Addiction (1998) 93(2), 163± 172

LEC T U RE

W hy do people change addictive behavior?

The 1996 H . David Archibald Lecture

W ILLIAM R. M ILLER

University of New M exico, Albuquerque, N ew M exico, USA

Abstract
Although there are specialty journals and professions focusin g on addiction s treatm ent, the reasons why people
change addictive behaviors are still not well understood. The sim plistic accoun t that people change because
they receive treatm ent is wanting in m any ways. M any people who recover do so without form al treatment.
Even relatively brief intervention s seem to trigger change, and the dose of treatm ent delivered is surprisingly
unrelated to outco m es. Client com pliance with m any different approaches, including placebo m edication, has
been linked to better outco m es. C lient outcom es also vary widely dependin g up on the therapist who delivers
treatm ent. Various models are brie¯ y considered that may help in understandin g this intrigu ing puzzle of
change.

Introduction shall we com prehend the transform ations to ab-

It is no wonder that people develop addictions. stinence or moderation that occur in a majority
Excesses with food and alcohol are as old as of individuals who were once entangled with
recorded history. The truth is that addictive be- alcohol or other drugs? What principles of physi-
haviors feel good, and we are wired to repeat ology and learning could account for the about-
such responses. With greater experience and face changes that one sees in people with or
technology, humankind has managed to re® ne without treatment, and in Alcoholics Anony-
drugs so exquisitely suited to reinforce the brain m ous?
that laboratory animals will self-adm inister them One answer to this question is beam ed at us
until they drop from exhaustion or starvation. through books, m agazines, television and pro-
One needs little more than physiology and learn- fessional training program s. The answer is
ing principles for a feasible account of how ad- ª treatmentº . Those who get treatment get bet-
dictions are established. If anything, the mystery ter. Indeed, there is evidence to support this
is in why we have not all succum bed. view. M any studies show outcome differences
What one does not ® nd is an adequate anim al between those who receive or do not receive a
model for human recov ery (Logan, 1993). W hy particular treatment (M iller et al., 1995). After a
do people with fully established addictions turn treatment event, health care costs tend to go
away from them, sometim es quite abruptly? How down dramatically (Holder & Blose, 1992). The

Correspondence: W illiam R. Miller PhD, U niversity of New M exico, Departm ent of Psychology, Logan H all,
Terrace & Redondo, Albuquerque, New M exico 87131-11 61, USA.
Subm itted 12th November 1996 ; initial review completed 3rd February 1997; ® nal version accepted 11th June
1997.

0965± 2140/98/020163 ± 10 $9.50 Ó Society for the Study of Addiction to Alcohol and Other Drugs

Carfax Publishing Lim ited

164 W illiam R. M iller

longer a person stays in treatment, the better he twice in New M exico, comparing 6 weeks
or she fares (e.g. Robson, Paulus & Clarke, (Miller & Taylor, 1980) and then 18 weeks
1965; cf. Finney, M oos & Chan, 1981; Pettinati (Miller, Taylor & W est, 1980) of outpatient
et al. , 1996). It is a simple and satisfying answer: treatment with the brief ª bibliotherapyº control
people defy the laws of addiction because of this condition. In both studies, clients random ly as-
entity called ª treatmentº . signed to the control condition again fared as
well as those treated by therapists.
Perhaps, I reasoned, it is all an illusion. Per-
Th e pieces of a puzzle haps clients come into these studies already
Yet something seems to be amiss with this way of prim ed for change, and it matters not at all what
thinking. The longer I have studied treatment one does. To check this possibility, Kay Harris
and change, the more I have come to question and I added two more control groups in which
this view in which I was trained. Seldom have I clients were assessed but then placed on a wait-
found what I expected in three decades of re- ing list for 10 weeks without any self-help in-
search. The data have challenged, disturbed and structions. One of these groups self-monitored
puzzled m e. their drinking on daily diary forms that were
m ailed back to us (which had been part of our
bibliotherapy condition), and the other did not.
Active controls W e did this because clients in our bibliotherapy
It started out simply enough. I wanted to com- condition had been telling us it was self-monitor-
pare the ef® cacy of different m ethods for treating ing that had the biggest impact on them . Clients
problem drinkers. In m y ® rst try at this (M iller, in the two waiting list control groups showed no
1978), I found no real differences among three reduction in their drinking, but once more both
rather different treatment methods, but there the bibliotherapy and the therapist-treated
was one tantalizing, post-hoc ® nding. I had pre- groups evidenced substantial and similar im-
pared, with the help of Ricardo MunÄ oz, a self- provement (Harris & Miller, 1990). The ª active
help guide to assist clients after treatm ent in ingredient,º it appeared, lay somewhere between
dealing not only with drinking, but with a the assessment and brief bibliotherapy compo-
broader range of the life problem s facing them nents.
(later published as Miller & M unÄ oz, 1982). I had
planned to give this to each client at the con-
clusion of therapy, but then decided I ought to Treatm ent intensity research
ensure that it would not in any way alter follow- Another piece of the puzzle em erged as we began
up ® ndings. Thus we randomly assigned clients, to review research on the effects of treatment
after 10 weeks of treatment, to receive or not intensity. Reid Hester and I collected and re-
receive this guidebook. Those who received the viewed all the studies we could ® nd in which
guidebook showed, as a group, continuing re- people had been assigned at random (or
duction in drinking, whereas those not receiving m atched to create equivalent groups) to treat-
it simply plateaued at their post-treatm ent level, m ents varying in apparent intensity: inpatient
so that 3 months later the groups were versus outpatient or day treatment, or longer
signi® cantly different. Then we gave the guide- versus shorter treatment. W ith surprising con-
book to the rest of our clients, and by 6 m onths sistency, these studies revealed no overall differ-
the difference had disappeared. ences in outcomes as a function of greater versus
What would happen if we provided the guide- lesser intensity of treatment (M iller & Hester,
book at the beginning , instead of delivering treat- 1986).
ment? In our next study, problem s drinkers were Relatedly, we began to assemble the literature
assessed and then assigned either to 10 sessions on relatively brief interventions for alcohol prob-
of outpatient treatment, or to a control condition lems. The ® ndings here were also rather consist-
in which they were given some encouragement ent across cultures. As Kay Harris and I had
and sent hom e with the self-help materials. To found, outcom es from brief intervention tend to
our surprise both groups fared well, and there be better than those with no intervention, but
was no real difference in outcom es (M iller, Grib- when brief intervention of 1± 3 sessions is com -
skov & M ortell, 1981). The study was repeated pared with more intensive treatment, often there
 W hy do people change addictive behavior? 165

