Behaviour Research and Therapy 38 (2000) 103117

www.elsevier.com/locate/brat

Invited essay
Can depression be de-medicalized in the 21st century:
scientic revolutions, counter-revolutions and the magnetic
eld of normal science
Neil S. Jacobson*, Eric Tomas Gortner
Department of Psychology, University of Washington, 1107 NE 45th Street, Suite 310, Seattle, WA 98105-4631, USA
Received 11 December 1998

Abstract

This article is about our scientic investigations of the change mechanisms in cognitive therapy (CT)
for depression. In a previous clinical trial, we found that so-called `cognitive' interventions were not
necessary for the success of CT: the behavioral activation (BA) component, a treatment precluding
attempts to change thinking, worked as well as the entire CT package, both in maximizing acute
treatment response and in relapse prevention over a two year period. We tentatively suggested at the
time of publication [Jacobson, N. S., Dobson, K. S., Truax, P. A., Addis, M. E., Koerner, K., Gollan,
J. K., Gortner, E. T., & Prince, S. E. (1996). A component analysis of cognitive-behavioral treament for
depression. Journal of Consulting and Clinical Psychology, 64, 295304; Gortner, E. T., Gollan, J. K.,
Dobson, K. S., & Jacobson, N. S. (1998). Cognitive-behavioral treatment for depression: relapse
prevention. Journal of Consulting and Clinical Psychology, 66, 377384.] that the `cognitive' components
of CT may not only be unnecessary but potentially a liability, since they result in a less parsimonious
treatment package that may be not be cost eective. In this article, we not only defend this contention,
but counteract the skepticism expressed by some CT advocates that the quality of our CT was decient.
Finally, we describe a study designed to conrm our conclusions from the earlier trial and, in the
process, reintroduce a contextual perspective on depression, one which counters the currently dominant
defect models reected in both Beck's cognitive model and in theories that emphasize biological
causation. # 2000 Elsevier Science Ltd. All rights reserved.

* Corresponding author.

0005-7967/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 5 - 7 9 6 7 ( 9 9 ) 0 0 0 2 9 - 7
104 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117

1. Introduction

Like many readers of this journal, we fancy ourselves clinical scientists, whose job it is to
search for truth and follow our data wherever those data may lead. At the same time, we are
well aware of the all too human fallibilities that produce allegiance to an a priori position, and
complicate the sober appraisal of ndings that are inconsistent with the most cherished beliefs
in the profession (Jacobson, in press). Although our laboratory has been engaged in the
scientic study of depression for almost 15 years, we nd ourselves in the midst of an
unanticipated controversy, one that was certainly not our guiding motive when we began our
work. This controversy reects the constant tension in clinical science between the reinforcers
associated with `being right' and the obligations to let ndings speak for themselves, regardless
of whether such ndings point in a direction that proves one's original hypotheses.
This tension is not new to our laboratory. After Jacobson documented the eectiveness of
behavioral couple therapy (BCT; Jacobson & Margolin, 1979) in a series of randomized clinical
trials (RCT's; Jacobson & Addis, 1993), BCT became the rst empirically supported treatment
for couples in the history of psychotherapy research. However, after our initial enthusiasm, we
began to look more carefully at the eectiveness of BCT. We discovered that although BCT
proved to be more eective than nothing, the rate of improvement was not clinically signicant
(Jacobson et al., 1984). Despite heartfelt allegiance and investment in the BCT model, as
scientists we were compelled to follow a fundamental rule of clinical research: let the data
speak. And this rule served us (and science, as a whole) well; it motivated us to develop a new
theory and, with the help of Andrew Christensen, spurred us on to develop a whole new
treatment model (Jacobson & Christensen, 1996; Christensen & Jacobson, in press), which is
now facing the same unbiased scrutiny that BCT was subjected to.
In the eld of depression, the data are not always allowed to speak for themselves (Jacobson
& Hollon, 1996): included among the plethora of creative thinkers in this eld are those with a
gift for coming up with post hoc arguments to discount ndings that are inconsistent with
prior cherished beliefs, thus leaving normal science intact. Jacobson and Hollon (1996) had
previously noted the facility with which disease model advocates interpret ambiguous ndings
in a way which arms their favorite anti-depressant medications and in the process discounts
the potential of psychosocial interventions. Since then, we have come to learn that our
biological colleagues do not have a monopoly on the post hoc defense. Since our ndings
became controversial (Jacobson et al., 1996; Gortner, Gollan, Dobson & Jacobson, 1998), we
have been on the defensive ourselves. Here is how it happened.

2. Our component analysis of cognitive therapy (CT): an attempt to uncover mechanisms of

change

We began a study in 1990 (Jacobson et al., 1996) attempting to test the cognitive theory of
change in depression put forth by Beck, Rush, Shaw and Emery (1979). The idea for this study
came from an earlier experiment where CT had served as a control group to evaluate BCT for
co-existing depression and marital discord (Jacobson, Dobson, Fruzzetti, Schmaling & Salusky,
1991; Jacobson, Fruzzetti, Dobson, Whisman & Hops, 1993). Jacobson was introduced for the
 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117 105

