it.1
2003, Lambert 2004, Nathan & Gorman
Mechanism is the weather of psychotherapy
2007). The changes can encompass social,
research. The focus of this review is on why
emotional, cognitive, behavioral, educational,
treatment works, through what processes, and
and physical spheres of functioning. We know
well that therapy works, i.e., is responsible 1
Mark Twain (Samuel L. Clemens) is credited with this
for change, but have little knowledge of why quote but some suggest that his coauthor A. Charles Dudley
or how it works. Warner wrote the statement (Twain & Warner 1874).
2 Kazdin
how change comes about, or more succinctly, therapy (or some independent variable) actu-
on the mechanism(s) of therapeutic change. ally unfolds and produces the change. Mecha-
This review also discusses the importance of nism explains how the intervention translates
studying mediators and mechanisms of ther- into events that lead to the outcome. This
apy, examines the limitations of existing data is easily confused with the notion of media-
evaluation and design strategies, and pro- tion. For example, cognitions may be shown
vides recommendations for changes needed in to mediate change in therapy, an important
research. lead perhaps. However, this does not explain
specically how the change came about, i.e.,
what are the intervening steps between cog-
CONCEPTUAL AND nitive change and reduced stress or anxiety. In
DEFINITIONAL ISSUES this review, the primary focus is on mediators
and mechanisms. The goal is to understand
Several related concepts are important to de- the mechanisms of change; the study of medi-
lineate in part because of their confusion but ators is often a rst step, as is illustrated below.
also because they are relevant to elaborating
Moderator refers to some characteristic

mechanisms (please see Table 1). It is useful to that inuences the direction or magnitude
begin with cause or causal relation. A random- of the relation between the intervention and
ized controlled trial (RCT) may show that outcome. If treatment outcome varies as a
treatment compared to no treatment leads function of characteristics of the patient or
to therapeutic change. From the demonstra- therapist (e.g., sex, ethnicity, temperament),
tion, we can say that the treatment caused the treatment delivery (e.g., individual versus
change, as the term cause is used in science. group treatment), or cohort (e.g., so-called
Demonstrating a cause does not say why generation X versus baby boomers), these lat-
the intervention led to change or how the ter variables are moderators. I return to mod-
change came about. To evaluate how change erators below because they have important
comes about, research often looks at me- bearing on mediators and mechanisms.
diators. As noted in the table, mediator is
a construct that shows important statistical
relations between an intervention and out-
come, but may not explain the precise pro- REASONS FOR STUDYING
cess through which change comes about. The MEDIATORS AND MECHANISMS
term mechanism refers to a greater level of Evaluating mediators and mechanisms of
specicity than does the term mediator and therapeutic change is important for sev-
reects the steps or processes through which eral reasons. First, there is an embarrassing
Table 1 Key terms and concepts

Cause: a variable or intervention that leads to and is responsible for the outcome or change.
Mediator: an intervening variable that may account (statistically) for the relationship between the
independent and dependent variable. Something that mediates change may not necessarily explain the
processes of how change came about. Also, the mediator could be a proxy for one or more other variables
or be a general construct that is not necessarily intended to explain the mechanisms of change. A mediator
may be a guide that points to possible mechanisms but is not necessarily a mechanism.
Mechanism: the basis for the effect, i.e., the processes or events that are responsible for the change; the
reasons why change occurred or how change came about.
Moderator: a characteristic that inuences the direction or magnitude of the relationship between and
independent and dependent variable. If the relationship between variable x and y varies is different for
males and females, sex is a moderator of the relation. Moderators are related to mediators and mechanisms
because they suggest that different processes might be involved (e.g., for males or females).
www.annualreviews.org Mediators and Mechanisms of Psychotherapy 3

wealth of treatments in use. For example, in ing the processes through which treatment
the context of child and adolescent therapy operates can help sort through those facets
alone, 550+ psychotherapies can be delin- that might be particularly inuential in treat-
eated (Kazdin 2000). Some of these are known ment outcome and permit better selection of
to produce change; it is not likely that the dif- suitable patients. For example, if changes in
ferent treatments produce change for differ- cognitive processes account for therapeutic
ent reasons. Understanding the mechanisms change, this nding might draw attention to
of change can bring order and parsimony to the pretreatment status of related processes
the current status of multiple interventions. (abstract reasoning, problem-solving, attribu-
Second, therapy can have quite broad out- tions) that might moderate who responds or
come effects, beyond the familiar benets fails to respond to treatment.
of reducing social, emotional, and behavioral Finally, understanding the mechanisms
problems (e.g., suicidal ideation, depression, through which change takes place is impor-
and panic attacks). Therapy also alters physi- tant beyond the context of psychotherapy.
cal conditions (e.g., pain, blood pressure), im- Many interventions or experiences in every-
proves recovery from surgery or illness, and day life improve adjustment and adaptive
increases the quality of life (see Kazdin 2000). functioning, ameliorate problems of mental
How do these effects come about? Elaborat- and physical health, help people manage
ing mechanisms of therapy will clarify the and cope with stress and crises, and more
connections between what is done (treatment) generally navigate the shoals of life. As exam-
and the diverse outcomes. ples, participating in religion, chatting with
Third, by understanding the processes that friends, exercising, undergoing hypnosis,
account for therapeutic change one ought to and writing about sources of stress all have
be better able to optimize therapeutic change. evidence in their behalf. Mechanisms that
Indeed, without understanding what is critical elaborate how therapy works might have gen-
to treatment and how it operates, we are at a erality for understanding human functioning
bit of a loss. Should we focus on more practice, beyond the context of therapy. The other
catharsis, chatting, homeworkwhat leads to side is also true. Mechanisms that explain
change and why? If we know how changes how other change methods work might well
come about, perhaps we can direct better, inform therapy. Basic psychological processes
stronger, different, or more strategies that (e.g., learning, memory, perception, persua-
trigger the critical change process(es). sion, social interaction) and their biological
Fourth, extending treatments from re- pathways (e.g., changes in neurotransmit-
search to clinic or real world settings will ters) may be common to many types of
be difcult without understanding how treat- interventions, including psychotherapy.
ment works. We enter the clinical arena with
one hand tied beyond our back if we apply
an unspecied and possibly low dose of some REQUIREMENTS FOR
treatment that we do not understand. To opti- DEMONSTRATING MEDIATORS
mize the generality of treatment effects from AND MECHANISMS OF CHANGE
research to practice we want to know what
is needed to make treatment work, what are
Multiple Criteria
the optimal conditions, and what components Establishing a mediator or mechanism has
must not be diluted to achieve change. several requirements. The requirements are
Fifth, understanding how therapy works highlighted because they provide the back-
can help identify moderators of treatment, ground for why changes are needed in
i.e., variables on which the effectiveness of research. I focus on mediation because
a given treatment may depend. Understand- this is an important interim step between
4 Kazdin
demonstrating a causal relation and under- come (A and C). This type of demonstra-
standing concretely the mechanism of action tion (e.g., RCT) is common and demonstrates
through which the effect occurs. Also, medi- cause. However, uncommon are experiments
ation is the primary focus of contemporary that manipulate the proposed mediator or
research. mechanism (B) and show the impact on out-
come (C). Experimental evidence strengthens
Strong association. Demonstration of a the case that a proposed mediator is responsi-
strong association between the psychothera- ble for a change in the outcome of interest.
peutic (A) intervention and the hypothesized
mediator of change (B) is an initial require- Timeline. A timeline must be established to
ment. Then of course, there ought to be an as- infer a causal relation or mediator of change.
sociation between the proposed mediator (B) Causes and mediators must temporally pre-
and therapeutic change (C). Indeed, if these cede the effects and outcomes. Demonstrat-
three variables are not related, the case for the ing a timeline between cause and an effect, al-
operation of a mediator is greatly weakened, beit obvious, is the Achilles heel of treatment
if not eliminated. studies, as I elaborate below.
Specificity. The second criterion refers to Gradient. Showing a gradient in which

the demonstration of the specicity of the as- stronger doses or greater activation of the
sociation between the intervention, proposed proposed mediator is associated with greater
mediator, and outcome. We would not want change in the outcome can help make the case
multiple mediators to account for the change, for a particular mediator. A common analysis
but rather show a more specic connection. A in medicine, epidemiology, and public health
demonstration that many plausible constructs is showing a dose-response relation. For ex-
do not account for therapeutic change, with ample, there is a dose-response (and linear)
the exception of one, strengthens the argu- relation between passive cigarette smoke (i.e.,
ment that the proposed construct mediates exposure to secondhand smoke) and coronary
change. heart disease (He et al. 1999). Demonstrating
a dose-response relation increases the plausi-
Consistency. Replication of an observed re- bility of an agent being causally involved and
sult across studies, samples, and conditions, may point to likely mechanisms as well. Of
i.e., consistency in the relation, contributes course, it is possible that there is no dose-
to inferences about mediators. We expect the response relation (e.g., a qualitative or on-
relations among A, B, and C not to be sam- off gradient) or that the relation is not linear.
ple specic. Inconsistency across two or more Such relations do not mean a particular con-
demonstrations does not necessarily mean struct is not causally related, but may make
that the proposed mediator is not involved. inferences more difcult or require supple-
The relation between a proposed mediator mentary information.
and outcome might be perfectly consistent but
moderated by a variable we have not yet iden- Plausibility or coherence. Plausibility or
tied. Yet, when consistency across studies is coherence of an explanation of how a me-
obtained, this greatly facilitates drawing in- diator or mechanism operates and integra-
ferences about whether a particular mediator tion of ndings with the broader scientic
may be involved. knowledge base contribute to the inferences.
In medicine, pathophysiology often is invoked
Experimental manipulation. Direct ma- to meet this criterion. That is, in light of the
nipulation through an experiment obviously ndings, is there a plausible, coherent, and
makes a strong case between therapy and out- reasonable process (e.g., buildup of plaque)