are no short- or long-term differences in out- de® ned by Truax & Carkhuff (1967). When
come (Bien, Miller & Tonigan, 1993b). client outcomes became known, we compared
therapists’ empathy ratings and their success
rates, which varied from 25% to 100% at 6
Com pliance and placebo effects m onths. W e were able to predict, from therapist
These ® ndings from controlled trials clash empathy rankings, two-thirds of the variance
markedly with the results of uncontrolled studies ( r 5 0.82) in client drinking (volum e of con-
of treatm ent intensity. It is a relatively consistent sumption per week) at 6 months, half (r 5 0.71)
® nding that clients who stay longer in or com ply at 12 m onths, and still one-fourth (r 5 0.51) at
more fully with treatment also tend to have 24 m onths (M iller & Baca, 1983).
better outcom es. This is som etimes interpreted A decade later we analyzed in greater
as evidence for the speci® c ef® cacy of treatment, detail this link between therapist style and
in that larger doses seem linked to better results. client drinking outcom es (Miller, Bene® eld
How can one reconcile the testimony of con- & Tonigan, 1993). First, we were able to
trolled and uncontrolled studies? An obvious replicate the ® nding of Patterson and Forgatch
confounding elem ent in the uncontrolled studies (1985) that client resistance or ª denialº
is that more com pliant clients are self-selected. can be experimentally in¯ uenced by therapist
Perhaps clients who work harder are initially behavior. High levels of resistance were
ª more motivatedº and for this reason fare better. associated with a more confrontational
In fact, better outcom es may be associated with therapist style, and low resistance with a more
doing m ore of alm ost anything to get better. The empathic counseling style. Therapist behavior
taking of some action toward recovery is a gen- counts from a recorded single counseling session
eral predictor of change (e.g. M iller et al. , 1996). were used to predict client outcom es at 12-
Clients who become more involved in Alcoholics m onth follow-up. From a single therapist re-
Anonymous after treatm ent are more likely to sponseÐ confrontingÐ we were able to predict
stay sober, although attendance alone m ay be an accurately ( r 5 0.59) the number of standard
insensitive predictor (e.g. M ontgomery, Tonigan drinks that the client would be consuming 1 year
& M iller, 1995). Clients who faithfully take their later: the m ore the therapist confronted, the
medication are more likely to show bene® t, even m ore the client drank.
if the medication is a placebo (Fuller et al. , These are not isolated ® ndings. There can be
1986). There seem s to be something about doing large variability in client outcomes across coun-
something to get better. selors. Therapists’ personality attributes, super-
Placebo effects are interesting in their own visor ratings, and even voice tone have been
right. In one of the largest studies of alcoholism found to predict how their clients will fare (for
treatment ever conducted, Fuller and his col- reviews see M iller, 1985; M iller & Rollnick,
leagues (1986) found that clients in both a 1991; Najavits & Weiss, 1994). Therapists’ posi-
disul® ram group and a placebo group showed tive expectancies for client outcomes, when m a-
signi® cantly better outcomes than did clients nipulated experim entally, have been shown to
given no pills. The disul® ram and placebo become self-ful® lling prophecies (Leake & King,
groups differed little from each other, but a 1977). Therapist effects are not, of course, all
therapeutic bene® t was evident in both. What positive. The therapist to whom a drug addict
they had in common was the non-speci® c effect (McLellan et al. , 1988) or an alcoholic (Valle,
of being given a medication to get better. 1981) is assigned at random can be the differ-
ence between high versus low risk of relapse.
C lients for three of the nine therapists in our
Sensitivity to therapist effects earlier study showed outcomes less favorable
My ® rst hint of therapist effects in addiction than those of clients in the self-help control
treatment came early, as an incidental ® nding of condition (M iller et al. , 1980). It is striking, too,
a study described above (Miller et al. , 1980). As that clients placed on a ª waiting listº did just
one part of the supervision of therapists deliver- thatÐ they waited without changing, whereas
ing the behavioral treatments that we were test- those told to begin changing on their own did so
ing, three supervisors rank-ordered the nine (Harris & M iller, 1990). Implicit in the com -
counselors on Rogerian em pathy as operationally m unication that one is on a ª waiting list for
166 W illiam R. M iller