rst time to state-of-the-art CT, as dened by expert trainer and supervisor Keith Dobson,
who had helped train and supervise cognitive therapists in the seminal Treatment for
Depression Collaborative Research Program (TDCRP; Elkin et al., 1989). As Jacobson listened
to tapes of our therapists performing CT, he was struck by both the power of the treatment
and the speed at which clients often showed a reduction in depressive symptoms. In many
cases, clients did most of their changing during the rst phase of therapy, before the so-called
`cognitive' interventions had even begun. We began to learn rst-hand that CT was multi-
faceted and included an abundance of straightforward behavioral activation techniques, in
addition to: (a) interventions to modify automatic dysfunctional thinking (AT) and (b) those
interventions aimed at modifying core beliefs (a.k.a. schemas or cognitive structures), which are
considered to be the most crucial techniques in CT. The theory of change put forth by Beck et
al. (1979) suggested that the schema-focused interventions were essential to change these
cognitive structures, and such change was viewed as the primary active ingredient accounting
for the relapse prevention potential of CT.
However, after familiarizing ourselves with the treatment, we considered two additional
hypotheses to be equally viable in accounting for the eects of CT: (1) an activation
hypothesis, where the clients were brought into contact with anti-depressant forces in the
environment through the early attempts to prepare them for later cognitive interventions and
(2) a coping skills hypothesis, where clients learned coping methods for dealing with
dysfunctional thinking by engaging in tasks such as completing dysfunctional thought records,
and empirically testing the negative beliefs that were connected with dysfunctional thinking.
In order to test these competing theories of change, we developed two component
treatments: (1) behavioral activation (BA), where only interventions designed to activate
subjects were allowed, without any attempts to change thinking and (2) an automatic thought
(AT) treatment where, in addition to activation, the therapist was allowed to attempt
modications of thinking. However, in neither AT nor BA were therapists allowed to attempt
changes in core dysfunctional schema. The total CT package was oered to only one third of
the subjects, so that the cognitive interventions designed to change core schema and underlying
assumptions could be tested for their power in preventing relapse as well as maximizing acute
treatment response.
In short, depressed outpatients were randomly assigned to one of the three treatments: BA,
AT or CT. The therapists were all cognitive behavioral in orientation and had been trained in
some version of that model. Although only one of them had actually received training in
Philadelphia at Beck's Center for Cognitive Therapy, competence ratings (based on the widely
used Cognitive Therapy Scale (CTS; Vallis, Shaw & Dobson, 1986)) provided by Dobson
indicated that all four of the therapists were performing at a high level of skill.
Dobson's primary responsibility in the collaboration with Jacobson was to ensure that CT
received a fair test. In other words, we specically brought Dobson into the study to control
for any possible allegiance eect resulting from our behaviorist leanings. Dobson was indeed
well-equipped to provide optimal training for the CT component of our study: he had edited
the authoritative Handbook of Cognitive therapy; published an inuential meta-analysis
demonstrating the ecacy of CT (Dobson, 1989) and helped ensure quality control in the
TDCRP. Moreover, he was a true believer in CT and was highly regarded enough by the CT
community to supervise and train therapists in the most important depression research trial to
106 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117

date, the TDCRP (Elkin et al., 1989). We were delighted to have the opportunity to
collaborate with Keith Dobson because, within the CT community, he is one of the true
luminaries.
Although Jacobson was a bit concerned that Dobson had an allegiance for CT and therefore
might bias the therapists, Dobson conducted himself as a true scientist throughout the
collaboration and went out of his way not to bias the therapists. Unfortunately, Dobson's faith
in CT was contagious and we found ourselves with a pool of therapists that were so biased in
favor of CT that they complained whenever they were assigned cases in either BA or AT
conditions. Our supervision meetings therefore required an inordinate amount of morale
boosting to keep their spirits up so that they would try their hardest despite their conviction
that BA and AT were inferior treatments.
All of our hard work with the therapists paid o. As we documented in our original report
(Jacobson et al., 1996), the therapists were remarkably good at sticking to the assigned
protocols. In order to document treatment integrity, we used the Collaborative Study
Psychotherapy Rating Scale designed for the TDCRP. The results of adherence ratings showed
that virtually no cognitive interventions slipped into BA cases and no schema-focused
interventions slipped into either the BA or AT conditions. Importantly, we also documented
that CT included a great deal of schema-focused activity, which is a necessary precursor for
establishing that CT was delivered competently (Waltz et al., 1993). Finally, Keith Dobson's
ratings documented that CT had been delivered in a competent fashion.
Despite our success in keeping the treatments distinct and despite the competence with which
CT was implemented, there were no signicant dierences between treatments at the end of the
acute treatment phase. In other words, BA was just as eective as AT, and both component
treatments worked as well as the full CT package. We looked at the data in numerous ways
and measured recovery using the consensus criteria recommended by leaders in the eld (Frank
et al., 1991). Our measures of recovery rates were derived from the LIFE interview (Keller et
al., 1987); the Beck Depression Inventory (BDI); and the Hamilton Rating Scale for
Depression (HRSD). In addition, parametric analyses were performed on both the BDI and
the HRSD. We thoroughly analyzed the data, providing numerous opportunities for CT to
demonstrate its superiority. Yet, no matter how we dened recovery, and no matter which
analyses were performed, both BA and AT yielded outcomes that were equivalent to CT.
Considering the allegiance of the therapists, their competence in CT and the documentation of
adherence to treatment protocols, our initial report strongly suggested that CT was no more
eective than either component treatment.
However, the true test of CT was its ability to prevent relapse: Beck's theory implies that
schema focused interventions should have been most indispensable in bringing about lasting
change (Gortner et al., 1998). Yet, at the two year follow-up our survival analyses indicated
that, if anything, trends favored BA, not CT, in preventing relapse. Although these dierences
were not statistically signicant, there was certainly no evidence that CT had any advantage
over the other treatments.
Furthermore, we compared the treatments throughout the follow-up phase on a measure
which could be considered more sensitive to overall functioning than survival analysis: the
number of well weeks. This measure, reecting the total number of weeks during the two year
follow-up that subjects functioned with minimal or no depressive symptoms, has been frowned
 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117 107