through which the disorder (e.g., atheroscle- change. We need a plausible account of how
rosis and heart attack) might be explained? the construct works and leads (in a testable
An explanation is plausible because it invokes way) to the outcomes.
other information and steps in some process-
outcome relation that are reasonable or sup-
ported by other research. General Comments
The use of plausibility and coherence Drawing inferences about a mediator of
to elaborate mechanisms is poignantly illus- change requires convergence of multiple cri-
trated in child abuse. Occasionally, parents teria because they act in concert. Interpreta-
bring their very injured and pained child to tion of what accounts for or explains a par-
an emergency room for treatment and tell ticular relation (mediator, mechanism) is not
the physician that the child has been injured. likely to come from a single investigation. By
Three examples from my own experience in- the very nature of one of the criteria (consis-
clude a child who allegedly fell off a bicycle, tency), replication is required. Yet, apart from
another who fell down the ve front stairs of that criterion, the case for a mediator is built
a cement porch at home, and a child who got by a sequence of studies that may vary in the
into a stght with a seven-year-old sibling. set of criteria they address and the clarity of
In each case, the physician (a different person the demonstration. After several studies, and
for each case) was suspicious because the in- when all or most of the criteria are met, one
juries consisted of large and deep bruise marks can state that some intervening process ac-
across the back (with lines resembling a belt) counts for change.
and a mark that could resemble a belt buckle
on the upper shoulder (child 1); three or possi-
bly four round burn marks on the childs back CURRENT STATUS OF
in the size of the end of a cigarette (child 2); RESEARCH ON MEDIATORS
and a black eye and open scalp wound un- AND MECHANISMS
der the hair (child 3). The physician in each Mechanisms of treatment are increasingly dis-
case was suspicious primarily based on the cussed, a likely precursor to more empirical
criterion of plausibility and coherence of the work on the topic (e.g., Brent & Kolko 1998,
mechanisms or process involved leading to Grawe 2004, Hofmann 2000, Kazdin 2006,
these outcomes. In light of how a child is likely Kazdin & Nock 2003, Weersing & Weisz
to fall off a bicycle or down the stairs or to 2002). I believe this has fostered the view
be hit by young sibling, respectively, the in- that we know about key processes leading to
juries were not very likely (plausible, coher- change and are using suitable methodologi-
ent). However, the injuries were very plausible cal, statistical, and design tools. Few empir-
by invoking another process or mechanism, ical studies are available that meet even two
namely, parent abuse of their children. (One or three of the criteria mentioned previously.
could use the term parsimonious here, but Consider briey two therapy areas where me-
I use plausible and coherent to focus on diators and mechanisms of action are often
a greater level of specicity, namely, looking discussed.
at the operation of a mechanism and how it
unfolds to produce an outcome.)
In relation to psychotherapy, plausibility Examples Where Mediators and
and coherence convey the importance of theo- Mechanisms are Discussed But Not
retically based investigation of mediators and Well Established
mechanisms of change. Here we need more Therapeutic alliance and treatment out-
than a global construct that can be used to ex- come. The therapeutic alliance refers to the
plain onset of a clinical problem or therapeutic collaborative nature of the patient-therapist
6 Kazdin
interaction, their agreement on goals, and esting to pursue. The broader point is more
the personal bond that emerges in treatment. pertinent here, to wit, in the vast majority
A consistent nding is that the stronger the of studies the timeline between alliance and
alliance the greater the therapeutic change symptom change has not been established.
(Horvath & Bedi 2002, Orlinsky et al. 2004).
Studies that evaluate alliance during (e.g., Cognitions in cognitive therapy for de-
early, middle) treatment often show that al- pression. There are very few forms of psy-
liance predicts improvement in symptoms at chotherapy as well established as cogni-
the end of treatment. Showing that alliance tive therapy (CT) for unipolar depression
predicts later symptom change by itself does among adults (American Psychiatric Associ-
not show that alliance plays a causal role, ation 2000, Hollon & Beck 2004). This treat-
leaving aside the more specic matter of re- ment is evidence based, and then some, in
ecting a potential mediator. Merely because light of the range of trials. But why does CT
symptoms are not assessed in the middle of work, i.e., through what mediators or mech-
treatment, does not mean they have not al- anisms? In fact, little can be stated as to why
ready changed. Perhaps very early in treat- treatment works. In the development of this
ment clients get a little better (some symptom treatment, the basis of therapeutic change was
improvement) and as a result form a positive thought to be changes in key cognitive pro-
alliance with the therapist. cesses (negative triad) that characterize many
For example, a study of psychodynami- depressed patients. CT is designed to change
cally oriented supportive therapy showed that these cognitions and in the process change
changes in alliance early in treatment pre- depression. The relation of cognitions and
dicted symptom change at the end of treat- cognitive change in treatment to therapeu-
ment, in keeping with a large body of evidence tic change has been studied in different ways
(Barber et al. 2000). However, a critical ad- by assessing symptom change and cognitive
dition was included. Both symptom change change at the end of treatment and show-
and alliance were assessed at multiple points. ing that one shares variance with the other,
Symptom changes early in treatment pre- or by evaluating whether cognitions assessed
dicted alliance and that alliance also predicted early or in the middle of treatment corre-
further symptom change. Thus, the famil- late with subsequent therapeutic change (e.g.,
iar alliance-outcome correlation in part re- DeRubeis et al. 1990, Kwon & Oei 2003). In
ects the relation of early and later symptom both of these methods, the timeline problem
change, and the timeline is symptom change is unresolved, i.e., we do not know the order-
to alliance as well as the reverse. Assessment of ing of cognitive change and symptom change.
both symptom change and alliance were com- This issue is similar to the concern raised in
pleted at multiple points during the course of relation to alliance, namely, in the vast ma-
treatment to identify these interesting rela- jority of studies, symptom change may have
tions. Other studies with assessments at mul- preceded or occurred concurrently with cog-
tiple points have shown that a positive al- nitive changes. From research as currently de-
liance may follow improvements in symptoms signed and discussed, it is not possible to say
(DeRubeis & Feeley 1990, Tang & DeRubeis that cognitive processes serve as the mediators
1999). of therapeutic change.
From these examples, I do not wish to as- Actually, unlike the research on alliance,
sert that alliance is invariably the effect rather perhaps one can say a bit more about me-
than a cause. Indeed, the correlational evi- diators and mechanisms of cognitive therapy.
dence does not permit statements about cause The research permits one to say more about
or mediation. The reciprocal or bidirectional what is not a likely mediator of the effects
relations of symptoms and alliance are inter- of CT. Tests of mediation and evaluation of

therapeutic changes quite early in the course feelings that life is meaningless? The time se-
of treatment suggest that improvements can quence problem is more basic, but how does
readily occur without changes in cognitions or one get from my therapist and I are bonding
in advance of implementing cognitive-change to my marriage, anxiety, and tics are better?
strategies in treatment (e.g., Burns & Spangler This is a leap with the intervening steps un-
2001, Tang & DeRubeis 1999). Challenges specied or untested, at least to my knowl-
to the cognitive bases of change in CT for edge. The steps are not academic. If we could
depression are not new (Ilardi & Craighead identify the steps, there may be other ways
1994, Whisman 1999). Perhaps we can state to activate them than through alliance alone.
more condently now than before that what- Also, we might identify novel moderators re-
ever may be the basis of changes with CT, it lated to the mechanisms that help us select
does not seem to be the cognitions as origi- individuals likely to vary in responsiveness to
nally proposed. the intervention.
General Comments
OVERVIEW OF METHODS FOR

I have highlighted two areas that are often dis- STUDYING MEDIATORS AND
cussed as if we know the basis for the effect, MECHANISMS IN
i.e., the mechanism involved. In both alliance PSYCHOTHERAPY
and cognitive therapy literatures, the timeline Dominant methods of evaluating mediators
problem is a methodological shroud that cov- in therapy research have limited what can be
ers most studies. Without clearly establishing concluded in large part because critical con-
that the putative basis for the effect invariably ditions mentioned previously are not met. I
comes before symptom change, conclusions highlight current methods and discuss why we
about mediation are in question. Of course, have not been able to learn very much about
it follows that more specic statements about mediators or mechanisms from them.
mechanisms are premature.
The two examples are intended to convey
how a key criterion, establishing a timeline, is Statistical Techniques
not met in otherwise well-studied areas where
Tests of mediation. Statistical evaluation
mechanisms are discussed. The examples were
can play a central role in addressing whether
used to illustrate this single point rather than
a particular construct accounts for change.
to review comprehensively the respective lit-
Multiple regression techniques, path analy-
eratures. Although I focused on the timeline
sis, structural equation modeling, and boot-
problem, other concerns in relation to these
strap methods are prominent options (Baron
literatures could be illustrated by applying all
& Kenny 1986, Holmbeck 2002, Hoyle &
of the criteria. Let me mention one to note
Smith 1994, Kenny et al. 1998, MacKinnon
that this is not a vacuous claim.
et al. 2002, Shrout & Bolger 2002). Multiple
We do not have a clear picture or set of
regression analyses have been the most com-
studies that test how the putative mechanism
monly used techniques, and an overview of the
unfolds in such a way as to alter symptoms. In
logic conveys the benets as well as the prob-
relation to plausibility and coherence of the
lems. Consider a hypothetical outcome study
mechanism-outcome relation, precisely what
in which we evaluate the following compo-
happens that leads to symptom change? For
nents:
example, through what process or sequence of
events along any dimensions (cognitive pro- A = an intervention (the treatment)
cesses, neurotransmitters, stress) does alliance B = a mediator or intervening variable
lead to reductions in depression, anxiety, or C = an outcome (therapeutic change)
8 Kazdin
The interrelations of A, B, and C, as eval- Even when the timeline is established,

uated statistically, are used to infer whether B mediation does not necessarily suggest the
can explain why treatment works. In demon- mechanism of action. If, for example, some
strating mediation statistically, four condi- cognitions are shown to come before symp-
tions and tests are usually proposed: tom change and statistically explain the
The treatment or intervention (A) must intervention-outcome (A-C) relation that by
be related to therapeutic change (C); itself does not show the construct is the
mechanism. What precisely is the process of
The treatment (A) must be related to
change, what are the steps from the construct
the proposed mediator (B);
to the change, and are other variables embed-
The proposed mediator (B) must be re-
ded in the measure? One might say that cog-
lated to therapeutic change (C); and
nitions as a mediator might be a rst step to
The relation between the intervention move to more ne-grained analyses, a defen-
(A) and therapeutic change (C) must be sible position. But one must also say that cog-
reduced after statistically controlling for nitions might not be the variable at all or at
the proposed mediator (B).
least the cognitions of interest to the inves-