treatmentº may be the instruction that change would intentionally manifest these com mon ele-
will not occur until professional help is delivered. m ents. Those receiving the DCU ® rst com plete
a reasonably thorough assessment of their drink-
ing and its consequences, then return for a se-
Natural change cond session to receive personal feedback
However, change can and does occur outside the delivered in an empathic (motivational inter-
context of treatm ent, and herein is yet another viewing) m anner. Embedded messages em pha-
piece of the puzzle. The stages and processes of size explicitly the client’ s freedom of choice,
change for addictive behaviors appear to be simi- personal responsibility for and ability to change.
lar whether change occurs with or without pro- Advice and additional information are provided
fessional assistance (Prochaska, DiClem ente & as requested by the client.
Norcross, 1992). Longitudinal studies of prob- Our ® rst two random ized trials of a two-
lem drinkers identi® ed in the general population session DCU were with self-referred problem
indicate that a majority improve within a few drinkers, most of whom had no history of formal
years, most of whom never receive professional treatment. Both showed relatively rapid and en-
help for their drinking (Sobell, Sobell & during decreases in drinking behavior, occurring
Toneatto, 1991). M ost people with alcohol within 6 weeks of the DCU, and maintained
problems get over them, and by any stretch of through 12 m onths of follow-up (Miller, Sover-
the imagination only a small minority are ever eign & Krege, 1988; M iller, Bene® eld & Toni-
reached by form al treatm ent systems. In a long- gan, 1993). Two other trials randomly assigned
term developm ental perspective, the course of clients to receive or not receive a DCU upon
alcoholism appears relatively unrelated to treat- entering a private residential substance abuse
ment events (Vaillant, 1983). Abrupt and sweep- program (Brown & M iller, 1993) or an outpa-
ing transformations that affect broad areas of tient treatment program for veterans with alcohol
behavior and personality similarly occur nat- problem s (Bien, M iller & Boroughs, 1993a). In
urally, outside the context of any formal helping both studies, clients who received the DC U at
relationship (M iller & C’ de Baca, 1994). the onset of treatment showed substantially
higher rates of abstinence and improvem ent at
3-month follow-up. Said another way, the sam e
M otivational intervention treatment programs had very different outcomes
If a client is ª not ready,º can one intervene to with versus without brief motivational interven-
enhance motivation for change? As it became tion at the outset. Bene® cial effects of motiva-
clearer that enduring behavior changes do occur tional interviewing have also been demonstrated
in the context of relatively brief interventions, we in randomized trials with pregnant drinkers
sought to identify comm on components of such (Handm aker, 1993) and heroin addicts (Saun-
interventions (Bien et al. , 1993b; M iller & ders, Wilkinson & Phillips, 1995).
Sanchez, 1994). A result was the mnemonic Other studies have contrasted motivational in-
acronym FRAMES, highlighting six elements terviewing with different treatment strategies.
frequently present in effective brief interventions: Heather et al. (1996) compared motivational in-
Feedba ck of personal status, emphasis on per- terviewing with behavioral coping skills training
sonal Responsibility for change, Adv ice to change, in the treatm ent of problem drinkers identi® ed in
offering a M enu of approaches by which change general health care settings. For clients who were
might be achieved, an E m pathic counseling style, initially unmotivated for change m otivational in-
and messages supporting client Self-ef® cacy for terviewing yielded greater behavior change,
change. The assessment and follow-up contacts whereas among those ready for change there was
present in brief intervention studies may also be no difference in the ef® cacy of the two ap-
important in¯ uences for change, providing at proaches. An adapted four-session version of
least implicit feedback, although in one study the Drinker’ s Check-up, Motivational Enhance-
assessm ents alone did not suf® ce, at least in the m ent Therapy (M ET), was tested as one of
context of waiting list instructions (Harris & three treatments included in Project M ATCH,
Miller, 1990). (1993, 1997). MET was compared in this ran-
The Drinker’ s Check-up (DCU) was an at- domized trial with two 12-session outpatient
tempt to build a motivational intervention that strategies: cognitive± behavioral skill training and
 W hy do people change addictive behavior? 167

