Journal of Consulting and Clinical Psychology Copyright 1995 by the American Psychological Association, Inc.

1995, Vol.63, No. 2,214-220 0022-006X/95/S3.00

SPECIAL FEATURE

Hypnosis as an Adjunct to Cognitive-Behavioral Psychotherapy:

A Meta-Analysis
Irving Kirsch, Guy Montgomery, and Guy Sapirstein
University of Connecticut

A meta-analysis was performed on 18 studies in which a cognitive-behavioral therapy was compared

with the same therapy supplemented by hypnosis. The results indicated that the addition of hypnosis
substantially enhanced treatment outcome, so that the average client receiving cognitive-behavioral
hypnotherapy showed greater improvement than at least 70% of clients receiving nonhypnotic treat-
ment. Effects seemed particularly pronounced for treatments of obesity, especially at long-term fol-
low-up, indicating that unlike those in nonhypnotic treatment, clients to whom hypnotic inductions
had been administered continued to lose weight after treatment ended. These results were particu-
larly striking because of the few procedural differences between the hypnotic and nonhypnotic
treatments.

Once relegated to the realm of the supernatural, hypnosis is Clinical hypnosis is a procedure in which a therapist suggests
increasingly accepted as a legitimate therapeutic procedure that a client experience changes in sensation, perception,
(Rhue, Lynn, & Kirsch, 1993). A century ago, hypnotherapy thought, and behavior. The hypnotic context is established by
often consisted of a hypnotic induction, followed by the admin- an induction procedure that usually includes instructions for
istration of suggestions for symptom removal. Consequently, relaxation. Hypotheses about how this procedure might en-
hypnotherapy has been viewed by some writers as a mode of hance therapy vary with theoretical conceptions of hypnosis.
therapy that might be compared with psychodynamic, cogni- Most therapists believe that hypnotic inductions produce an al-
tive-behavioral, or other therapeutic approaches (e.g., Smith, tered state of consciousness in susceptible individuals (see
Glass, & Miller, 1980). However, suggestions for symptom relief Kirsch, 1993). Among the presumed characteristics of the hy-
play a relatively minor role in contemporary hypnotherapy. In- pothesized hypnotic state are hypersuggestibility, more vivid
stead, hypnotherapy generally consists of the addition of hyp- imagery, more primary process thinking, greater availability of
nosis to some recognized form of psychotherapy (Rhue et al., childhood memories, and a tolerance of logical incongruities
1993). As a result, the question to be asked is not whether hyp- often referred to as "trance logic" (Fromm, 1992; Hilgard,
nosis works better than another treatment but rather whether it 1965; Orne, 1959). From a traditional state perspective, the
enhances the effectiveness of a treatment. benefits of adding hypnosis to treatment are due to these char-
Before 1980, research on the efficacy of hypnotherapy was acteristics of the hypothesized trance state.1
largely confined to psychodynamic hypnotherapy (Smith et al., In contrast to this view, cognitive-behavioral theorists have
1980). More recently, empirical studies have focused on the use rejected the hypothesis that there is a distinctly hypnotic state
of hypnosis in behavior therapy, cognitive therapy, and cogni- of consciousness (Barber, 1969; Kirsch, 1990; Sarbin & Coe,
tive-behavior therapy (Spinhoven, 1987). The distinction be- 1972; Spanos & Chaves, 1989). From a nonstate perspective,
tween these latter modes of therapy is not entirely clear. Cogni- hypnosis has been hypothesized to augment therapy outcome
tive processes (e.g., imagery) are a component of many behav- through its effects on clients' beliefs and expectations (Barber,
ior therapies, and behavioral tasks are a component of virtually 1985; Coe, 1993; Fish, 1973; Kirsch, 1985, 1990). Thus, state
all cognitive therapies. In this article, we use the term cognitive- theorists and cognitive-behavioral theorists agree that hypnosis
behavioral psychotherapy to refer to treatment procedures de- can enhance treatment effects, albeit for different reasons. The
scribed as behavioral, cognitive, or cognitive-behavioral. purpose of this review is to assess the empirical data bearing on
this question of whether cognitive-behavioral psychotherapies
are enhanced by the addition of hypnosis.
Irving Kirsch, Guy Montgomery, and Guy Sapirstein, Department of
Psychology, University of Connecticut.
1
The research reported in this article was supported by a grant from Some hypnosis theorists use the terms state and trance in a purely
the University of Connecticut Research Foundation. descriptive sense, without ascribing any causal properties to the concept
Correspondence concerning this article should be addressed to Irving (Hilgard, 1969; Kihlstrom, 1985). With respect to the altered state is-
Kirsch, Department of Psychology, U-20, Room 107, University of sue, this view is virtually identical to the cognitive-behavioral concep-
Connecticut, 406 Babbidge Road, Storrs, Connecticut 06269-1020. tion of hypnosis.