The logic seems compelling because the tigator. Cognitions might be a proxy variable
many conditions, if met, suggest that the im- for some other construct or be a global con-
pact of treatment (A) on therapeutic change struct that includes multiple distinguishable
(C) really depends on some intervening pro- components (Kraemer et al. 2001).
cesses (B). My simple description ignores In cross-sectional studies, the language
many nuances. For example, the extent to used to describe the data-analytic strategies
which the A-C relationship is reduced by con- and the ndings lends itself to misconception
trolling for B is a matter of degree, and there in relation to the timeline. For instance, re-
is no cut point (beyond statistical signicance gression analyses identify variables as pre-
[e.g., Sobel test]) to decide whether there is dictors or independent variables and others
or is not support for a particular mediational as outcomes or dependent variables. Yet,
view. Also, the various statistical analyses are the timeline is only established by the ex-
not free from controversy or challenge (e.g., perimental design. The distinction between
Kraemer et al. 2001, 2002). I merely wish to antecedent (independent) and outcome (de-
convey that the statistical analyses highlighted pendent) variables, from the standpoint of the
here are used to evaluate mediators of change steps of the statistical analyses (and printouts)
in therapy. is arbitrary. Similarly, the accompanying dia-
As is so often the case in statistical analyses, grams of the results with a ow chart of some
the concerns here are not about the statistics kind to convey the authors views of mediation
per se but about their use and interpretation. (e.g., a structural equation model with arrows
A key interpretive limitation is the fact that pointing to the right) may lead the author or
the timeline between the mediator and the reader to conclude erroneously that there is a
outcome is not necessarily established. Most timeline.
psychotherapy studies of mediation evaluate
the mediator and symptoms at pre and post Percentage of variance. Occasionally, re-
or evaluate the mediator but not the symp- searchers focus on the notion of percentage
toms during treatment. Change in the media- of variance accounted for, or explained by, a
tor is shown to correlate, predict, and account variable and consider this as proof that a criti-
for variance in relation to the outcome. The cal process or the critical explanation has been
statistical analysis alone cannot establish that identied. If two variables are correlated (r),
one inuence preceded, and therefore possi- then one can identify the proportion of com-
bly mediated, the other. mon or shared inuence (r2 ). For example,

therapeutic processes (e.g., alliance) predict studies that assess integrity do so at the end
therapeutic change. Researchers often note of treatment, at the same time that symptom
that alliance accounts for a signicant pro- change is evaluated, raising some timeline is-
portion of variance and sometimes even more sues we can forego here.) The investigator
variance than other inuences (e.g., treat- may analyze the data with only those clients
ment technique). Further interpretation is of- who received the intervention as intended or
ten added to suggest this must mean that the who received some minimal dose or by includ-
alliance is why treatment leads to change or ing all the data and showing a correlation be-
is the most signicant/important inuence in tween how well treatment was implemented
therapy. and the degree of therapeutic change. Receiv-
Nothing in the measure of percentage of ing the appropriate levels of treatment is not
variance speaks to mediators or mechanisms. randomly distributed and may well be con-
First, shared variance of alliance and outcome founded with client or client x therapist char-
could be huge, but that could be due to symp- acteristics. For example, getting more, better,
tom change occurring before alliance. Sec- or more carefully implemented treatment may
ond, the therapeutic alliance can account for relate to the personality of the client (or ther-
treatment outcome variance but itself be ex- apist), severity of his or her problems, match
plained by one or more other variables, such of values and interests between the therapist
as common method variance in the alliance and patient, and more.
outcome measures or even characteristics of The example is on treatment integrity but
the patients before they came to treatment the broader point is critical. In studying any
(e.g., Kazdin & Whitley 2006, Zigler & Glick intervening process or construct of therapy,
1986). In short, amount variance may or may the investigator may wish to include in the
not point toward mediators or mechanisms. data analyses only those individuals for whom
Whether the relation (any correlation) pro- the putative mediator was invoked or oc-
vides meaningful leads will stem from the con- curred. After random assignment of cases to
ditions required for establishing mediators, as different groups, keeping or using only cases
enumerated above. where the mediator was effectively manipu-
lated changes the equality of the sample and
Biases in the data analysis. The way the introduces other constructs that are likely to
data analyses are completed occasionally can be confounded with the variable (mediator) of
foster the view that a critical inuence or me- interest.
diator has been identied. An example in psy-
chotherapy research pertains to integrity or
delity of treatment, that is, the notion that Design Methods for Studying
treatment was carried out as intended. In- Mediators and Mechanisms
vestigators evaluate whether clients who re- Randomized controlled trials. RCTs re-
ceived the treatment as intended show greater main the primary method of demonstrating
change than those who did not (for a review a causal relation between treatment and ther-
see Perepletchikova & Kazdin 2005). Obvi- apeutic change. The most common limitation
ously, if critical procedures of treatment are of RCTs pertinent to this discussion is the
responsible for change, adherence to these failure to establish a timeline between a pro-
procedures ought to make a difference in out- posed mediator or mechanism and outcome,
come. A measure of treatment integrity may as I illustrated above. Assessing the proposed
allow the investigator to delineate the ex- mediator during treatment is necessary but
tent to which clients received the full dose not sufcient to show the timeline between
or proper implementation of treatment. (Re- the mediator and outcome. The assessment of
lated to this article, but not this section, many symptom change is required during treatment
10 Kazdin
as well. Changes in assessment and design of mechanisms of change, one ought to begin
treatment trials, noted later in the chapter, can with the criteria or requirements mentioned
address this concern. above. Statistical analyses and experimental
designs (arrangements) are tools to address
Component analyses of treatment tech- these requirements. One completes the sta-
niques. One way that investigators attempt tistical analysis and then reverts to one of
to get at mediators and mechanisms is by an- the criteria to ask, Was this criterion met?
alyzing a treatment that is known to be effec- Whether the timeline of a supposed mediator
tive. In this context, treatment is considered a or mechanism of change or whether a con-
package, i.e., several distinguishable ingre- struct is a plausible and coherent explanation
dients or components (e.g., x, y, and z). Dis- of therapeutic change are not questions about
mantling studies provide all the components statistical analyses per se but about interpre-
of the package to one group and variations tation of those analyses.
without all of the components to one or more

other groups. The complementary approach
PATHS TO IDENTIFYING AND
is the constructive strategy in which one be-

gins with one component and adds others (to ELABORATING MEDIATORS
other groups or individuals in a crossover de- AND MECHANISMS
sign) (see Kazdin 2003). The idea behind each Understanding mediators and then mecha-
strategy is to identify necessary, sufcient, nisms is not a matter of one study but is a mat-
and facilitative ingredients for treatment to ter of creeping up on the process that draws
achieve change. Under ideal circumstances on a series of projects often seemingly unre-
there is a theory underlying the package with a lated or from different disciplines or types of
more specic statement that some ingredient research. It is useful to think of the search
(e.g., focus on specic activities or exercises) is for mechanisms as a chess game. Even though
essential for therapeutic change. If a compo- there might be a nal winning move (check-
nent is shown to contribute greatly to change mate), the game is won on multiple fronts,
and little or no change occurs without that an integrated sequence of actions, and con-
component, investigators often interpret this verging moves that make checkmate possible.
as evidence for a mediator or mechanism. Critical to a chess game is that there is move-
Identifying a critical component of treatment toward a goal; whether psychotherapy
ment, while valuable for many reasons (e.g., research shows this movement, at least in re-
for extending abbreviated versions of treat- lation to understanding mechanisms, is not so
ment to clinical work), does not provide direct clear. By movement, I refer to a progression
support for a mediator or mechanism. A com- that reects depth or type of understanding of
ponent might achieve its effects for all sorts the factors that produce therapeutic change.
of reasons (processes) that must be assessed. Consider the progression in understand-
Investigators might note that dismantling is ing based on the concepts of correlate, risk fac-
a rst steptrue in principle. In practice, tor, and cause (Kraemer et al. 1997) in relation
there are scores of very informative disman- to the familiar example of cigarette smoking
tling studies that have not moved to the next and lung cancer. The connection between the
steps to understand why a component might characteristic and outcome began as a corre-
be important. late (e.g., cross-sectional study ndings that
people with lung cancer reported a higher
rate of smoking) and moved to a risk factor
General Comments or a correlate where one characteristic clearly
In demonstrating causal relations and identi- precedes another (e.g., longitudinal study
fying candidates that might be mediators or ndings that those who smoked had higher

mortality rates for lung cancer). Then the are conned to correlation. The problem is
connection moved to being identied as a elsewhere, namely, little theory about key
causal factor. Both quasi experiments with constructs (mediators) and how they could
humans and true experiments with animals be studied, little effort to identify steps or
showed that the amount of smoking altered processes (mechanisms) by which the con-
the outcome. The dose-response relation of struct leads to an outcome, and little use
the ndings as well as direct experimenta- of convergent lines of inquiry that could
tion supported the causal role of smoking and strengthen inferences about causes, media-
disease. Once a causal role is demonstrated, tors, and mechanisms. One does not need true
one can ask more analytically how, or through experiment necessarily. One needs to build
what mechanism, does the cause operate? the case by meeting the requirements outlined
Much of research on treatment, but also above. There are many strategies to under-
on psychiatric disorders, identies correlates stand mechanisms or at least to move the ball
and risk factors (predictors) of the outcome. forward signicantly beyond correlation, as I
A difculty is that the work rarely progresses address below.
to the next step that would provide a more in-

depth evaluation of how the factor operates.
Research on heart disease and cholesterol il- Meticulous Description
lustrates the progress and more in-depth eval- Understanding mechanisms is readily framed
uation. Cholesterol has been known to be a as an explanation of some phenomenon of in-
risk factor and predictor of heart disease. Re- terest. In research (and life) one can read-
search progressed to evaluate whether choles- ily distinguish description (what is happen-
terol is causally involved, and indeed, it is. ing) from explanation (why it is happening
Changing (reducing) cholesterol, in fact, al- or through what forces, processes, or mech-
ters the subsequent risk for heart disease. anisms). This is a helpful distinction to learn
Psychological research often moves from and to teach but for this discussion to blur. In
risk factor to causal risk factor casually and many instances in science, one can conceive
without actual tests. In the usual instance, a of description and explanation as related and
study identies, let us say, two risk factors for as opposite ends of a continuum. Depending
some deleterious outcome. The investigator on the detail, level of analysis, and sequence
then suggests that we ought to change, ad- of moving from one to the other, description
dress, or in someway attend to these risk fac- can become explanation. Let us continue the
tors to make people better. This is a non se- example of cigarette smoking and lung cancer.
quitur. We do not know that the risk factors Spanning decades, cross-sectional and lon-
bear any causal relation to the outcome. The gitudinal studies and research with humans
practice of identifying a risk factor and then and animals have established a causal role
discussing an intervention to alter that risk between cigarette smoking and lung cancer,
factor is so common that one looks for sys- which is where we left off in the comments
tem issues that might foster it (e.g., journals above. Establishing a causal relation does not
or funding agencies that insist on clinical im- automatically explain the mechanisms, i.e.,
plications on what might [merely] be a basic the process(es) through which lung cancer
critical nding). comes about. The mechanism has been un-
Psychological research rarely moves the covered by describing what happens in a se-
evaluation from correlate, to risk factor, to quence from smoking to mutation of cells into
causal agent. A usual reason given is that cancer (Denissenko et al. 1996). A chemical
one cannot experimentally manipulate crit- (benzo[a]pyrene) found in cigarette smoke in-
ical variables in humans (e.g., child-rearing duces genetic mutation at specic regions of
practices, attachment style), and therefore we the genes DNA that is identical to the damage
12 Kazdin
evident in lung cancer cells. This nding is erator. As is well known, children with a his-
considered to convey precisely how cigarette tory of physical abuse are at risk for later anti-
smoking leads to cancer at the molecular level. social behavior. Most people who are abused
This is an example of where the what (de- as children do not engage in antisocial be-
scription) can be sufciently ne grained to havior. A genetic characteristic moderates the
convey the how. relationship. Abused children with a genetic
In therapy, proposed mechanisms might polymorphism (related to the metabolism of
encompass such constructs as the therapeutic serotonin) have much higher rates of antiso-
relationship. Research needs to go beyond the cial behaviors than those without this poly-
demonstrated correlation and even the pre- morphism (Caspi et al. 2002). Among boys
dictive portion (i.e., on the assumption that with the allele and maltreatment, 85% de-
the timeline can be rmly established). One veloped some form of antisocial behavior
way to move closer to understanding mecha- (diagnosis of conduct disorder, personality
nisms would be to describe social interaction assessment of aggression, symptoms of adult