twelve-step facilitation therapy. During the 3- (see Davidson, 1992; and Farkas et al. , 1996,
month period of treatment, all three treatment with accom panying com mentaries). This
groups showed dramatic improvem ents, al- in¯ uential model is predominantly descriptive
though M ET lagged slightly behind. Through- and explicitly transtheoretical. Its stages and pro-
out the year following treatment, however, cesses were derived empirically by factor and
comparable changes were observed in MET rela- cluster analytic m ethods. In this sense, the model
tive to the other two groups on the primary is more directed toward how than why people
outcome variables. change, yet the extensive em pirical research un-
derlying the model may offer a framework within
which to explore theoretical constructs underly-
M etapho ric m odels ing change. The model posits a sequence of
In sum , the explanation that ª people change stages involving various kinds of affective, behav-
because they get treatmentº is wanting in several ioral and cognitive appraisal shifts. A number of
ways (Humphreys, M oos & Cohen, 1997). comm on processes of change have also been
Many who change addictive behaviors do so identi® ed, and tied to earlier and later stages of
without any formal treatment. Although the change. Concepts of theoretical importance,
seeking of help may be a signal event motivation- such as self-ef® cacy and the relative balance of
ally, the dose of treatment received, when stud- pros and cons for change, have been studied in
ied experimentally, has shown little relationship relation to progression through the stages.
to outcomes. People get better with placebo W ithin this view, change is promoted by helping
medications, and in minimal help conditions that the individual to advance from one stage to the
were regarded to be control groups. The style of next. A key for the therapist, then, is in sensing
interaction between counselor and client appears the emotional, behavioral or cognitive shifts that
to have a substantial ability to accelerate or need to occur for the client to progress toward
impede the natural change process. Relatively change. The optimal procedures for triggering
brief motivational interventions exert a sizeable these shifts are only beginning to be discovered,
effect on outcomes with or without additional and it may be that relatively brief interventions
treatment. Furthermore, ª treatmentº is anything can suf® ce to trigger at least some of these
but a homogeneous phenomenon (Institute of transitions. It is possible that a triggered tran-
Medicine, 1990), and attributes of treatment sition between certain stages is a suf® cient cata-
represent only one class of determ inants of client lyst for an ongoing process of change, which
outcome (Moos, Finney & Cronkite, 1990). could account for some of the sharp de¯ ections
Treatment appears to be neither a necessary nor in behavior observed with brief interventions.
a suf® cient condition for change.
Why then do people change addictive behav- Self-ef® cacy
iors? What is it about the human cortex that Bandura’ s (1977) construct of self-ef® cacy has
allows us to escape the ensnaring physiology of been similarly in¯ uential in the ® eld of addictive
addiction? I will consider brie¯ y here ® ve broad behaviors. The individual’ s belief in his or her
and interlocking metaphoric m odels that seem ability to change has ® gured prominently in
promising lenses through which to focus on the health belief models such as the protection mo-
uniquely human process of recovery. tivation theory of Ronald Rogers (Rogers &
Prentice-Dunn, 1997), and has been central in
various treatm ent approaches (e.g. M arlatt &
Stages of change Gordon, 1985; Annis, 1986). The individual’ s
The transtheoretical model of Prochaska and his cognitive appraisal of the ability to change is set
colleagues, and its attendant stages and pro- forth as a necessary condition for change to
cesses of change, are by now quite familiar to occur. Self-ef® cacy has been regarded as a key
most who work in the ® eld of addictions. This m oderating variable in understanding the incon-
model arose from research on addictive behav- sistent effects of feedback (Kluger & DeNisi,
iors, and has been used widely as a heuristic for 1996) and fear arousal (Leventhal, 1971) on
understanding change (M iller & Heather, 1986; behavior. Within this view, behavior change is
Prochaska et al. , 1992). Both proponents and promoted by the enhancem ent of self-ef® cacy, as
detractors have offered extensive commentary by training in effective coping skills.
168 W illiam R. M iller

It is a testable hypothesis whether an observed Smith, 1995), the mechanisms underlying the
shift in self-ef® cacy is a necessary or suf® cient ef® cacy of brief intervention logically lie within
antecedent of behavior change. It is a com mon the individual’ s intrinsic motivational structure.
® nding that cognitive appraisal shifts can shadow In self-regulation theory (Kanfer, 1986; M iller &
behavior change rather than preceding it, and Brown, 1991), behavior change is triggered by a
other ® ndings from our laboratory suggest that detected discrepancy between feedback of cur-
cognitive markers of outcom e could be passive rent status and a goal or standard with which it
re¯ ections of behavior change (e.g. M iller et al. , is compared. M iller & Rollnick (1991) described
1996). Nevertheless, shifts in self-ef® cacy seem a a therapeutic process of developing discrepancy ,
promising place to look in trying to understand enhancing salient awareness of the cons of cur-
what triggers change. rent behavior and the pros of change. A dis-
crepancy between present status and important
goals has long been described as a necessary but
Am bivalence not suf® cient condition for change, and progress
Miller & Rollnick (1991) chose ambivalence as a has been made toward identifying factors that
central construct in their description of the m oderate the m otivational impact of discrepan-
mechanisms of motivational interviewing. In the cies (Kluger & DeN isi, 1996). Further, goals are
language of decisional balance (Janis & M ann, generally understood as being hierarchically or-
1977) ambivalence may be thought of as a coun- ganized, and the likelihood and intensity of
terbalancing set of pros and cons that immobilize change may be related to the level of goals with
the change process. Often the m otivational vec- which a current behavior con¯ icts (Rokeach,
tors related to addictive behavior are arrayed 1973).
along not one dimension, but m any dimensions. If the tenets of self-regulation theory are
The pros and cons of drug using and of change sound, then the instigation of a perceived dis-
are likely linked to a variety of personal goals. crepancy between current behavior and
From an operant perspective, ambivalence can signi® cant goals would be expected to set in
be understood as simultaneous competing con- m otion a chain of corrective action. Physical
tingencies, and recovery as the transferring of homeostatic system s such as thermostats show
control for a behavior from one contingency (or this property of prolonged periods without
set of contingencies) to another. M ore generally, change, followed by an abrupt shift in state
resolution of ambivalence involves a tipping of triggered by a detected discrepancy. Herein is
the m otivational balance, removing barriers and another plausible account of the genesis of be-
disincentives (cons) for change and enhancing havior displacem ents that from other perspec-
the number or salience of pros for change. tives might seem surprisingly sudden or large.
Bivalent con¯ icts (such as single or double
approach-avoidance) are notoriously immobiliz-
ing. The processes by which such con¯ icts are M eaning
resolved in humans are com plex indeed. Thera- The making of meaning is a characteristic of
peutic procedures that lead to a sudden resol- higher cognitive functions. Discrete disjunctive
ution of such ambivalence (becoming unstuck) changes in organized behavior, such as are some-
might be expected to liberate behavior to change, times seen in recovery from addictions, m ay be
even dramatically. Related shifts in controlling related to an abrupt shift in the perceived m ean-
contingencies might account for both acute ing of the addictive behavior itself. In the lan-
con¯ ict resolution and abrupt behavior change. guage of contemporary behaviorism, this could
be thought of as a shift of the behavior from one
stimulus equivalence class to another (Dougher
Discrepancy & M arkham, 1996). An illustrative story of this
Whatever is happening in motivational interview- type was told by David Premack (1970, p. 115)
ing, it does not appear to be a change in actua l of a m an who:
contingencies affecting the target behavior. Al-
though som e treatm ent approaches do seek to dates his quitting from a day on which he had
alter directly the environmental contingencies gone to pick up his children at the city library.
controlling addictive behavior (e.g. M eyers & A thunderstorm greeted him as he arrived
 W hy do people change addictive behavior? 169