214
 HYPNOSIS AS AN ADJUNCT 215

There are a number of reasons for supposing that if hypnosis and college students. Participants were assigned to treatment randomly
enhances psychotherapy outcome, its effects are likely to be rel- in 16 of these comparisons and sequentially in 2 (Edelson & Fitzpatrick,
atively modest. First, most of the procedures conducted in hyp- 1989; Lazarus, 1973). Method of participant assignment was unclear
in two of the reports (Howard & Reardon, 1986; O'Brien, Cooley,
notherapy are the same as those conducted in nonhypnotic psy- Ciotti, & Henninger, 1981).
chotherapy. This is a consequence of the fact that hypnosis is an Effect sizes ( d ) were calculated for each outcome variable as the stan-
adjunct to therapy rather than a mode of therapy. Second, cli- dardized mean difference between the hypnosis group and the corre-
ents vary in their responsiveness to hypnosis. From a traditional sponding no-hypnosis group. Where sufficient data were not provided
state viewpoint, only those with sufficient hypnotic talent or for direct calculation of effect sizes, they were estimated using the pro-
ability are likely to benefit substantially from the addition of cedures described by Smith et ah (1980).
hypnosis to a treatment (Levitt, 1993). Third, clients vary in Three different units of analysis can be used in calculating mean
their attitudes and expectancies regarding hypnosis. Enhance- effect sizes, individual dependent variables (e.g., Smith et ah, 1980),
ment of outcome should be limited to clients with positive atti- treatments (e.g., Barker, Funk, & Houston, 1988), or studies (e.g., Ly-
tudes and expectations, whereas a degradation of treatment out- ons & Woods, 1991). In the latter two methods, effect sizes are averaged
across dependent variables. Using individual dependent variables as the
come might be expected among clients with negative attitudes
unit of analysis results in substantial violations of the assumption of
(Kirsch, 1990, 1993). Finally, typical hypnotic inductions statistical independence when standard inferential statistical tests
closely resemble conventional relaxation training. In fact, all are applied to the results. In contrast, statistical independence is
that is needed to convert relaxation training into a hypnotic assured when only one effect size is used from each study (Hunter &
induction is the addition of the word hypnosis. Instead of saying Schmidt, 1990).
"more and more deeply relaxed," the therapist says "more and Some studies contain evaluations of more than one treatment. In
more deeply hypnotized." Because relaxation training is a fre- such cases, the use of a single effect size for each study obscures differ-
quent component of behavior therapy, the addition of hypnosis ences between these treatments. In the present sample, there were two
to behavior therapy may consist of little more that the use of the studies in which more than one hypnotic treatment were compared with
word "hypnosis." a comparable nonhypnotic treatment. Goldstein (1981) included two
The small magnitude of anticipated mean effects, combined hypnosis groups, in one of which the participants were given an arm
levitation suggestion as a means demonstrating the effects of hypnosis
with the relatively small samples used in many therapy outcome and thereby enhancing the participants' treatment outcome expecta-
studies, are likely to lead to inconsistent outcomes, in which tions. Barabasz and Spiegel (1989) also used two hypnosis groups, one
some studies show significant effects and others do not. In situ- in which the same hypnotic suggestions were used for all participants
ations of this sort, meta-analyses can provide more definitive and another in which suggestions were individualized on the basis of
answers than individual studies, narrative reviews, or box scores participant characteristics.
of significant results (Hunter & Schmidt, 1990). Meta-analysis Treatment was chosen as the unit of analysis in this meta-analysis
allows comparison of outcomes across studies by the calculation because it avoids shortcomings associated with other options. Using
of effect sizes, defined as the standardized mean difference be- studies as the unit of analysis would not have allowed complete assess-
tween the experimental group and the control group. The meta- ment of variations in cognitive-behavioral hypnotherapy, thereby im-
analysis reported here assesses the effect of adding hypnosis to peding the search for moderator variables. Using individual effects as
the unit of analysis would have biased the results in the direction of
cognitive-behavioral psychotherapy. Because hypnosis is not a
studies with large numbers of dependent variables. Calculating a mean
treatment in its own right, we limited our sample of studies to effect for each treatment avoided both of these shortcomings. Also, be-
those in which a cognitive-behavioral treatment administered cause there were only two studies in which more than one hypnotic
in a hypnotic context was compared to the same treatment ad- treatment was included, the use of treatment as the unit of analysis
ministered without hypnosis. In calculating effect sizes, the affected statistical independence only minimally. As a further precau-
nonhypnotic treatment was treated as the control condition. tion, standard inferential statistics were replaced by the calculation of
confidence intervals calculated as 1.96 times the standard deviation of
the sampling error (i.e., 1.96 times the standard deviation of the ob-
Method served effect sizes divided by the number of effects; Hunter & Schmidt,
1990, pp. 437-438).
Studies of cognitive-behavioral hypnotherapy were obtained from
previous reviews (Rhue et ah, 1993; Spanos, 1991; Spinhoven, 1987)
and a computer search of the PsycLIT database from 1974 to 1993 using Results
the search terms, cognitive-behavioral hypnotherapy, hypnosis and psy-
chotherapy, hypnosis and behavior therapy, and hypnotherapy or hypno- Presenting problems, treatments, sample sizes, and effect
sis and adjunct to therapy. Inclusion criteria were as follows: (a) A cog- sizes are presented in Table 1. The mean effect size across stud-
nitive-behavioral psychotherapy was administered to at least one group ies was .87 standard deviations. This effect differed significantly
of participants in a hypnotic context; (b) the same therapy was admin- from zero, indicating that hypnosis enhanced the efficacy of
istered to at least one group of clients in a nonhypnotic context; and (c)
sufficient data were reported to allow calculation of effect sizes.
cognitive-behavioral treatments. Inspection of Table 1 reveals a
These criteria resulted in a sample of 18 studies in which 20 hypnotic wide range in sample size, which is a source of potential bias.
treatments were compared with identical nonhypnotic treatments and Calculation of the correlation between sample size and effect
from which we were able to calculate the magnitude of 90 effects. These size indicated that significantly larger effects were reported in
comparisons involved 577 participants. Fourteen treatment compari- studies with larger samples (r = . 50, p< .05). To correct for this
sons were derived from data of clinical patients as participants, 5 used bias, we weighted effects by the size of the samples from which
college students as participants, and 1 involved a mix of clinical patients they were obtained and calculated the mean weighted effect size
216 I. KIRSCH, G. MONTGOMERY, AND G. SAPIRSTEIN