outside of the context of therapy in relation personality disorder, or court conviction of vi-
to neurological or other biological indices olent crime) by the age of 26. Individuals with
(e.g., Adolphs 2003, Meyer-Lindenberg et al. the combined allele and maltreatment consti-
2005). What changes take place in social in- tuted only 12% of the sample, but accounted
teraction? There is still a huge leap between for 44% of the cohorts violent convictions.
these descriptions and explaining how a rela- Further research has replicated and extended
tion in therapy leads to symptom change, but the nding by noting that parent neglect as
this is a start and moves beyond where we are well as abuse in conjunction with the polymor-
today in the therapy literature. phism increase risk for conduct problems and
violence (Foley et al. 2004, Jaffee et al. 2005).
So far, this is a fascinating illustration of
Moderators as a Path to Identifying moderation. However, closer scrutiny is help-
Mediators and Mechanisms ful here because it hints at mechanism. Caspi
Moderators refer to characteristics that inu- and colleagues (2002) looked at the allele for
ence the direction or strength of the relation monoamine oxidase A (MAO-A) because:
between an intervention and outcome. For ex- The gene that encodes the MAO-A en-
ample, we know that childhood signs of an- zyme that metabolizes neurotransmit-
tisocial behavior predict later delinquency in ters is linked with maltreatment victim-
adolescence for boys but not for girls, i.e., sex ization and aggressive behavior;
moderates the relationship (Tremblay et al. A rare mutation causing a null allele at
1992). This suggests that different mediators
the MAO-A locus in human males is as-
and mechanisms are likely to be involved in
sociated with increased aggression;
the onset of delinquency for boys and girls.
Animal gene knockout studies show that
The nding is very useful indeed, because any
deleting this gene increases aggression;
search for mediators that combined boys and
and
girls might not nd an effect; a clear effect
for boys might be diluted or nullied by the
Restoring this gene expression de-
absence of any effect among girls. creases aggression.
Moderators can play a more direct role in In one sense we have identied a
elaborating mediators and mechanisms of ac- moderatorthe inuence of an indepen-
tion, and these have yet to be exploited. Con- dent variable (abuse in the home) and out-
sider an example of the effect of experience come (antisocial behavior years later) is inu-
during childhood on subsequent criminal be- enced by some other characteristic or variable
havior, where a genetic characteristic is a mod- (MAO-A allele). Clearly, we have much more

because the work and the results it gener- conditioning of fear in humans and animals.
ated are beginning to point to the genetic and Conditioning as an explanation of fear acqui-
molecular underpinnings. We do not know sition and extinction as an explanation of fear
how the allele and abuse traverse specic steps reduction or elimination are useful paradigms
from a to z in which aggression emerges, but for the processes that might be involved in
we are getting close. For example, recent nd- treatment. Research has suggested that ex-
ings show the neural mechanisms through tinction is not merely unlearning (elimina-
which the genetic inuence is likely to operate tion of a previously established connection)
(Meyer-Lindenberg et al. 2006). The MAO-A because the connection is not erased or lost,
allele is associated with diminished brain cir- but rather is actively suppressed through re-
cuitry related to impulse control that would learning of an acquired inhibition (Myers &
promote aggression. Davis 2002).
The type of moderator work illustrated Understanding the neurological under-
here has some characteristics uncommon in pinnings of extinction has moved to interven-
the usual moderator research in relation to tion research. Conditioning and extinction of
therapy. In the illustration, the moderator was fear depend on a particular receptor in the
identied based on considering mechanisms amygdala (N-methyl-D-aspartate) (see Davis
that might be involved. Theory about poten- et al. 2006). Chemically blocking the recep-
tial mechanisms, prior correlational evidence tor shortly before extinction training blocks
(abuse and victimization), and other studies extinction in animal research, a nding that
indirectly related served as background. In shows a dose-response relation. Blocking the
much of treatment research and moderator receptor after extinction training also blocks
research in clinical psychology more gener- extinction, which suggests that the consoli-
ally, moderators of convenience are used, such dation process can be interrupted. A com-
as information routinely obtained and global pound (D-cycloserine) binds to the receptor
indices (e.g., socioeconomic class, ethnicity, and makes the receptor work better, i.e., en-
comorbidity). There is little sound theory be- hances extinction when given before or soon
hind the research or predictions that derive after extinction training.
from proposing precisely what facets of the The laboratory research has moved to
moderator might be important in explain- therapy trials where exposure therapy, based
ing the relation. Thus, there is a vast liter- on an extinction model, was evaluated to test
ature with analyses showing boys and girls, whether enhancing a mechanism of extinction
younger versus older, and this ethnic group would improve treatment outcome. An initial
versus that ethnic group differ. This is ne as controlled trial was completed with individu-
a start, but much of the research never gets als who suffered acrophobia (fear of heights)
past the start. Moderation can lead to in- (Ressler et al. 2004). Exposure therapy, one
sights about mediation, as the example of ag- of the most well demonstrated treatments for
gression shows, but it requires tests of ideas anxiety, was used as the treatment. The goal
about what the mechanisms are or could be. was to extinguish fear; exposure to heights was
provided in presentations via virtual reality.
Presumably, activation of the critical receptor
Direct Intervention and (with D-cycloserine) would improve the
Manipulation effects of exposure therapy (i.e., augment ex-
Direct manipulation of a proposed mecha- tinction). Indeed, that was found. Participants
nism is of course a powerful way to move our who received the drug (oral administration
understanding forward. Consider the work on two to four hours before each session) showed
fear conditioning and psychotherapy. There greater improvements than those who re-
have been decades of research on Pavlovian ceived a placebo. The results were reected
14 Kazdin
on several measures of avoidance, anxiety, In the 1960s, Patterson and his colleagues
global improvement, and self-exposures to began an extensive research program designed
real-world heights as well as skin conduc- to understand the emergence and mainte-
tance, as a measure of anxiety during and after nance of aggressive child behavior (Patterson
treatment. The effects were evident one week 1982, Patterson et al. 1992). The studies in-
and three months after treatment. The en- cluded directly observing child and parent in-
hanced outcome effects (with D-cycloserine) teraction in the home in a detailed fashion
have been replicated for the treatment of (29 different behaviors and interactions oc-
social anxiety (Hofmann et al. 2006). curring from moment to moment including
The model of the research program, i.e., such behaviors as attending to and unwit-
movement from moderators and mediators tingly reinforcing child deviant behavior, us-
to mechanisms and from basic to applied re- ing commands, delivering harsh punishment,
search, more than two outcome studies needs and failing to attend to appropriate child be-
to be replicated. Understanding mechanisms havior). Among the many interaction pat-

of learning and extinction, but also memory, terns, those involving coercion have received
belief, persuasion, control, stress alleviation, the greatest attention (Patterson et al. 1992,
anticipation, and so on are within empirical Snyder & Stoolmiller 2002). Coercion refers
reach in a similar way. Once such mechanisms to a sequence of parent and child actions
are studied, potential targets can be identied, and reactions that increase the frequency and
with a similar paradigm of manipulating the amplitude of angry, hostile, and aggressive
mechanisms. Manipulation might be through behaviors. The sequence may begin with an
psychological interventions as well as biolog- argument over some action that has or has
ical ones. not been performed. This intensies through
verbal statements (e.g., yelling, screaming) to
more intensive actions (e.g., hitting, shoving).
Converging Lines of Work Ultimately, a high-intensity action of one per-
The prior examples emphasize key elements son (usually the child) ends the aversive be-
of the demonstrations such as studying mod- havior of the other person (usually the par-
erators or intervening directly on an intended ent). Thus through negative reinforcement
mechanism. Actually, the emphases are useful (increase in likelihood of a behavior that ter-
but the examples are part of a broader strat- minates an aversive condition), children are
egy. Multiple lines of evidence are likely to inadvertently rewarded for their aggressive
be needed to converge on precisely what the interactions. Their escalation of coercive be-
mechanism is. The examples I have provided havior is increased in the process, and children
focus on moderators and mechanisms and un- are likely to be more aggressive (more often,
derpinnings that are biological. This is not higher intensity) in the future. The parent be-
a coincidence; the technological advances for haviors are part of the discipline practices that
studying biological processes are astounding sustain aggressive behavior. The interaction
and in some cases, processes (e.g., neurotrans- does not place a single-unidirectional causal
mitter or synapse activity) can be observed relation between the parent and child. Rather,
in real time. Studying mediators and mecha- a dynamic interaction exists in which aversive
nisms and key theses of this review have noth- behavior on the part of both parties escalates
ing inherently to do with biology. The focus and does so in a way that systematically pro-
on mechanisms and the convergence of mul- grams, fosters, and develops greater deviance
tiple lines of work can be gleaned from study- in the child.
ing psychological processes and human inter- The parent-child interaction does not nec-
action, as illustrated in research on parenting essarily determine the next behavior but in-
practices in the homes of young children. creases the probability that the behavior

would move in one direction and toward some causal relations toward understanding me-
end rather than another. Given x (behavior of diators and mechanisms. I have omitted
the parent), y (behavior of the child) is much studies on mediation, which have become
more likely to occur and so on in the sequence. relatively common. The reason for omitting
Conditional probabilities of behaviors were these was explained in the discussion of
used to describe the interactions leading to statistical tests of mediation, namely, the
aggression. I mentioned previously that there studies rarely establish the critical conditions
is a way in which meticulous description can for establishing a timeline and a mediator
blend with and become an explanation. Much is not necessarily a mechanism. When the
of the sequence of interactions was of this timeline is not established, it is even too
type, namely, showing that the interactions much of a leap to imply there is a mediation
fostered aggression and that the timeline was relation beyond a statistical connection in
clear. which the mediator and outcome could be
The studies showing that specic inept reversed.