there; and at the same tim e a search of his needs to progress toward such integration, a
pockets disclosed a familiar problem: he was work in early progress and in need of others’
out of cigarettes. Glancing back at the library, insights.
he caught a glimpse of his children stepping First, an understanding of recovery and of why
out in the rain, but he continued around the people change addictive behaviors m ust, l think,
corner, certain that he could ® nd a parking be focused on higher cognitive processes, the
space, rush in, buy the cigarettes and be back very elem ents that are lacking in animal models
before the children got seriously wet. The view (Logan, 1993). There has been nothing in effec-
of himself as a father who would ª actually tive brief interventions to shift actual environ-
leave the kids in the rain while he ran after m ental contingencies, to teach coping skills, to
cigarettesº wasÐ I think the word is appropri- establish reconditioning, or to alter physiology
ateÐ humiliating, and he quit smoking. and pharm acology. Rather, the changes emerg-
ing in such a short time are likely to be funda-
More recent developments in self theory (e.g. m entally motivational in nature. What we
Hermans, Kempen & van Loon, 1996) provide a observe clinically has the quality of a shift in
fram ework within which to understand the shift- complex appraisal, which is often followed by
ing meaning of a particular behavior when it enduring behavior change.
collides with higher-level values and conceptions Secondly, I believe that the hierarchical orga-
of the self. nization of goals and values is a particularly
The behavioral implications of a deep shift in fruitful m etaphor for understanding such ap-
meaning are poorly understood, but enough is praisal shifts (cf. Cox, Klinger & Blount, 1991).
known to suggest that related changes in affect, M otivational interviewing is a narrative process
cognition and behavior can be large and broad- of evoking from the client reasons for and com -
ranging (Rokeach, 1973; M iller & C’ de Baca, m itm ent to change. Study of the language by
1994). The high-pro® le examples of cults and which commitments are requested and m ade is
brainwashing are only a small, albeit dram atic itself a well-developed area of research, and I am
piece of the picture here. The reorganization of currently collaborating with psycholinguist Paul
cognitive± perceptual structure can range from a Am rhein (1992) to determine whether speech
simple peripheral level (reframing) to the kind of events within motivational interviewing have a
core shift in identity imm ortalized by Charles reliable sequence and systematically predict be-
Dickens in A Christm as Carol . It is reasonable havior change. The self can itself be understood
that the m ore central the restructuring, the as a dialogical and storytelling process, an idea
greater the sweep of change. It is sensible, then, dating at least to William James (Hermans,
to study abrupt changes in alcohol or drug use 1992).
within the larger context of the individual. Just as The progress of understanding in science of-
problem behavior theory posits alcohol/drug ten, although not always, proceeds from descrip-
problems to be just one part of a larger cluster of tion to prediction to control. We have intriguing
life problems (Jessor, 1987), so remission in one change phenomena to explain. We already have
problem behavior may be just one element of a m ultiple descriptive models for what motivates
more comprehensive pattern of life change. change. We have clear demonstrations of exper-
Within Alcoholics Anonymous, sobriety is m uch imental control, but not yet a clear understand-
more than being dry. The larger pattern is ing of the processes through which such control
missed if we look only through the microscope of is exerted. The intervening links at present are
quantifying alcohol/drug use. m etaphoric and clinical/intuitive (M iller & Roll-
nick, 1991). A step forward would be the devel-
opment and testing of predictive models that
integrate or compare the various explanatory
M ixing m etaphors m etaphors. The needed theoretical pieces may
It would be satisfying to present at this point already be available and simply in need of as-
a fully-developed theoretical model that inte- sembly, or it may be necessary to form ulate a
grates the above metaphors and accounts for new conceptual framework to understand these
most of the observed data. Instead I have only a phenom ena. As consistent predictive relation-
vision of a direction in which I believe the ® eld ships are clari® ed, the re® nem ent and testing of
170 W illiam R. M iller

therapeutic interventions (experimental control) change. We are only beginning to comprehend