upon by some investigators because it has the potential to be misleading: once a patient has
relapsed, they are likely to seek further treatment, thus potentially contaminating eorts to
determine treatment eects beyond the point of initial relapse. Only prior to the rst relapse
can observed dierences between treatment groups be unequivocally attributed to the
treatments, free of the confounding inuences created by returning to treatment. This is the
logic for relying on `time to rst relapse' as the barometer for assessing relapse prevention.
However, we were concerned about a dierent, and potentially more serious problem with
outcome measures such as `time to rst relapse'. Suppose that subject No. 1 had a month-long
relapse shortly after treatment ended, but recovered from that relapse without further
treatment and remained free of depression for the remainder of the two year follow-up.
Compare this subject with another hypothetical subject, No. 2, who had his rst relapse 18
months after treatment ended and stayed depressed for the entire 6 month period culminating
in the two year follow-up. Subject No. 1 actually functioned better if we take the entire two
year follow-up period into account, because he was only depressed for 1 month out of 24. In
contrast, subject No. 2 was depressed for 6 of the 24 follow-up months. Yet, survival analysis
favors subject No. 2 because the rst relapse occurs 18 months after treatment termination. In
our view, the subject's functioning over the entire two year period is more important than
when they have their rst relapse. Furthermore, it is an empirical question whether there is any
systematic relationship between the seeking of treatment and survival time. In fact, we found
virtually no correlation between the two in our study. Therefore, the number of well weeks was
a more sensitive index of functioning during the two year period than `time to rst relapse'.
Yet, even on this measure, at no follow-up point, and on no measure of depression severity,
did CT outperform either of its component treatments.
We even used traditional mutivariate analysis of variance to search for evidence that CT was
superior to its components. Both survival and well weeks analysis were based on only a subset
of the intent to treat sample: those who remained in the study and improved enough during
the course of treatment to be considered treatment responders. Since randomization is virtually
never preserved when follow-up analyses are limited to treatment responders, it is technically
inappropriate to base group comparisons on statistical procedures which assume random
assignment. Moreover, analyses based solely on treatment responders tell us nothing about the
ultimate fate of those who were non-responsive during the acute treatment phase. Therefore,
we performed traditional MANOVA's which compared the treatment groups every six months
on measures of depression severity regardless of whether they were classied as positive
responders at the end of acute treatment. These analyses preserved random assignment and
provided unique information about the entire sample. We examined how subjects in all three
treatment conditions were functioning at 6, 12, 18 and 24 months after the end of acute
treatment. By examining changes in functioning from post-test to each of the follow-up points,
we were able to determine whether treatments dierentially eected changes in depressive
symptom severity. Once again, the three treatment conditions were equivalent at each follow-
up point, for each measure of depressive symptoms. By comparing the treatment conditions in
degree of change from pre-test to follow-up, we were able to examine the ultimate impact of
therapy at each follow-up point, taking into account both acute and long-term treatment
response. The equivalence of the three treatments was apparent, even in these parametric
analyses.
108 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117

The conclusions based on this wide variety of analyses were inescapable: both BA and AT
performed as well as CT, in both the short and long-term. This was true despite a group of
therapists with a clear allegiance to CT and a supervisor with a similar bias. We have ruled out
the possibility that the null ndings were due to unwanted overlap between treatment
conditions and our low (8%) drop-out rate during acute treatment makes it unlikely that our
results were inuenced by dierential attrition. The relapse prevention data provide an even
more convincing argument that the component treatments performed as well as CT did,
because Beck's theory of change suggests that CT is particularly likely to pay o in its relapse
prevention eects: only the full CT package contained those interventions directly designed to
change the proposed active ingredient in preventing relapse. Since over 90% of our subjects
were retained throughout the follow-up period (the exact percentage depended on our
denition of `acute recovery'), it is highly unlikely that dierential attrition suppressed any
relapse prevention eects.
Although the results of this trial, like those of any clinical trial, have to be replicated, they
call into question the necessary and sucient conditions for long-term recovery in CT. They
also raise questions about the theory of change put forth by Beck and his colleagues to explain
the change mechanisms of CT. Finally, the ndings suggest the possibility that a simpler, more
parsimonious treatment for depression, BA, might be at least as eective as the more
complicated, multi-faceted CT treatment package. Since there are fewer intervention options in
BA than there are in CT and a distinct dierence in the apparent amount of training time
required (see below), the potential value of BA includes the apparent ease with which it can be
taught to therapists with less training and experience than that required for CT. Our study
suggested the arrival of a promising new psychotherapy for depression: behavioral activation.