child-rearing practices contributed to aggres-
sive behavior served as a model for inter-

vention (Reid et al. 2002). Several studies RECOMMENDATIONS FOR
have shown that changing parent-child in- RESEARCH
teractions (via parent management training) Psychotherapy research has a long history of
signicantly reduces aggressive behavior and discussing processes of therapy, but little re-
related conduct problems (see Kazdin 2005, search has addressed the conditions neces-
Reid et al. 2002 for reviews). Thus, a converg- sary to establish mediators or mechanisms.
ing set of studies showed a sequence of coer- In general, the investigation of mediators and
cive parent-child interactions leading to esca- mechanisms of therapy can be improved in
lation, support for a model that explains the several ways. Table 2 lists recommendations
interaction (coercion theory and reinforce- to enhance our understanding of therapeutic
ment leading to escalated aggression), and an change.
intervention that changes putatively critical
parenting processes that controvert aggressive
child behavior. The unfolding of coercive be- Use Theory as a Guide
havior and effective intervention go very far The guiding question for treatment research
to suggest the mechanisms involved in onset is how does treatment achieve change? The
and elimination of aggression in the home, at answer may involve basic psychological pro-
least for some children. cesses (e.g., memory, learning, information
processing) or a broader theory (e.g., moti-
vation). It is no longer sufcient to provide
General Comments global conceptual views (e.g., psychodynamic,
The discussion highlights examples of cognitive, or familial) that foster a treatment
treatment-related research that moves from approach or orientation toward what to do
Table 2 Recommendations for research in the sessions. Rather, to ensure progress,
1. Use theory as a guide
specic conceptual models are needed to ex-
2. Include measures of potential mediators in treatment studies plain those processes that are responsible for
3. Establish the timeline of the proposed mediator or mechanism and therapeutic change. What is needed further is
outcome greater specicity in conceptualizing not only
4. Assess more than one mediator or mechanism the critical construct but also how that oper-
5. Use designs that can evaluate mediators and mechanisms ates to produce symptom change. We need
6. Examine consistencies across different types of studies more than tests of mediation to understand
7. Intervene to change the proposed mediator or mechanism mechanisms. Mediation tests of plausible
16 Kazdin
constructs can provide a screening device of a change in some process results in symptom
sorts to identify potential avenues to be pur- change. As a prerequisite to understanding,
sued in a ne-grained fashion. assessments of potential mediators ought to
It would be helpful for intervention re- be included in treatment studies.
search to identify candidate mediators
and mechanisms or plausible constructs that
would explain or account for (statistically) Establish the Timeline of the
therapeutic change, manipulate the proposed Proposed Mediator or Mechanism
mechanism, assess to ensure it has been ma- and Outcome
nipulated, and then evaluate change. For ex- It is important to establish that the proposed
ample, in relation to tobacco use among mediator is changing before the outcome.
teenagers, several mediators that may serve as The timeline has two requirements: (a) the
useful targets have been identied, including proposed mediator must be assessed before
coping skills of the youth, peer inuences, and the proposed outcome, and (b) the outcome
availability of tobacco, among others (Mac- must also be assessed early to ensure the me-
Kinnon et al. 2002). The targets can be the diator has in fact changed before the outcome.
focus of intervention. If one of these targets Even during the middle of treatment, long
leads to change in tobacco use, this would before the investigator may be interested in
serve as an excellent basis for further work therapeutic change, it is quite possible that
to understand exactly how the inuence pro- improvements occur in the client and these
duces change. We need next-step research improvements come before change in the pu-
that begins with theory but tests directly how tative mediator.
the proposed mediator operates. Assessment is the main change needed in
In research training, there is often a strong research. Assessment on multiple occasions
demand of the investigator to begin with a the- during treatment can provide information on
ory or conceptual model. The study that fol- the timeline of mediators and mechanisms and
lows is a test of that theory. However, the goal outcomes and the possibility of bidirectional
of research is to end up with an understand- changes, i.e., each one inuences the other
ing of how therapy works. This goal can be in some way and at different points. Assess-
achieved by research that generates hypothe- ment on a session-by-session basis (i.e., every
ses and theories in addition to research that occasion over the course of treatment) per-
tests hypotheses. There is far too little re- mits evaluation of the mediator of change and
search that focuses on generating hypotheses symptom reduction and considers individual
from careful observation and on building the- differences in the course of these changes.
ory that can be tested (Kazdin 2003, McGuire
1997).
Assess More than One Mediator or
Mechanism
Include Measures of Potential The accumulation of evidence would prot
Mediators in Treatment Studies from the assessment of more than one media-
The mediator or mechanism ought to be spec- tor in a given study. It is rare that one mediator
ied so it can be measured. Studies occasion- is studied, and hence there may be little value
ally include such measures (Hofmann 2000, in raising the bar even higher by recommend-
Weersing & Weisz 2002), although their ad- ing the assessment of two or more mediators.
ministration has not allowed evaluation of Recommending the assessment of more than
timelines. Yet, measures are available. More one mediator during treatment means that the
ne-grained analyses will be needed to study assessment battery (e.g., how many measures)
the unfolding of processes over time and how will increase as each mediator is added to the

design. In laboratory (efcacy) studies of ther- same time at pre- and post-treatment. With
apy, the addition of one or two measures dur- this variation, conclusions cannot be reached
ing the course of treatment may not be par- about whether improvements in symptoms in-
ticularly onerous. uenced the proposed mediator or vice versa,
The assessment of multiple mediators in or whether both were altered by another
a given study has enormous benets. If two variable.
or more mediators are studied, one can iden- The third design variation assesses symp-
tify if one is more plausible or makes a greater toms at pre- and post-treatment, but during
contribution to the outcome. In addition, the the course of treatment (on one or more oc-
assessment of multiple potential mediators casions) the proposed mediator is assessed.
within individual studies is economically ef- The data analyses then evaluate whether the
cient, given the tremendous amount of time process during treatment contributes to (pre-
and resources needed for any treatment inves- dicts, accounts for) treatment outcome. This
tigation. Across many studies, some mediators research gives a strong but misleading im-
may repeatedly emerge as possible contenders pression that a timeline is established between
while others fall by the wayside. some process (e.g., cognitions, alliance) and
therapeutic change. The failure to measure
symptoms at the same time or indeed before
Use Designs that Can Evaluate the mid-assessment of the supposed media-
Mediators and Mechanisms tor precludes conclusions about whether the
Table 3 lists ve designs that vary in the as- mediator comes before symptom change. Just
sessment of potential mediators or mecha- because symptoms were not assessed in the
nisms of change and treatment outcome. As- middle of treatment does not mean they did
sume all to be RCTs in which treatment is not improve or indeed even improve before
compared with no treatment. The rst and the putative process variable.
most commonly used design variation omits The fourth design variation improves on
assessment of potential mediators. RCTs are the prior designs by including assessment
excellent in demonstrating a causal relation of the proposed mediator and the outcome
between the intervention and therapeutic (symptoms) during treatment. Ideally, there
change. Yet, the designs that resemble the rst will be more than one assessment occasion
variation can say nothing about mediators or during treatment. This variation can evaluate
mechanisms, even though we as authors often the time sequence, i.e., whether changes in
do. In the second design variation, symptoms the mediator preceded symptom change and
and possible mediators are assessed at the whether symptom change preceded change in
Table 3 Outcome study designs and evaluation of mediators or mechanisms

Mechanism assessment Outcome assessment
Design variation Pre During Post Pre During Post
1. Usual outcome design N N N Y N Y
2. Concurrent study of mechanisms and Y N Y Y N Y
outcomes
3. Assessment of mechanism during treatment N Y N Y N Y
4. Assessment of mechanisms and outcomes Y Y Y Y Y Y
during treatment
5. Assessment of mechanisms and outcomes Y Y, Y, Y, . . . Y Y Y, Y, Y, . . . Y
all or most sessions
Y = yes, assessment is conducted; N = no assessment.
18 Kazdin
the putative mediator (which may make the understanding key mechanisms of extinction
mediator an effect rather than a cause). How- has already improved the effectiveness of
ever, if the assessment is only on one occasion extinction-based treatment (Davis et al.
during treatment, it is possible that both the 2005). Therapeutically relevant phenomena
proposed mediator and symptom change oc- (e.g., attachment, separation, social support)
curred or appear to have occurred at the same can be studied in animal research to identify
time. Their relation might not be easy to dis- processes (e.g., changes in the structure or
cern and the possibility exists that a third vari- function of the brain) and their consequences
able led to both changes in the mediator and in behavior. These in turn might direct re-
symptoms. search to plausible underpinnings to support
A disadvantage of the fourth design vari- a conceptual view of the mechanism of ther-
ation is that it presumes that the course of apeutic change. Such tests, far removed from
change for both the mediator and outcome is therapy settings, provide important tests of
captured by measuring each of these at only principle. For example, maternal caregiving
one (or even two) xed points during treat- behaviors (e.g., nursing, licking, grooming)
ment. There could be great variation in when among rats inuence the responsiveness to
the change is made among patients receiv- stress in the offspring; the effects can be seen
ing the same treatment. Both the mediator in behavioral as well as from the neurological
and symptoms may change at different points and endocrine responses of the offspring
among a set of patients. The fth design varia- (e.g., Champagne et al. 2003, Pruessner
tion, an extension of the prior design, provides et al. 2004). This might well be pertinent
a more ne-grained analysis of change in me- to understanding stress, coping with stress,
diator and symptoms and overcomes this con- and interventions designed to ameliorate
cern. Assessments are made so that one can stress.
examine the course of change of the media-
tor and symptoms and can take into account Naturalistic studies. If one is proposing a
individual differences in when the changes mediator of change, is there a sample, popula-
occur. tion, or setting in which this mediator may be
expected to vary naturally, i.e., without inves-
tigator intervention? For example, if changing
Examine Consistencies Across parenting style is proposed to explain why a
Different Types of Studies parent- or family-based treatment of a child
Understanding mediators and mechanisms clinical problem is effective, naturalistic stud-
through which therapeutic change occurs ies examining families with and without these
could prot from different types of studies, practices and the short- and long-term child
beyond those that might be construed as ther- behaviors with which these are associated are
apy research. Conclusions from these studies relevant. Among naturally occurring instances
may be consistent and converge in making a of the process or construct, is there a dose-
particular process plausible. response relation?
As an example and following up on the
Animal laboratory research. Granted, prior example of maternal caregiving among
many of proposed mediators of therapy rats, naturalistic studies of normal moth-
may not be amenable to mouse or zebra ering have revealed that stress reactivity in
sh models. Yet, some of the mediators and human infants is inuenced by maternal
mechanisms of therapy might be studied in caregiving (e.g., sensitivity, availability, lack of
the lab, and we ought not to be shy about intrusiveness) during routine activities (e.g.,
them or shy away from them. I mentioned feeding, meal preparation) very much in keep-
above the work on fear conditioning and how ing with the animal research highlighted