become possible. what truly causes change in addictive behavior.
This current situation is sim ilar in many ways There is every indication that there are powerful
to the state of knowledge regarding relapse. The forces yet to be understood, and perhaps chan-
phenomenon of return to prior addictive behav- neled. Change, it seems, m ay not behave by the
ior is clear, and presents a com pelling need for rules we have imagined. I can think of no more
understanding. A variety of explanatory m odels fascinating or satisfying topic to which to devote
have been proposed, and a few speci® ed inter- one’ s scienti® c life than a solution to this endur-
ventions have been demonstrated to exert longi- ing puzzle.
tudinal in¯ uence on the probability of relapse.
Com parative testing of predictive models of re-
lapse has only begun, and should lead to a A cknow ledgem ents
clearer understanding of which variables do in The address that formed the basis for this article
fact predict addictive behavior prospectively, and was originally presented at the Addiction Re-
how they interact (Allen, Lowman & M iller, search Foundation in Toronto, Ontario, Canada.
1996). Once such predictive relationships are Preparation of this work was supported in part
better understood, the way is paved for clarifying by grant K05-AA00133 from the National Insti-
why certain m ethods of relapse prevention work, tute on Alcohol Abuse and Alcoholism.
and for improving the effectiveness of such inter-
ventions. R eferences
One example of this approach is found in a A LLEN , J., L OW M AN , C. & M ILLER , W . R. (Eds) (1996 )
recently com pleted prospective study of relapse Perspectives on precipitants of relapse, A ddiction, 91,
to drinking (M iller et al. , 1996), designed to test supplement.
A NNIS , H . M . (1986 ) A relapse prevention m odel for
Marlatt’ s conceptual model of relapse as a devel- treatment of alcoholics, in: M ILLER , W . R. &
opmental process (Cumm ings, Gordon & M ar- H EATHER , N. (E ds) Treating Addictive Behaviors: pro-
latt, 1980). We sought to measure each cesses of change, pp. 407± 433 (N ew York, Plenum
component of the model prospectively, every 2 Press).
A M RH EIN , P. C. (1992 ) T he comprehension of quasi-
months for a year from the onset of treatment for
perform ative verbs in verbal com mitments: new evi-
alcohol problem s. We wanted to determ ine dence for componential theories of lexical meaning,
whether the theoretical com ponents of M arlatt’ s Journal of M emory and Language, 31, 756 ± 784.
model actually would predict the occurrence of B ANDU RA . A. (1977 ) Self-ef® cacy: toward a unifying
relapse. The extent of exposure to high-risk situ- theory of behavioral change, Psychological Review,
84, 191± 215.
ations was unrelated to relapse risk, but the B IEN , T. H., M ILLER , W . R. & B OROU G HS , J. M.
client’ s coping style was a signi® cant predictor of (1993a ) M otivational interviewing with alcohol out-
outcome. Avoidant coping was associated with patients, Behaviou ral and Cognitive Psychotherapy, 21,
an increased risk of relapse, whereas positive 347± 356.
B IEN , T. H., M ILLER , W . R. & T O NIG AN , J. S. (1993b )
thinking was linked to lower relapse risk. Once
Brief interventions for alcohol problem s: a review ,
coping behavior had been taken into account, Addiction, 88, 315± 336 .
neither self-ef® cacy nor alcohol expectancies in- B ROW N , J. M. & M ILLER , W . R. (1993 ) Impact of m oti-
creased prospective predictability of relapse. The vational interviewing on participation in residential
extent of client endorsem ent of a disease model alcoholism treatment, Psychology of Addictive Behav -
iors, 1, 211 ± 218.
of alcoholism (M oyers & Miller, 1993), however, C O X , W . M., K LING ER , E. & B LOU NT , J. P. (1991 ) Al-
did contribute unique variance; disease beliefs cohol use and goal hierarchies: system atic motiva-
were consistently associated with a higher risk of tional counseling w ith alcoholics, in M ILLER , W . R. &
relapse across all follow-up points. R O LLNICK , S., Motivational Interview ing: preparing
people to change addictive behavior, pp. 260 ± 271 (N ew
It would be fruitful, I believe, to use such
York, Guilford Press).
multivariate m odeling to disentangle the various C U M M INGS , C., G ORDON , J. R. & M ARLATT , G . A.
strands of motivation for change, and to under- (1980 ) Relapse: prevention and prediction, in:
stand how they emerge and interact over time. M ILLER , W . R. (Ed.) The A ddictive Behaviors: treat-
Beneath the descriptive stages of change lie mul- ment of alcoholism, drug abuse, smoking, and obesity,
pp. 291± 321 (N ew York, Pergam on Press).
tiple tides, the processes that som etimes cross, D AVIDSO N , R. (1992 ) Prochaska and DiClem ente’ s
sometim es cancel each other out, and now and model of change: a case study? British Journal of
again converge to produce a great wave of Addiction, 87, 821± 823 .
 W hy do people change addictive behavior? 171