3. Reactions by and to our critics

The rst article describing our study was published as a Special Feature of the Journal of
Consulting and Clinical Psychology (Jacobson et al., 1996). To us, this was both gratifying and
somewhat surprising. It was gratifying that both the Editor and the reviewers seemed to have
such a positive reaction to the study. However, it was also surprising, since the study had a
serious methodological limitation: in the absence of a control group, it was impossible to know
for sure whether all three of our treatment conditions were equally eective or equally
ineective. In retrospect, it appears that the potential theoretical and clinical ramications of
the ndings were more signicant to the referees than the methodological weakness. In
particular, they have praised us for going to such extraordinary lengths to rule out alternatives
to the null hypothesis, in anticipation of null ndings: the strong evidence of treatment
integrity; the data on the therapists' competence in CT; the minimal drop out rate during acute
treatment and the almost perfect retention rate during the follow-up period. Moreover, in the
Gortner et al. (1998) report on our relapse prevention ndings, we carefully compared our
outcomes for CT with those found in previous clinical trials. We demonstrated that our CT
outcomes, both during acute treatment and at various points during the follow-up, compare
favorably to those from previous clinical trials. These comparisons are extremely valuable
because they further support the hypothesis that BA's strong showing was not the result of CT
 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117 109

under-performing relative to other trials. Instead, our results demonstrated that BA held its
own against a widely-touted psychosocial alternative to anti-depressant medication: cognitive
therapy. BA worked well, not because CT performed less impressively than it usually does, but
despite the strong showing of CT.
However, there were concerns expressed to us, mostly among self-described advocates of CT.
These critiques were not made frequently, but they were presented to us in ways which made it
dicult to respond: sometimes they were second-hand verbal reports; at other times they were
anonymous reviews to either grant proposals or journal articles and, on still other occasions,
the comments were o-hand and casual remarks during panel discussions at scientic meetings.
It appears to us that CT has become so popular that many of our clinical research colleagues
simply didn't like our results and the knee-jerk reaction was to discredit them. We are
delighted to have a forum in this article for discussing the issues raised in these critiques,
because the questions raised about our study go far beyond the specics of comparisons
between BA and CT and raise important issues pertinent to any clinical science.
The most common concern we heard came from some CT advocates, who expressed
skepticism that our CT was performed competently. It seems unlikely that this skepticism
would have arisen had CT outperformed its component treatments; but, given the faith that
CT advocates have in their treatment, they quite naturally wondered whether it had been
adequately tested in our study. However, we continued to hear these suspicions raised even
after we documented that our cognitive therapists were competent.
However, since CT did not perform as well as many expected in the TDCRP (Elkin et al.,
1989), the quality of supervision in that study and the capabilities of the supervisors, were
questioned by some. Keith Dobson was the supervisor of our therapists and also one of the
supervisors in the TDCRP. When our results were published, we heard some of these same
criticisms.
These post hoc arguments were new to us. For example, in the eld of Behavioral Couple
Therapy, there is no precedent for raising questions about clinical competence based on the
results of a study. If someone is considered to be an expert, that assessment doesn't change
simply because the treatment doesn't perform well. When a pattern of negative results began to
emerge in the eld of BCT we questioned the treatment, not those who implemented the
treatment. This is how normal science should guide the evaluation and development of
treatments: methodology should dictate our evaluation of treatment implementation, results
should not. Were it otherwise, one's own favorite treatment could never be held accountable to
results: poor showings would always nd refuge behind allegations of incompetent treatment
implementation.
The methodology for documenting adherence and competence in our CT component study
was based on guidelines we ourselves had developed (Waltz et al., 1993) and these guidelines
hold investigators to a very high standard. Few studies documented treatment integrity as
thoroughly as ours did. However, our study's results appear to have been provocative to those
with strong opinions about the ecacy of CT and, consequently, our results, as opposed to our
methodology, have inappropriately become the arbiter of criticisms leveled at our
implementation of CT. It doesn't seem reasonable to question the competence of treatment
implementation because the results turned out a certain way, especially when the trainer and
supervisor is someone whose credentials are as strong as Dobson's. Nor does it seem
110 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117

reasonable to use the ndings of the TDCRP to question supervisor qualications: they were
chosen for the job because of their qualications and the only basis for criticism is the
relatively weak performance of CT in that study. Thus, in both instances, supervisors were
chosen because of their qualications for the job and then later criticized because CT did not
perform as well as advocates might have hoped. It seems clear to us that a viable alternative
interpretation could be used to explain both the disappointing (to advocates of CT) ndings of
our study and in the TDCRP: perhaps CT is less eective than those invested in the model
might have wished.
More recently, in the process of applying for extra-mural funds to replicate our previous
study, we began to hear other concerns; not just about the quality of CT in our previous trial,
but also about whether or not CT can be trusted in any study where the therapists are not
either under the supervision of its originators in Philadelphia or the therapists have received
two years of intensive training by someone recognized as an expert by the Center in
Philadelphia. Let us be clear about the source of this concern: it was never Aaron T. Beck
himself and it would be inaccurate to attribute this suggested training requirement to anyone
currently associated with the Center for Cognitive Therapy in Philadelphia. Nevertheless,
whomever the source, this is a standard that is far more exacting than any we can recall in the
annals of psychotherapy training, with the exception of requirements in psychoanalytic
institutes. However, this would not concern us if the standards were justied. In our view, they
are not. First, there is no empirical basis for them: that is, there is no evidence that competence
as a cognitive therapist varies depending on the proximity to Philadelphia of the training or
supervision; nor is there evidence that outcomes improve with greater amounts of training.
Second, if true, the standards would virtually disqualify CT from playing a major role in the
mental health delivery system. If competence in CT for depression requires two years of
intensive supervision from one of a handful of recognized experts, that means at any point in
time, there will only be a small group equipped to provide this treatment. In other words, in
insisting upon such standards in the absence of empirical evidence supporting them, CT
advocates are not only invalidating most previous CT research, they are arguing themselves
out of business. No treatment requiring such intensive training and supervision could be
transportable enough to play a major role in the mental health delivery system of this or any
other country.
As we and our colleagues have already expressed skepticism about the relationship between
clinical training and psychotherapy outcome (Christensen & Jacobson, 1993), it is probably not
surprising to hear that we are skeptical about these training requirements suggested by some
CT advocates. In contrast to these training requirements, let us harken back to the
development of marital treatments. Jacobson co-authored the rst treatment manual that was
used in BCT research (Jacobson & Margolin, 1979). However, once this manual was
disseminated and at least a dozen investigators throughout the world used it or derivations of
it in their own clinical trials (Hahlweg & Markman, 1988; Jacobson & Addis, 1993), no one,
including Jacobson, ever questioned the results of outcome studies because he had not been the
one to provide the quality control. We were further struck by the contrast between the
concerns raised about the quality of our CT and the lack of similar concerns in the eld of
anxiety disorders, where literally hundreds of outcome studies had been done based on
techniques originally developed by Wolpe (1958). No one, to our knowledge, had required or
 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117 111