above (Hume & Fox 2006). Low-quality ing who changes and how change unfolds, and
caregiving was associated with greater stress who does not change and what might be op-
reactivity of their infants (e.g., fearfulness, erative there.
more right frontal brain asymmetry), an effect
that could not be explained by infant tempera- Laboratory studies of therapeutic pro-
ment. Caregiving in relation to stress response cesses. Such studies are viewed with ambiva-
and reactivity behaves in a similar way across lence because they do not show whether treat-
different research paradigms and draws atten- ment works in real-life settings. Controlled
tion to mediators or mechanisms that might studies of therapy in research rather than clin-
be pertinent to therapy (e.g., trauma, stress, ical settings are more important now than ever
coping). before. The careful control afforded such re-
Naturalistic studies by themselves may not search is precisely what is needed to identify
permit strong causal conclusions. Yet, such mediators and mechanisms. Translational re-
evidence can be enormously helpful. Many search without knowing what to translate will
advances in understanding cancer, heart dis- have a checkered yield in clinical applications
ease, and stroke began by looking for variation of treatment.

in putative mediators (e.g., in health habits,
diet) among individuals with varied outcomes
(e.g., morbidity, mortality). Observing pro- Intervene to Change the Proposed
cesses that may be operative in the natural Mediator or Mechanism
environment and their short- and long-term An excellent strategy is to conduct an exper-
correlates can be very useful, for both gener- iment in which the proposed mediator is in
ating and testing hypotheses about mediators fact altered or varied across groups, as illus-
and mechanisms. trated in the treatment study cited above on
extinction of fear (Davis et al. 2005). Groups
Qualitative research. Qualitative research randomly composed might be assigned to low,
is an approach to the subject matter of hu- high, and medium levels of a proposed medi-
man experience and focuses on narrative ac- ator (as a general concept) or mechanism (as a
counts, description, interpretation, context, more specic set of steps expected to lead di-
and meaning. Among the key characteris- rectly to the outcome). Strong support would
tics is the in-depth study of the phenomena be evident from ndings that outcome varies
of interest. Individual participants or cases directly as a function of levels of the manipu-
are focused on intensely to examine pro- lated dose.
cesses, meaning, characteristics, and contexts. Intervening to change a mediator is an
Qualitative research is a rigorous, veriable, excellent strategy. Here, too, assessing more
empirical, and replicable set of methodolo- than one mediator would be helpful in under-
gies that encompasses many different disci- standing why change occurred. Intervening to
plines and diverse design, assessment, and alter the mediator and assessing the level of
data-analytic strategies (Berg 2001, Denzin & that mediator (as a check on the manipula-
Lincoln 2005). In the context of the present tion) and two or more plausible, other medi-
discussion, qualitative research might study ators that are not manipulated would be an
the process of therapy, how the patient and elegant way to evaluate mechanisms. In such
therapist experience that process, and what a study, one can rule out or make implausible
might be critical actions or cognitions and some mediators while providing evidence on
how they relate to improvements outside of behalf of another mediator.
treatment. Qualitative research can provide a A variation of the intervention approach
ne-grained analysis by intensively evaluating is worth distinguishing, and I refer to this
the richness and details of the process, includ- here as therapy knockout studies. The term
20 Kazdin
draws from genetic work (e.g., gene knockout lation and have identied the mechanism; in
studies with mice) where a particular gene is others, we know of increased risk. The per-
omitted or altered and the effects are evident vasiveness of the inuence of smoking on so
on behavior or some other facet suspected many conditions can introduce complexities
to be controlled by the gene. The general in the search for mechanisms because so many
model of this research would be a wonder- biological systems are involved. There may be
ful extension to psychotherapy mechanisms. multiple and different mechanisms for the sin-
More specically, if the investigator believes gle agent but different outcomes. On the other
or theory predicts that a specic mechanism hand, some common pathways may exist that
accounts for change, it would be useful to help focus research.
provide the therapy with an added interven-
tion that is designed to knock out (inacti- Multiple influences, single outcomes.
vate) the mechanism. If role-play, practice, or Similar outcomes may be reached through
warm fuzzy relations are critical to the tech- multiple paths. Thus, we do not expect to
nique, give two variations of the treatment: see all people with a particular characteris-
the original and the original with an effort tic (high kindness, bipolar disorder) to have
to inactivate the mechanism. As with any sin- reached these delineations through the same
gle study, supportive evidence that treatment path. There are multiple paths. The paths may
worked wonders only when the mechanism reect similar mechanisms activated by differ-
was allowed to operate could be explained ent experiences or different mechanisms. For
in multiple ways. Even so, this evidence example, low IQ could result from genetic,
would be a superb addition to accumulating prenatal, cultural, and postnatal toxic (e.g.,
evidence. lead) inuences. This single outcome has
many paths. Essential to work on mediators
and mechanisms is distinguishing different
SPECIAL CHALLENGES AND courses or paths and moderating inuences.
OBSTACLES Looking for one explanation or mechanism
There are multiple challenges in considering for one group, one therapy, or one out-
mediators and mechanisms that extend become may yield little. Conceptual work on
yond a few changes in designs or measure- possible moderators and exploratory stud-
ment strategies. Consider some of the key ies (and yes, shing expeditions) followed by
challenges briey. conceptual work will be critical to look for
subgroups.
Mechanism-Outcome Relations Nonlinear relations. We are trained to

The discussion has implied a simple model in think, love statistics that generate, and per-
which a single mechanism leads to a single haps even are victims of a cognitive heuristic
outcome or the effects are strong, simple, and (I call it slippery-slope thinking) that attracts
linear. Yet multiple variants have implications us to linear relations. Many critical relations
for conceptualizing, designing, and interpret- are nonlinear. For example, cholesterol and
ing research. risk for heart disease is positive and linear;
higher cholesterol increases risk. Cholesterol
Single agent (influence), multiple out- and stroke are U shaped, i.e., nonlinear so
comes. One complexity occurs when a sin- that low and high cholesterol increase risk.
gle inuence produces multiple outcomes. Nonlinear relations propose a challenge in the
For example, cigarette smoking leads to sev- sense that dose-response relation (as a linear
eral physical and psychological conditions. In function) is one clue on the path toward mech-
some of these, we know there is a causal re- anisms, although it is not essential. Looking

at means of groups and using statistics that example, a standard dose of psychotropic
evaluate and search for linear relations can medication (e.g., for clinical depression) can
speed or delay progress. We may nd a weak be an overdose or underdose for people of
relation or no relation between an agent and different ethnicities and countries (Lin et al.
outcome (e.g., cholesterol and stroke) for the 1993). Medication effects are moderated by
sample as a whole. Analyses of subgroups and ethnicity. Among the intriguing issues, would
tests of nonlinear relations to identify reliable medication effects operate similarly if doses
patterns of mediator-outcome relations are a were adjusted to each group, or is this not
starting point. merely a matter of dose? Either way, evalu-
ating or analyzing data for an overall (main)
Timing of Change. Assume for a moment effect of medication and ignoring ethnicity
that 10 patients receive identical treatment would lead to a weak effect and not encourage
over the course of 20 sessions and that treat- pursuit of mechanisms.
ment works for each of them for the identical It is possible that the mediator or mech-
reason, i.e., the same mechanism is responsi- anism of change in psychotherapy varies as a
ble for change. It is not likely that the process function of a moderator variable. Searching
of change will follow the identical time course for moderators (a priori or post hoc), test-
so that by session 8, for example, the mecha- ing them (statistical power from dividing of
nism has changed in a critical way and symp- the sample into subgroups), and interpret-
tom change is underway. Indeed, apparently ing them (e.g., is the moderator a proxy for
the timeline of therapeutic change can be al- some other variable?) have their own special
tered by what patients are told about the dura- challenges. Rather than looking for main ef-
tion of treatment prior to beginning treatment fects of an intervention and a uniform mech-
(Barkham et al. 1996). anism of change, we may need to identify
Some patients may make rapid or sudden and characterize subgroups, very much in
gains at a particular point in treatment (e.g., the way that genetic researchers often prot
Tang & DeRubeis 1999). One could say that from looking at special groups and individual
at a given point, some have and some have outliers.
not made change in some qualitative or cate-
gorical fashion. Alternatively, one could con-
sider that the point of therapeutic change for Measurement Development
all individuals is normally distributed with a Mediators and mechanisms in therapy are of-
mean and standard deviation. In either sce- ten discussed but validated measures of key
nario (sudden gains but not at the identical constructs are not readily available. Many ad-
point or normally distributed changes across vances in biotechnology as represented by the
several points), assessment of the mechanism continued advances in neuroimaging have had
is a challenge. Assessment of the mecha- enormous impact on the search for neurolog-
nism at any one or two points in a study ical mechanisms, although these assessments
may not capture when change in the mech- have nontrivial interpretive challenges. If by
anism has occurred. A challenge for research mediators or mechanisms, we will be search-
is ensuring that one can evaluate mechanism ing for psychological explanations of action,
and change that may vary in course among we will need valid measures. In the parent
individuals. training literature, mentioned above, behav-
ioral observations were used to show that
parent-child interactions unfolded in a se-
Everything in Moderation quence leading to (conditional probabilities)
The effects of an intervention may be moder- child aggressive behavior. Among the many
ated in ways that exert enormous impact. For virtues of this work was the assessment of
22 Kazdin
observable actions and charting a sequence everyday life, I have tried one or two myself.
that promotes aggression. This is a quarrel about the necessary assess-
Presumably, many mediators of change ment and design requirements that are in-
begin as broad constructs (e.g., changes in frequently included in research. In addition,
cognition). We need valid measures of such assuming the timeline were unequivocally
constructs and then demonstration of how established, we need next-step research
the constructs operate. There are promising that claries how a relationship in ther-
leads. For example, break down of coping apy leads to symptom change, i.e., through
skills in high-risk situations is associated with what specic steps. These steps need to be
cocaine and other substance abuse. Treatment evaluated.
often targets coping skills as the critical me- Prior research has provided important
diator of change in reducing substance use. groundwork on which future studies could
A role-play measure (Cocaine Risk Response build. For instance, an increasing number
Test) has been developed, evaluated, and in- of studies are including assessments during
tegrated into controlled treatment research the course of treatment (e.g., Beauchaine
to evaluate the mediator (Carroll et al. 1999, et al. 2005, Eddy & Chamberlain 2000, Kolko
2005). Assessment of mediators and mecha- et al. 2000, Kwon & Oei 2003). The de-
nisms raises all of the usual questions in mea- signs used in these investigations represent
surement development. The topic has not a great improvement over prior studies and
been accorded sufcient attention. signal progress in research on mechanisms of
change. Yet, existing studies have attempted to
evaluate only a handful of potential mediators
CONCLUSIONS and mechanisms of change.
Enormous progress has been made in psy- The scientic study of mechanisms of
chotherapy research. This has culminated in change is certainly not an easy path on
recognition of several treatments that have which to embark. A given treatment might
strong evidence in their behalf. Despite this work for multiple reasons. Just as there is
progress, research advances are sorely needed no simple and single path to many diseases,
in studying the mediators and mechanisms of disorders, or social, emotional, and behav-
therapeutic change. It is remarkable that after ioral problems (e.g., lung cancer, attention-
decades of psychotherapy research, we can- decit/hyperactivity disorder), there may be
not provide an evidence-based explanation for analogous complexity in mechanisms for a
how or why even our most well studied inter- given treatment technique or therapeutic out-
ventions produce change. come. Two patients in the same treatment
Many researchers might regard the rather conceivably could respond for different rea-
large body of research on the therapeutic rela- sons. The complexities are critically impor-
tionship as a potential exception. Yet, the vast tant to understand because of a point made
majority of studies rarely rule out the possibil- above, namely, the best patient care will come
ity that the relationship is the result of symp- from ensuring that the optimal variation of
tom change or some other variable rather than treatment is provided. Understanding mech-
a mechanism responsible for it. I am not chal- anisms of treatment is the path toward im-
lenging the importance of relationshipsin proved treatment.
ACKNOWLEDGMENT
Completion of this chapter was facilitated by support from the National Institute of Mental
Health (MH59029).