D OU GH ER , M . J. & M ARKH AM , M. R. (1996 ) Stimulus L EVENTH AL , H . (1971 ) Fear appeals and persuasion:
classes and the untrained acquisition of stimulus- the differentiation of a m otivational construct,
function, in Z ENDALL , T. R. & S W EETS , P. M . (E ds) American Journal of Public H ealth, 61, 1208± 1224.
Stimulus class functions in humans and animals, Ad- L OG AN , F. M . (1993 ) Anim al learning and motivation
vances in Psychology series (Am sterdam , North Hol- and addictive drugs, Psychological Reports, 73, 291±
land). 306.
F ARKAS , A. J., P IERC E , J. P., Z HU , S. - H, R O SB RO OK , B. et M ARLATT , G . A. & G ORDON , J. R. (1985 ) R elapse Pre-
al. (1996 ) Addiction versus stages of change m odels vention: maintena nce strategies in the treatmen t of addic-
in predicting smoking cessation, Addiction, 91, tive behaviors (N ew York, G uilford Press).
1271± 1280. M C L ELLAN , A. T., W OO DY , G. E., L UBO RSKY , L. &
F INNEY , J. W ., M OOS , R. & C HAN , D. (1981 ) Length of G OEH L , L. (1988 ) Is the counselor an ª active in-
stay and program com ponent effects in the treatm ent gredientº in substance abuse rehabilitation? An
of alcoholism : a com parison of tw o techniques for examination of treatm ent success among four coun-
process analysis, Journal of Consulting and Clinical selors, Journal of Nervous and Menta l Disease, 176,
Psychology, 49, 120± 131. 423± 430.
F ULLER , R. K., B RANC HEY , L., B RIG HTW ELL , D. R. et al. M EYERS , R. J. & S M ITH , J. E. (1995 ) Clinical Guide to
(1986 ) Disul® ram treatm ent of alcoholism : a Veter- Alcohol Treatment: the community reinforcem ent ap-
ans Adm inistration cooperative study, Journal of Ner- proach (N ew York, G uilford Press).
vous and M ental Disease, 256 , 1449 ± 1455 . M ILLER , W . R. (1978 ) Behavioral treatm ent of problem
H ANDM AKER , N. S. (1993 ) Motivating pregnant drinkers: a com parative outcome study of three con-
drinkers to abstain: prevention in prenatal care clin- trolled drinking therapies, Journal of Consulting and
ics, Doctoral dissertation, U niversity of New Mex- Clinical Psychology, 46, 74± 86.
ico. M ILLER , W . R. (1985 ) M otivation for treatm ent: a
H ARRIS , K. B. & M ILLER , W . R. (1990 ) Behavioral self- review w ith special emphasis on alcoholism , Psycho-
control training for problem drinkers: com ponents of logical Bulletin, 98, 84± 107.
M ILLER , W . R. & B AC A , L. M. (1983 ) Tw o-year follow-
ef® cacy, Psychology of Addictive Behaviors, 4, 82± 90.
up of bibliotherapy and therapist-directed controlled
H EATHER , N., R OLLNIC K , S., B ELL , A. & R ICHM OND , R.
drinking training for problem drinkers, Behavior
(1996 ) Effects of brief counselling among m ale
Therapy, 14, 441 ± 448.
heavy drinkers identi® ed on general hospital wards,
M ILLER , W . R., B ENEF IELD , R. G . & T O NIG AN , J. S.
Drug and A lcohol Review, 15, 29± 38.
(1993 ) Enhancing motivation for change in problem
H ERM ANS , H. J. M., K EM PEN , H. J. G . & VAN L OO N , R.
drinking: a controlled com parison of two therapist
J. P. (1992 ) The dialogical self: beyond individualism
styles, Journal of Consulting and Clinical Psychology,
and rationalism , Am erican Psychologist, 47, 23± 33.
61, 455± 461.
H OLDER , H. D. & B LO SE , J. O. (1992 ) The reduction of
M ILLER , W . R. & B ROW N , J. M. (1991 ) Self-regulation
health care costs associated with alcoholism treat-
as a conceptual basis for the prevention and treat-
m ent: a 14-year longitudinal study, Journal of Studies
ment of addictive behaviours, in: H EATH ER , N,
on A lcohol, 53, 293± 302. M ILLER , W . R. & G REELEY , J. (E ds) Self-control and
H U M PHREYS , K., M OO S , R. H . & C OHEN , C. (1997)
the Addictive Behaviou rs, pp. 3± 79 (Sydney, M axw ell
Social and com m unity resources and long-term re- Macmillan Publishing Australia).
covery from treated and untreated alcoholism , M ILLER , W . R., B RO WN , J. M ., S IM PSON , T. L. et al.
Journal of Studies on Alcohol, 58, 231± 238 . (1995 ) W hat works? A m ethodological analysis of
INSTITU TE OF M EDICIN E , N ATIONA L A CADEM Y OF S CI- the alcohol treatm ent outcom e literature, in H ESTER ,
ENCES (1990) Broadening the Base of Treatmen t for
R. K. & M ILLER , W . R. (Eds) Handbook of Alcoholism
Alcohol Problems (W ashington, DC, National Acad- Treatment Approaches: effectiv e alternativ es, 2nd edn,
emy Press). pp. 12± 44 (Boston, MA, Allyn and Bacon).
J ANIS , I. L. & M ANN , L. (1977) Decision-making: a psy- M ILLER , W . R. & C ’ DE B ACA , J. (1994 ) Quantum
chological analysis of con¯ ict, choice, and commitment change: toward a psychology of transformation, in:
(N ew York, Free Press). H EATHER TO N , T. & W EINBERG ER , J. (E ds) Can Per-
J ESSOR , R. (1987 ) Problem -behavior theory, psychoso- sonality Change?, pp. 253 ± 280 (W ashington, DC,
cial developm ent, and adolescent problem drinking, Am erican Psychological Association).
British Journal of A ddiction, 82, 331 ± 342. M ILLER , W . R., G RIBSKO V , C. J. & M ORTELL , R. L.
K ANF ER , F. M. (1986 ) Implications of a self-regulation (1981 ) Effectiveness of a self-control m anual for
m odel of therapy for treatm ent of addictive behav- problem drinkers with and w ithout therapist contact,
iors, in: M ILLER , W . R. & H EATHER , N. (E ds) International Journal of the Addictions, 16, 1247±
Treating Addictiv e Behaviors: processes of change, pp. 1254.
29± 47 (N ew York, Plenum Press). M ILLER , W . R. & H EATHER , N. (1986) Treating Addic-
K LU G ER , A. N. & D E N ISI, A. (1996 ) The effects of tive Behaviors: processes of change (N ew York, Plenum
feedback interventions on perform ance: a historical Press).
review , a m eta-analysis, and a prelim inary feedback M ILLER , W . R. & H ESTER , R. K. (1986 ) Inpatient alco-
intervention theory, Addictive Behaviors, 119, 254± holism treatm ent: who bene® ts? American Psycholo-
284. gist, 41, 794± 805 .
L EAKE , G . J. & K ING , A. S. (1977 ) Effect of counselor M ILLER , W . R. & M U NÄ OZ , R. F. (1982 ) H ow to Control
expectations on alcoholic recovery, Alcohol H ealth Your Drinking, 2nd edn (Albuquerque, U niversity of
and R esearch W orld, 11, 16± 22. New M exico Press).
172 W illiam R. M iller