even hinted that a study of desensitization or other exposure techniques for phobias was
tainted because Wolpe himself, or those who worked with him, had not provided the training.
It seemed to us an arbitrary and scientically indefensible criticism, especially given the
recognized expertise of Keith Dobson. Nevertheless, we wanted to quell concerns about our
study, and in doing so we entered the Pandora's box of competence certication.

4. The politics of competence certication

Responding to continued questioning about the quality of our CT, we subjected our tapes to
the ratings of outside experts, recognized master CT therapists who had no other connection to
the study. After consulting with the Center in Philadelphia, we identied two experts who
seemed satisfactory to all concerned. Each of them rated a randomly selected portion of our
tapes on the Cognitive Therapy Scale, which is an instrument specically developed to measure
competence in CT. Table 1 shows the inter-rater reliabilities (computed through intra-class
correlation coecients) of the three competence raters: the two outside experts and Keith
Dobson. We were stunned by the results: the ratings conducted by the outside raters were
virtually uncorrelated with each other, or with those of Keith Dobson. There was wide
variability in the ratings provided by the experts; therefore, restriction of range could not
explain these remarkably low reliability coecients.
As we examined the reliability of these ratings more closely, both by item analysis and by
looking at sub-scales within the CTS, we kept coming to the same conclusion: either the CTS
was an unreliable instrument or there was no agreement among cognitive therapy experts as to
what competent cognitive therapy was.
Unfortunately, our attempt to clarify the issue of competence through outside expert CTS
ratings only revealed further aws in the technology and conceptual backdrop to measuring
therapeutic competence. Indeed, the measurement of competence remains a murky,
controversial area within the CT community. Recently, Hollon (personal communication,
October 12, 1998) reported similar problems in getting expert raters to agree. However,
Dobson was one of the raters in this new study and Hollon remarked on the close
correspondence between Dobson's ratings and those of Hollon, himself a recognized leader in
CT. Although the jury is still out on the ultimate utility of the CTS, we took comfort in the
reliability between Dobson and the experts organized by Hollon. This suggested that Dobson's
competence ratings in our study were trustworthy after all. Thus, Hollon's work in progress
provides good news for us, because it appears to rule out one more interpretation (poor

Table 1
Reliability of CTS competence ratings

Expert 1 (osite) Expert 2 (osite) Expert 3 (Keith Dobson)

Expert 1 (osite) 1.00 0.07 (N = 25) 0.01 (N = 30)

Expert 2 (osite) 1.00 0.08 (N = 33)
Expert 3 (Keith Dobson) 1.00
112 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117

treatment implementation) which competes with the one we favor (BA is at least as eective as
CT).
From the vantage point of a research laboratory outside of the mainstream CT `world', the
reactions to our ndings have the appearance of political advocacy, as opposed to the open-
minded and considered discussion that science normally facilitates. Our impression is that some
within the CT community are inclined to explain away unappealing ndings with many an
unsupported assertion, if that is what it takes to maintain the viability of CT. Importantly,
Beck himself does not seem to share these concerns: as he indicated in a letter to Jacobson on
November 11, 1998, he has the utmost condence in both the quality of cognitive therapy
conducted in our previous trial; the abilities of Keith Dobson as a supervisor and the
methodology of the trial. The critics, it seems, have dug themselves a hole: given the alleged
need for intensive and long-term training, CT will have to work substantially better than less
expensive treatments to justify its existence. For example, we estimate that therapists with no
experience in BA can do it competently as long as they are closely supervised for the rst 6
months. Of course, both their assertion (that competent CT requires long-term training and
supervision) and ours remain empirical questions. However, they will be tested in a study that
is currently underway.