LITERATURE CITED
Adolphs R. 2003. Cognitive neuroscience of human social behavior. Nat. Rev. Neurosci. 4:165
178
American Psychiatric Association. 2000. Practice guideline for the treatment of patients with
major depressive disorder (revision). Am. J. Psychiatry 157(Suppl. 4):145
Barber JP, Connolly MB, Crits-Christoph P, Gladis L, Siqueland L. 2000. Alliance predicts
patients outcome beyond in-treatment change in symptoms. J. Consult. Clin. Psychol.
68:102732
Barkham M, Rees A, Stiles WB, Shapiro DA, Hardy GE, Reynolds S. 1996. Dose-effect
relations in time-limited psychotherapy for depression. J. Consult. Clin. Psychol. 64:92735
Baron RM, Kenny DA. 1986. The moderator-mediator variable distinction in social psycholog-
ical research: conceptual, strategic, and statistical considerations. J. Personal. Soc. Psychol.
51:117382
Beauchaine TP, Webster-Stratton C, Reid MJ. 2005. Mediators, moderators, and predictors
of 1-year outcomes among children treated for early-onset conduct problems: a latent
growth curve analysis. J. Consult. Clin. Psychol. 73:37188

Berg BL. 2001. Qualitative Research Methods for the Social Sciences. Needham Heights, MA: Allyn
& Bacon. 4th ed.
Brent DA, Kolko DJ. 1998. Psychotherapy: denitions, mechanism of action, and relationship
to etiological models. J. Abnorm. Child Psychol. 26:1725
Burns DD, Spangler DL. 2001. Do changes in dysfunctional attitudes mediate changes in
depression and anxiety in cognitive behavioral therapy? Behav. Ther. 32:33769
Carroll KM, Nich C, Ball SA. 2005. Practice makes progress? Homework assignments and
outcome in treatment of cocaine dependence. J. Consult. Clin. Psychol. 73:74955
Carroll KM, Nich C, Frankforter TL, Bisighini RM. 1999. Do patients change in ways we
intend? Assessing acquisition of coping skills among cocaine-dependent patients. Psychol.
Assess. 11:7785
Caspi A, McClay J, Moftt TE, Mill J, Martin J, et al. 2002. Role of genotype in the cycle of
violence in maltreated children. Science 297:85154
Champagne FA, Francis DD, Mar A, Meaney M. 2003. Variations in maternal care in the
rat as a mediating inuence for the effects of development on behavior. Physiol. Behav.
79:35971
Davis M, Myers KM, Chhatwal J, Ressler KJ. 2006. Pharmacological treatments that facilitate
extinction of fear: relevance to psychotherapy. NeuroRx 3:8296
Davis M, Myers KM, Ressler KJ, Rothbaum BO. 2005. Facilitation of extinction of conditioned
fear by D-cycloserine. Curr. Direct. Psychol. Sci. 14:21419
Denissenko MF, Pao A, Tang M, Pfeifer GP. 1996. Preferential formation of benzo[a]pyrene
adducts at lung cancer mutational hotspots in P53. Science 274:43032
Denzin NK, Lincoln YS, eds. 2005. The SAGE Handbook of Qualitative Research. Thousand
Oaks, CA: Sage. 3rd ed.
DeRubeis RJ, Evans MD, Hollon SD, Garvey MJ, Grove WM, Tuason VB. 1990. How does
cognitive therapy work? Cognitive change and symptom change in cognitive therapy and
pharmacotherapy for depression. J. Consult. Clin. Psychol. 58:86269
DeRubeis RJ, Feeley M. 1990. Determinants of change in cognitive therapy for depression.
Cogn. Ther. Res. 14:46982
Eddy JM, Chamberlain P. 2000. Family management and deviant peer association as mediators
of the impact of treatment condition on youth antisocial behavior. J. Consult. Clin. Psychol.
68:85763
24 Kazdin
Foley D, Wormley B, Silberg J, Maes H, Hewitt J, et al. 2004. Childhood adversity, MAOA
genotype, and risk for conduct disorder. Arch. Gen. Psychiatry 61:73844
Grawe K. 2004. Psychological Therapy. Cambridge, MA: Hogrefe & Huber
He J, Vupputuri S, Allen K, Prerost MR, Hughes J, Whelton PK. 1999. Passive smoking and
the risk of coronary heart diseasea meta-analysis of epidemiologic studies. New Engl. J.
Med. 340:92026
Hofmann SG. 2000. Treatment of social phobia: potential mediators and moderators. Clin.
Psychol. Sci. Pract. 7:316
Hofmann SG, Meuret AE, Smits JA, Simon NM, Pollack MH, et al. 2006. Augmentation
of exposure therapy with D-cycloserine for social anxiety disorder. Arch. Gen. Psychiatry
63:298304
Hollon SD, Beck AT. 2004. Cognitive and cognitive behavioral therapies. In Bergin and
Garelds Handbook of Psychotherapy and Behavior Change, ed. MJ Lambert, pp. 44792.
New York: Wiley. 5th ed.

Holmbeck GN. 2002. Post-hoc probing of signicant moderational and mediational effects in
studies of pediatric populations. J. Pediatr. Psychol. 27:8796
Horvath AO, Bedi RP. 2002. The alliance. In Psychotherapy Relationships That Work: Therapist
Contributions and Responsiveness to Patients, ed. JC Norcross, pp. 3769. New York: Oxford
Univ. Press
Hoyle RH, Smith GT. 1994. Formulating clinical research hypotheses as structural equation
models: a conceptual overview. J. Consult. Clin. Psychol. 62:42940
Hume AA, Fox NA. 2006. Ordinary variations in maternal caregiving inuence human infants
stress reactivity. Psychol. Sci. 17:55056
Ilardi SS, Craighead WE. 1994. The role of nonspecic factors in cognitive-behavior therapy
for depression. Clin. Psychol. Sci. Pract. 1:13856
Jaffee SR, Caspi A, Moftt TE, Dodge K, Rutter M, et al. 2005. Nature x nurture: genetic
vulnerabilities interact with physical maltreatment to promote behavior problems. Dev.
Psychopathol. 17:6784
Kazdin AE. 2000. Psychotherapy for Children and Adolescents: Directions for Research and Practice.
New York: Oxford Univ. Press
Kazdin AE. 2003. Research Design in Clinical Psychology. Needham Heights, MA: Allyn & Bacon.
4th ed.
Kazdin AE. 2005. Parent Management Training: Treatment for Oppositional, Aggressive, and An-
tisocial Behavior in Children and Adolescents. New York: Oxford Univ. Press
Kazdin AE. 2006. Mechanisms of change in psychotherapy: advances, breakthroughs, and
cutting-edge research (do not yet exist). In Strengthening Research Methodology: Psychological
Measurement and Evaluation. ed. RR Bootzin, PM McKnight, pp. 77101. Washington,
DC: Am. Psychol. Assoc.
Kazdin AE, Nock MK. 2003. Delineating mechanisms of change in child and adolescent
therapy: methodological issues and research recommendations. J. Child Psychol. Psychi-
atry 44:111629
Kazdin AE, Weisz JR, eds. 2003. Evidence-Based Psychotherapies for Children and Adolescents. New
York: Guilford
Kazdin AE, Whitley MK. 2006. Pretreatment social relations, therapeutic alliance, and im-
provements in parenting practices in parent management training. J. Consult. Clin. Psychol.
74:34655
Kenny DA, Kashy DA, Bolger N. 1998. Data analysis in social psychology. In Handbook of
Social Psychology, ed. D Gilbert, ST Fiske, G Lindzey, Vol. 1, pp. 23365. New York:
McGraw-Hill. 4th ed.