M ILLER , W . R. & R OLLNIC K , S. (1991) M otivational Alcoholism: Clinical and Experimental Research, 20,
Interview ing: preparing people for change (N ew York, 847± 852.
G uilford Press). P REM AC K , D. (1970 ) M echanisms of self-control, in:
M ILLER , W . R. & S ANC HEZ , V. C. (1994 ) Motivating H U NT , W . A. (Ed.) Learning M echanisms in Smoking,
young adults for treatment and lifestyle change, in pp. 107± 123 (C hicago, Aldine).
H O WARD , G . (E d.) Issues in A lcohol Use and Misuse by P RO CHASKA , J. O., D IC LEM ENTE , C. C. & N ORCRO SS , J.
Young Adults, pp. 55± 82 (N otre Dame, IN, U niver- C. (1992 ) In search of how people change: applica-
sity of Notre Dame Press). tions to addictive behaviors, American Psychologist,
M ILLER , W . R., S OVEREIG N , R. G . & K REG E , B. (1988) 47, 1102± 1114.
Motivational interviewing with problem drinkers: II. P RO JEC T M ATC H R ESEARC H G RO UP (1993 ) Project
The Drinker’ s Check-up as a preventive interven- MATCH: rationale and m ethods for a multisite
tion, Behaviou ral Psychotherapy, 16, 251 ± 268. clinical trial m atching patients to alcoholism treat-
M ILLER , W . R. & T AYLO R , C. A. (1980 ) Relative effec- ment, Alcoholism: Clinical and Experimental Research,
tiveness of bibliotherapy, individual and group self- 17, 1130± 1145.
control training in the treatm ent of problem P RO JEC T MATC H R ESEARC H G ROUP (1997 ) Matching
drinkers, Addictive Behaviors, 5, 13± 24. alcoholism treatments to client heterogeneity: Proj-
M ILLER , W . R., T AYLOR , C. A. & W EST , J. C. (1980 ) ect MAT CH postreatment drinking outcom es,
Focused versus broad spectrum behavior therapy for Journal of Studies on Alcohol, 58, 7± 29.
problem drinkers, Journal of Consulting and Clinical R OB SO N , R. A. H., P AU LUS , I. & C LARKE , G . G . (1965 )
Psychology, 48, 590± 601. An evaluation of the effect of a clinic treatm ent
M ILLER , W . R., W ESTERB ERG , V. S., H ARRIS , R. J. & program on the rehabilitation of alcoholic patients,
T ONIGAN , J. S. (1996 ) W hat predicts relapse? Quarterly Journal of Studies on Alcohol, 26, 264± 278 .
Prospective testing of antecedent models, Addiction, R OG ERS , R. & P RENTIC E- D U NN , S. (1997 ) Protection
91 (suppl.), S155 ± S171. motivation theory, in: G O CHM AN , D. (E d.) Handbook
M O NTG OM ERY , H . A., M ILLER , W . R. & T ONIGAN , J. S. of Health Behavio r Research: Vol. 1. Determinants of
(1995 ) Does Alcoholics Anonymous involvem ent health behavior: personal and social, pp. 113 ± 132
predict treatment outcom e? Journal of Substance (N ew York, Plenum Press).
Abuse Treatment, 2, 241± 246. R OKEAC H , M . (1973) The Nature of Human Values
M O OS , R. H ., F INNEY , J. W . & C RONKIT E , R. C. (1990 ) (N ew York, Free Press).
Alcoholism Treatment: context, process, and outcome S AU NDERS , B., W ILKINSO N , C. & P HILLIPS , M . (1995)
(N ew York, Oxford University Press). The impact of a brief m otivational intervention w ith
M O YERS , T . B. & M ILLER , W . R. (1993 ) Therapists’ opiate users attending a methadone programm e,
conceptualizations of alcoholism : m easurement and Addiction, 90, 415± 424 .
implications for treatm ent, Psychology of Addictive S OBELL , L. C., S OBELL , M. B. & T ONEATT O , T . (1991)
Behaviors, 7, 238± 245. Recovery from alcohol problem s without treatm ent,
N AJAVITS , L. M. & W EISS, R. D. (1994 ) Variations in in: H EATHER , N., M ILLER , W . R. & G REELEY , J. (Eds)
therapist effectiveness in the treatm ent of patients Self-control and the A ddictive Behaviou rs, pp. 198 ± 242
with substance use disorders: an empirical review, (Botany, Maxwell Macmillan Publishing Australia).
Addiction, 89, 679± 688. T RUAX , C. B. & C ARKHU FF , R. R. (1967 ) Toward Effec-
P ATTERS ON , G . A. & F ORG ATCH , M . S. (1985 ) Thera- tive Counseling and Psychotherapy (C hicago, Aldine).
pist behavior as a determ inant for client noncom pli- V AILLAN T , G . (1983) The Natural History of Alcoholism:
ance: a paradox for the behavior m odi® er, Journal of causes, patterns, and paths to recovery (C ambridge,
Consulting and Clinical Psychology, 53, 846± 851. MA, Harvard University Press).
P ETTINA TI , H . M., B ELDEN , P. P., E VANS , B. D., V ALLE , S. K. (1981 ) Interpersonal functioning of alco-
R U ETSC H , C. R., M EYERS , K. & J ENSEN , J. M . (1996 ) holism counselors and treatment outcom e, Journal of
The natural history of outpatient alcohol and drug Studies on Alcohol, 42, 783 ± 790.
abuse treatm ent in a private health care setting,