5. Toward a more cost eective method of treatment depression: behavioral activation

In response to the praise as well as the criticism we received for our last trial, we tried to
design a replication and extension that would more denitively test the potency of these brief
psychosocial treatments for depression. We modied our design in several ways, partly to
overcome some of the limitations that we ourselves perceived in our previous trial and partly
to respond to the assertions of both CT and pharmacotherapy advocates. We want to be sure
that, by the end of this study, we will have provided tests of our comparison treatments that
are above reproach, even by the most die-hard advocates of those approaches.
First, we chose a group of cognitive therapists that we have since labeled the `Dream Team'.
All three of them have many years of training and supervision, two directly by the Center for
Cognitive Therapy in Philadelphia. The third has spent time getting additional training from
the Center, but even prior to that training has many years of experience and is widely
recognized as a cognitive therapy expert.
Second, we reconceptualized the BA treatment so that, instead of just being a component of
CT, it is designed to be a bona de treatment in its own right. We kept the same set of
treatment options, but created a behavior analytic theoretical framework to aid in the decision
rules for guiding therapists in their intervention targets. The behavior analytic framework
emphasizes functional analyses of the environmental events that have impinged on individual
clients to generate the depression, and formulates cases in a way that looks outside rather than
inside the person for targeting change. That is, instead of emphasizing faulty thinking, our
revisionist BA treatment conceptualizes depression in terms of environmental events that have
created contextual shifts, which in turn have denied the client access to those reinforcers which
normally functioned as anti-depressants. These contextual shifts have to be corrected through
targeted activation, activation aimed at altering the environment in such a way that anti-
 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117 113

depressant reinforcers are more easily accessed. People who have lost their jobs are aided in
nding employment. Those who have lost their friends or lovers are aided in nding new
people. Those with relationships that have gone sour are aided in xing them.
In short, rather than general activation aimed at increases in self-dened pleasant events, in
our new BA treatment activation strategies follow directly from the functional analyses.
Moreover, individuals are taught the skills of conducting their own functional analyses, as
ways of providing for coping methods in future cases where the environment has failed them.
In direct contrast with CT, where individuals are taught to think dierently in an eort to cope
with negative circumstances, BA assumes that negative thinking is a realistic by-product of
stressful environments and that the thinking will change automatically as reinforcers return to
the client's life.
This functional analysis of how depression is created does not deny the importance of
diatheses in vulnerable individuals. These vulnerabilities that make people more or less
sensitive to negative contextual shifts could be either hard-wired or evolved from particular
learning histories. Thus, the behavior analytic framework attempts to analyze episodes of
depression, including proximal causes, while remaining silent about more distal causes
commonly thought of as risk factors. The model also accounts for the perpetuation of
depressive behavior through avoidance of anti-depressant reinforcers: because contact with
these potential reinforcers is often punishing in the short run, depressed people avoid contact,
to their short-term relief but their long-term detriment. For example, a depressed man in our
current study discovered that his wife had been unfaithful: this discovery triggered depressive
behavior, including avoidance of interpersonal contact, contact which was ultimately necessary
to provide access to anti-depressant reinforcers. The BA therapist must unravel this secondary
layer of avoidance, prior to the frontal assault designed to undo the depressogenic contextual
shifts. Finally, the BA therapist must teach the model to clients, so that they become behavior
analysts and use these new skills to prevent relapse.
Our functional analytic version of BA is part of a larger eort to provide an alternative to
the defect models of depression that have dominated the discourse among clinical scientists:
most notably, defects in brain chemistry; but also defects in thinking, represented by CT. These
defect models represent the triumph of the disease model among the mental health professions,
a model that was once challenged by clinical psychologists, but is now embraced by the bulk of
the scientic community, not just for bipolar depression but also for the heterogeneous
category known as major depressive disorder (MDD). The disease model of depression has
been incorporated into the thinking of even those who advocate psychosocial treatments and
can be considered the normal science of our time. Included in this view are a number of
assumptions.
First, there is the assumption that depression is qualitatively dierent from the dysphoria
experienced by those who seek therapy complaining of the blues but are either sub-threshold
on the DSM or not severe enough on measures of symptoms. Second, this qualitatively MDD
is assumed to be episodic, with a course that generally lasts from about 912 months before it
remits even in the absence of treatment. There is assumed to be a period following an episode
where the risk for relapse is elevated, making the distinction between relapse-return of
suppressed symptoms from a previous episode-and recurrence-onset of a new episode-
114 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117