Kolko DJ, Brent DA, Baugher M, Bridge J, Birmaher B. 2000. Cognitive and family therapies
for adolescent depression: treatment specicity, mediation, and moderation. J. Consult.
Clin. Psychol. 68:60314
Kraemer HC, Kazdin AE, Offord DR, Kessler RC, Jensen PS, Kupfer DJ. 1997. Coming to
terms with the terms of risk. Arch. Gen. Psychiatry 54:33743
Kraemer HC, Stice E, Kazdin AE, Offord DR, Kupfer DJ. 2001. How do risk factors work
together? Mediators, moderators, independent, overlapping, and proxy-risk factors. Am.
J. Psychiatry 158:84856
Kraemer HC, Wilson GT, Fairburn CG, Agras WS. 2002. Mediators and moderators of
treatment effects in randomized clinical trials. Arch. Gen. Psychiatry 59:87783
Kwon S, Oei TPS. 2003. Cognitive processes in a group cognitive behavior therapy of depres-
sion. J. Behav. Ther. Exp. Psychiatry 34:7385
Lambert MJ, ed. 2004. Bergin and Garelds Handbook of Psychotherapy and Behavior Change.
New York: Wiley. 5th ed.

Lin K, Poland RE, Nagasaki G, eds. 1993. Psychopharmacology and Psychobiology of Ethnicity.
Washington, DC: Am. Psychiatr. Press
MacKinnon DP, Lockwood CM, Hoffman JM, West SG, Sheets V. 2002. A comparison of
methods to test mediation and other intervening variable effects. Psychol. Methods 7:83104
MacKinnon DP, Taborga MP, Morgan-Lopez AA. 2002. Mediation designs in tobacco pre-
vention research. Drug Alcohol Depend. 68:S6983
McGuire WJ. 1997. Creative hypothesis generating in psychology: some useful heuristics.
Annu. Rev. Psychol. 48:130
Meyer-Lindenberg A, Buckholtz JW, Kolachana B, Ahmad R, Hariri AR, et al. 2006. Neural
mechanisms of genetic risk for impulsivity and violence in humans. Proc. Natl. Acad. Sci.
USA 103:626974
Meyer-Lindenberg A, Hariri AR, Munoz KE, Mervis CB, Mattay VS, et al. 2005. Neural
correlates of genetically abnormal social cognition in Williams syndrome. Nat. Neurosci.
8:99193
Myers KM, Davis M. 2002. Behavioral and neural analysis of extinction: a review. Neuron
36:56784
Nathan PE, Gorman JM, eds. 2007. Treatments That Work. New York: Oxford Univ. Press. 3rd
ed. In press
Orlinsky DE, Rnnestad MH, Willutzki U. 2004. Fifty years of psychotherapy process-
outcome research: continuity and change. In Bergin and Garelds Handbook of Psychotherapy
and Behavior Change, ed. MJ Lambert, pp. 30789. New York: Wiley. 5th ed.
Patterson GR. 1982. Coercive Family Process. Eugene, OR: Castalia
Patterson GR, Reid JB, Dishion TJ. 1992. Antisocial Boys. Eugene, OR: Castalia
Perepletchikova F, Kazdin AE. 2005. Treatment integrity and therapeutic change: issues and
research recommendations. Clin. Psychol. Sci. Pract. 12:36583
Pruessner JC, Champagne F, Meaney MJ, Dagher A. 2004. Dopamine release in response to
a psychological stress in humans and its relationship to early maternal care: a positron
emission tomography study using [C]raclopride. J. Neurosci. 24:282531
Reid JB, Patterson GR, Snyder J, eds. 2002. Antisocial Behavior in Children and Adolescents: A
Developmental Analysis and Model for Intervention. Washington, DC: Am. Psychol. Assoc.
Ressler KJ, Rothbaum BO, Tannenbaum L, Anderson P, Graap K, et al. 2004. Cognitive
enhancers as adjuncts to psychotherapy. Use of D-cycloserine in phobic individuals to
facilitate extinction of fear. Arch. Gen. Psychiatry 61:113644
Shrout PE, Bolger N. 2002. Mediation in experimental and nonexperimental studies: new
procedures and recommendations. Psychol. Methods 7:42245
26 Kazdin
Snyder J, Stoolmiller M. 2002. Reinforcement and coercion mechanisms in the development

of antisocial behavior: the family. In Antisocial Behavior in Children and Adolescents: A De-
velopmental Analysis and Model for Intervention, ed. JB Reid, GR Patterson, J Snyder, pp.
65100. Washington, DC: Am. Psychol. Assoc.
Tang TZ, DeRubeis RJ. 1999. Sudden gains and critical sessions in cognitive-behavioral therapy
for depression. J. Consult. Clin. Psychol. 67:894904
Tremblay RE, Masse B, Perron D, LeBlanc M, Schwartzman AE, Ledingham JE. 1992. Early
disruptive behavior, poor school achievement, delinquent behavior, and delinquent per-
sonality: longitudinal analyses. J. Consult. Clin. Psychol. 60:6472
Twain M, Warner ACD. 1874. The Gilded Age: A Tale of Today. Hartford, CT: Am. Publ.
Weersing VR, Weisz JR. 2002. Mechanisms of action in youth psychotherapy. J. Child Psychol.
Psychiatry 43:329
Whisman MA. 1999. The importance of the cognitive theory of change in cognitive therapy
of depression. Clin. Psychol. Sci. Pract. 6:3004

Zigler E, Glick MA. 1986. A Developmental Approach to Adult Psychopathology. New York: Wiley

AR307-FM ARI 2 March 2007 14:4
Annual Review of
Clinical Psychology
Contents Volume 3, 2007
Mediators and Mechanisms of Change in Psychotherapy Research

Alan E. Kazdin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 1
Evidence-Based Assessment
John Hunsley and Eric J. Mash p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 29

Internet Methods for Delivering Behavioral and Health-Related
Interventions (eHealth)
Victor Strecher p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 53
Drug Abuse in African American and Hispanic Adolescents: Culture,
Development, and Behavior
Jose Szapocznik, Guillermo Prado, Ann Kathleen Burlew, Robert A. Williams,
and Daniel A. Santisteban p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 77
Depression in Mothers
Sherryl H. Goodman p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p107
Prevalence, Comorbidity, and Service Utilization for Mood Disorders
in the United States at the Beginning of the Twenty-rst Century
Ronald C. Kessler, Kathleen R. Merikangas, and Philip S. Wang p p p p p p p p p p p p p p p p p p p p p137
Stimulating the Development of Drug Treatments to Improve
Cognition in Schizophrenia
Michael F. Green p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p159
Dialectical Behavior Therapy for Borderline Personality Disorder
Thomas R. Lynch, William T. Trost, Nicholas Salsman,
and Marsha M. Linehan p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p181
A Meta-Analytic Review of Eating Disorder Prevention Programs:
Encouraging Findings
Eric Stice, Heather Shaw, and C. Nathan Marti p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p207
Sexual Dysfunctions in Women
Cindy M. Meston and Andrea Bradford p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p233
Relapse and Relapse Prevention
Thomas H. Brandon, Jennifer Irvin Vidrine, and Erika B. Litvin p p p p p p p p p p p p p p p p p p p257
vii
AR307-FM ARI 2 March 2007 14:4
Marital and Family Processes in the Context of Alcohol Use and

Alcohol Disorders
Kenneth E. Leonard and Rina D. Eiden p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p285
Unwarranted Assumptions about Childrens Testimonial Accuracy
Stephen J. Ceci, Sarah Kulkofsky, J. Zoe Klemfuss, Charlotte D. Sweeney,
and Maggie Bruck p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p311
Expressed Emotion and Relapse of Psychopathology
Jill M. Hooley p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p329
Sexual Orientation and Mental Health
Gregory M. Herek and Linda D. Garnets p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p353
Coping Resources, Coping Processes, and Mental Health

Shelley E. Taylor and Annette L. Stanton p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p377
Indexes
Cumulative Index of Contributing Authors, Volumes 13 p p p p p p p p p p p p p p p p p p p p p p p p p p p403

Cumulative Index of Chapter Titles, Volumes 13 p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p405
Errata
An online log of corrections to Annual Review of Clinical Psychology chapters (if any)
may be found at http://clinpsy.AnnualReviews.org
viii Contents

Kazdin 2007 PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Kazdin 2007 PDF

Загружено:

Авторское право:

Доступные форматы

ANRV307-CP03-01 ARI 20 February 2007 18:34

Mediators and Mechanisms

Department of Psychology, Yale University, New Haven, Connecticut 06520-8205;

Annu. Rev. Clin. Psychol. 2007. 3:127 Key Words

This articles doi: Abstract

Multiple Criteria . . . . . . . . . . . . . . . . . 4 Include Measures of Potential

RESEARCH ON MEDIATORS Establish the Timeline of the

INTRODUCTION Discussions and theory about why psy-

Moderator refers to some characteristic

Table 1 Key terms and concepts

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 3

if not eliminated. studies, as I elaborate below.

Specificity. The second criterion refers to Gradient. Showing a gradient in which

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 5

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 7

OVERVIEW OF METHODS FOR

The interrelations of A, B, and C, as eval- Even when the timeline is established,

least the cognitions of interest to the inves-

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 9

without all of the components to one or more

is the constructive strategy in which one be-

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 11

to the next step that would provide a more in-

nisms would be to describe social interaction assessment of aggression, symptoms of adult

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 13

to be replicated. Understanding mechanisms havior). Among the many interaction pat-

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 15

The studies showing that specic inept reversed.

sive behavior served as a model for inter-

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 17

Table 3 Outcome study designs and evaluation of mediators or mechanisms

Y = yes, assessment is conducted; N = no assessment.

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 19

ease, and stroke began by looking for variation of treatment.

Mechanism-Outcome Relations Nonlinear relations. We are trained to

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 21

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 23

growth curve analysis. J. Consult. Clin. Psychol. 73:37188

New York: Wiley. 5th ed.

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 25

New York: Wiley. 5th ed.

Snyder J, Stoolmiller M. 2002. Reinforcement and coercion mechanisms in the development

of depression. Clin. Psychol. Sci. Pract. 6:3004

www.annualreviews.org Mediators and Mechanisms of Psychotherapy 27

Contents Volume 3, 2007

Mediators and Mechanisms of Change in Psychotherapy Research

John Hunsley and Eric J. Mash p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 29

Marital and Family Processes in the Context of Alcohol Use and

Coping Resources, Coping Processes, and Mental Health

Cumulative Index of Contributing Authors, Volumes 13 p p p p p p p p p p p p p p p p p p p p p p p p p p p403

Вам также может понравиться