meaningful. Third, the distal, primary, or ultimate cause of MDD is assumed to be within the
person-a predisposition-either genetic or learned.
The once popular but now largely discredited counterpart to these assumptions can be dimly
detected in the contextualism of the 1970's (Rehm, 1981): our contextual revisionism views
MDD as a quantitatively but not qualitatively distinct manifestation of mood disturbances
which are best described in a continuity model which uses dimensions rather than categories;
we assume (and have evidence (cf. Gortner et al., 1998)) that the probability of relapse remains
constant over time, thus discrediting the notion that depression in outpatients (nonpsychotic,
unipolar) is episodic and runs its systemic course-independent of context and, nally, that the
distal causes of depression include not just genetic vulnerability but also learning history. Our
study is an initial step toward building this model as well as a clinical trial comparing BA with
CT and pharmacotherapy.
We are one and a half years into our new trial. The subject pool consists of HMO
outpatients and we have controlled for the allegiance eect (Jacobson, in press) by adding to
our team of project coordinators (Sona Dimidjian, Jackie Gollan and Lisa Roberts) and co-
Investigator Keith Dobson a team of experts with allegiance to CT (Steve Hollon, Sandra
Coman); BA (Michael Addis) and pharmacotherapy (Dave Dunner; Jan Fawcett; Don Klein).
Each approach is supervised by experienced practitioners and advocates of their particular
model. Competence ratings from recognized experts outside the study will be obtained. We will
ultimately enter 500 subjects into this trial, making it the largest outpatient clinical trial ever
conducted in the area of depression.
It is far too early to know what we will nd. However, there are some interesting facts
emerging which are-thus far-consistent with some of the discoveries from our previous trial.
First, our BA therapists began the trial with no training and no pilot cases. Given the extensive
costs of training CT therapists, at least as some advocates dene the prerequisites to
competence, equivalent ndings would go a long way toward favoring BA as the new
psychosocial treatment of choice for major depression. Second, we have closely examined and
tracked the course of people seeking treatment for depression in this study, as we did in the
last study. Our inclusion and exclusion criteria are standard and widely used in the eld.
However, our help seekers are not under experimental control; they are simply the typical
people who seek treatment for depression in Seattle. Thus far, patients seeking treatment for
depression on an outpatient basis have about a 4% chance of being accepted into our study. A
small number refuse random assignment. Occasionally, people are rejected because they exhibit
bipolar, psychotic or excessive suicidal symptoms. However, the majority of those rejected are
depressed, just not depressed enough to make it into the study. Either they endorse three
rather than four symptoms in addition to the pervasive sadness that leads them to call
themselves depressed, thus not meeting criteria for MDD; their BDI scores are 18 instead of 20
or greater or their HRSD scores are 13 rather than 14 or greater. All of these people, often
referred to as sub-threshold, are depressed, and are seeking therapy for it. They are seen every
day in the oces of our colleagues, in greater numbers than those we accept into our clinical
trials. It is hard to make the case that they constitute a qualitatively distinct group from those
who are both diagnosed MDD and have high enough scores on our measures of severity to be
included into the study. In an outpatient population, our clinical trials cover only a small
minority of the local help-seekers for depression. Not only does this possibility raise serious
 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117 115

questions about the generalizability of our ndings to outpatient clinical practice, but they help
account for the cynicism often expressed by clinicians about the millions of dollars spent on
RCT's. As one articulate colleague put it, ``Where are these people? I don't know where you
guys nd them. I have been treating depression for 40 years and I never see them''. It is easy
to forget about that 4%.
Third, we hypothesize, based on our previous study, that-unlike the ndings reported from
the largely inpatient studies in the early years of depression research (Hollon & De Rubeis,
1981), MDD is typically neither episodic nor systemic in course, that the probability of relapse
does not decrease as we become temporally removed from the time of remission. All in all, we
expect to nd support for a continuity model of MDD: results that are consistent with the
notion that MDD is an arbitrary extreme, the area at the peak of the dimension, with cuto
points that are equally arbitrary. Further, we expect our results to be consistent with a model
of external causation. By the end of our study, if we are correct, there will be reason for some
of us to view depression as a quite understandable response to life events, not an irrational
response or one that is primarily a function of alterations in brain chemistry. Although we do
not believe that all are equally vulnerable to depression, we do hypothesize that the role of
social context was prematurely dismissed by advocates of defect models.
Thus, although we are pleased with our previous trial, we are trying to be responsive to the
critics in the current study. We have even included three pharmacotherapy conditions, all of
which are control groups. First, there is a pill placebo group, which is there for no other
reason than to demonstrate a drug eect during acute treatment (statistically signicant
dierences between drugs and placebo) to satisfy those critics who view all comparisons
between psychotherapy and pharmacotherapy as meaningless unless they also establish that the
sample is drug-responsive. Despite the circularity of the argument put forth by exponents of
this point of view (McNally, 1996; Klein, 1996), we want our results to be above reproach,
even when we disagree with the requirements of our colleagues (Jacobson & Hollon, 1996).
Thus, just as we have used competence ratings and a Dream Team of cognitive therapists to
satisfy our critics from the CT camp, so have we included pharmacotherapy, in one of two
phases ((a) 16 weeks of acute treatment to remission, followed by transfer to placebo under
double blind conditions and (b) continuation of treatment for 12 months following remission),
to establish stringent control groups for establishing the relapse prevention eects of our
psychosocial treatments. We expect that both BA and CT will demonstrate relapse prevention
eects, eects that we also expect to nd with pharmacotherapy as long as clients continue to
receive medication. However, in the end, when the cost of CT (e.g. training and intensive
supervision) and pharmacotherapy (nancial, side eects, etc.) are factored into the equation,
we expect BA to emerge as the most cost-eective long-term treatment for MDD. Not all of
our co-Investigators hypothesize as we do. However, the control for allegiance eects within
one sight makes our study unusual. We may obtain some interesting answers to long-standing
questions.

6. Conclusion

We began with a description of a RCT which produced null ndings. The null ndings,
116 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117

although open to multiple interpretations, pointed to the possibility that cognitive interventions
are neither necessary nor desirable in an optimal psychosocial treatment for depression. Our
study stirred up controversy, enough to inspire this paper, despite the absence of written
critiques. We ended with the description of a new study, which is in part an attempt to
replicate our ndings to the satisfaction of our critics and has become something more: an
eort aimed at reconsidering outpatient MDD from a functional analytic framework. What we
have presented is the evolution of a research program that is still evolving. And it all started
with a study of BCT.
We oer this narrative as a snapshot of how our research program on depression has
progressed over the last decade. Although we strive to follow our data, one of the ongoing
lessons is that the search for truth is often complicated by quasi-scientic and even political
matters. Some of these non-scientic matters, when added to the mix, can result in better
science. We hope that our current study provides such an example. In any case, we're not
complaining. These are interesting times to be doing depression research and controversy is not
necessarily a bad thing.

Acknowledgements

Preparation of this article was supported by National Institute of Mental Health Grants
2R01 MH44063-06 and 5K02 MH00868-05.

References

Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy of depression. New York: Guilford
Press.
Christensen, A., & Jacobson, N. S. (in press). Reconcilable dierences. New York: Guilford.
Christensen, A., & Jacobson, N. S. (1993). Who or what can do psychotherapy: the status and challenge of nonpro-
fessional therapies. Psychological Science, 5, 814.
Dobson, K. S. (1989). A meta-analysis of the ecacy of cognitive therapy for depression. Journal of Consulting and
Clinical Psychology, 57, 414419.
Elkin, I., Shea, T., Watkins, J. T., Imber, S. C., Sotsky, S. M., Collins, J. F., Glass, D. R., Pilkonis, P. A., Leber,
W. R., Fiester, S. J., Docherty, J., & Parlo, M. B. (1989). NIMH Treatment of Depression Collaborative
Research Program. Archives of General Psychiatry, 46, 971982.
Frank, E., Prien, R. F., Jarrett, R. B., Keller, M. B., Kupfer, D. J., Lavori, P. W., Rush, A. J., & Weissman, M. M.
(1991). Conceptualization and reationale for consensus denitions of terms in major depressive disorder: remis-
sion, recovery, relapse, and recurrence. Archives of General Psychiatry, 48, 851855.
Gortner, E. T., Gollan, J. K., Dobson, K. S., & Jacobson, N. S. (1998). Cognitive-behavioral treatment for de-
pression: relapse prevention. Journal of Consulting and Clinical Psychology, 66, 377384.
Hahlweg, K., & Markman, H. (1988). Eectiveness of behavioral marital therapy: empirical status of behavioral
techniques in preventing and alleviating marital distress. Journal of Consulting and Clinical Psychology, 56, 440
447.
Hollon, S. D., & De Rubeis, R. J. (1981). Placebo-psychotherapy combinations: inappropriate representations of
psychotherapy in drug-psychotherapy comparative trials. Psychological Bulletin, 90(3), 467477.
Jacobson, N. S. (in press). The role or the allegiance eect in psychotherapy research: controlling and accounting
for it. Clinical Psychology: Research and Practice.
 N.S. Jacobson, E.T. Gortner / Behaviour Research and Therapy 38 (2000) 103117 117

Jacobson, N. S., & Addis, M. E. (1993). Research on couples and couple therapy: what do we know? Where are we
going? Journal of Consulting and Clinical Psychology, 61, 8593.
Jacobson, N. S., & Christensen, A. (1996). Integrative couple therapy: promoting acceptance and change. New York:
W.W. Norton.
Jacobson, N. S., Dobson, K. S., Truax, P. A., Addis, M. E., Koerner, K., Gollan, J. K., Gortner, E. T., & Prince,
S. E. (1996). A component analysis of cognitive-behavioral treament for depression. Journal of Consulting and
Clinical Psychology, 64, 295304.
Jacobson, N. S., Dobson, K., Fruzzetti, A. E., Schmaling, K. B., & Salusky, S. (1991). Marital therapy as a treat-
ment for depression. Journal of Consulting and Clinical Psychlolgy, 59, 547557.
Jacobson, N. S., Follette, W. C., Revenstorf, D., Baucom, D. H., Hahlweg, K., & Margolin, G. (1984). Variability
in outcome and clinical signicance of behavioral marital therapy: a reanalysis of outcome data. Journal of
Consulting and Clinical Psychology, 52, 497504.
Jacobson, N. S., Fruzzetti, A. E., Dobson, K., Whisman, M., & Hops, H. (1993). Couple therapy as a treatment for
depression: II. The eects of relationship quality and therapy on depressive relapse. Journal of Consulting and
Clinical Psychology, 61, 516519.
Jacobson, N. S., & Hollon, S. D. (1996). Cognitive-behavior therapy versus pharmacotherapy: now that the jury's
returned its verdict, it's time to present the rest of the evidence. Journal of Consulting and Clinical Psychology, 64,
7480.
Jacobson, N. S., & Margolin, G. (1979). Marital therapy: strategies based on social learing and behavior exchange
principles. New York: Brunner/Mazel.
Keller, M. B., Lavori, P. W., Friedman, B., Nielsen, E. et al. (1987). The longitudinal internal follow-up evaluation:
a comprehensive method for assessing outcome in prospective longitudinal designs. Archives of General
Psychiatry, 44(6), 540548.
Klein, D. F. (1996). Preventing hung juries about therapy studies. Journal of Consulting and Clinical Psychlolgy, 64,
8187.
McNally, R. J. (1996). Methodological controversies in the treatment of panic disorder. Journal of Consulting and
Clinical Psychology, 64, 8891.
Rehm, L. P. (1981). Behavior therapy for depression: present status and future directions. New York: Academic Press.
Vallis, T. M., Shaw, B. F., & Dobson, K. S. (1986). The cognitive therapy scale: psychometric properties. Journal of
Consulting and Clinical Psychology, 54(3), 381385.
Waltz, J., Addis, M. E., Koerner, K., & Jacobson, N. S. (1993). Testing the integrity of a psychotherapy protocol:
assessment of adherence and competence. Journal of Consulting and Clinical Psychology, 61(4), 620630.
Wolpe, J. (1958). Psychotherapy by reciprocal inhibition. Stanford, CA: Stanford University Press.