Table 1
Description of Studies and Mean Effect Sizes
Study Presenting problem Cognitive-behavioral treatment
McAmmond et al. (1971) Pain Relaxation 18 -0.20
Borkovec & Fowles (1973) Insomnia Relaxation 18 -0.10
Deabler etal. (1973) Hypertension Relaxation 30 0.51
Lazarus(1973) Mixed Varied 20 1.45
Sullivan etal. (1974) Anxiety Relaxation 16 1.40
Graham etal. (1975) Insomnia Relaxation 22 -0.01
Bornstein & Devine (1980) Obesity Covert modeling 18 0.83
Deyoub&Wilkie(1980) Obesity Imagery + coping suggestions 48 -0.17
Goldstein (1981) Obesity Self-monitoring + stimulus control + self- 40 5.57
(Hypnosis with arm levitation) reinforcement
Goldstein (1981) (hypnosis Obesity Self-monitoring + stimulus control + self- 40 -0.08
without arm levitation) reinforcement
O'Brien etal. (1981) Snake phobia Systematic desensitization 18 0.73
Wadden & Flaxman (1981) Obesity Covert modeling 22 -0.22
Bolocofsky etal. (1985) Obesity Self-monitoring + goal setting + stimulus control 109 3.65
Howard & Reardon (1986) Self-concept & athletic Cognitive restructuring 16 0.02
performance
Barabasz & Spiegel (1989 Obesity Self-monitoring + goal setting 30 0.62
[standard suggestions])
Barabasz & Spiegel (1989 Obesity Self-monitoring + goal setting 29 0.75
[individualized suggestions])
Edelson & Fitzpatrick (1989) Chronic pain Cognitive strategy 18 0.16
Tosi etal. (1989) Duodenal ulcer Cognitive restructuring 12 0.93
Tosi etal. (1992) Hypertension Cognitive restructuring 21 1.11
Schoenberger(1993) Public speaking Relaxation + imagery + cognitive restructuring 32 0.40
anxiety + in vivo practice

( D ) following the procedures described by Hunter and Schmidt ence of a moderator variable. Although the variances of popu-
(1990). This revealed a significant effect of 1.36 standard devi- lation effect sizes suggested the presence of a moderator only for
ations resulting from the addition of hypnosis to cognitive-be- physiological outcomes, we were also interested in examining
havioral psychotherapy. This indicates that the average client theoretically predicted potential moderators. Therefore, we be-
receiving cognitive-behavioral hypnotherapy showed more im- gan assessing possible moderating variables using the treatment
provement than 90% of clients receiving nonhypnotic effect sizes listed in Table 1.
treatment.2 The use of hypnosis in psychotherapy entails the provision of
Inspection of Table 1 reveals two effects that might be classi- a hypnotic induction followed by therapeutic suggestions. Most
fied as outliers (Bolocofsky, Spinier, & Coulthard-Morris, 1985; hypnotic inductions (including all of those described in the
Goldstein, 1981). To provide conservative estimates of effect studies reviewed here) contain relaxation instructions that are
sizes, we winsorized with g = 1 and g = 2, which resulted in very similar to those used in relaxation training. In some of the
weighted mean effect sizes of 1.23 and .66, respectively. The studies reviewed here, the only difference between the hypnotic
more conservative of these estimates indicates that the average induction and the relaxation instructions used in the nonhyp-
client receiving therapy with hypnosis was better off at the end notic condition was the use of the term hypnosis (e.g., Lazarus,
of it than 75% of clients receiving the same therapy without 1973; Schoenberger, 1993). Because relaxation training was
hypnosis. used in only some of the nonhypnotic cognitive-behavioral
Besides calculating overall effects, we examined effect size as treatments described in this review, the enhancement of treat-
a function of type of dependent variable. Physiological variables ment outcome that we observed may have been due to relax-
were assessed in 12 studies, behavioral measures in 5 studies, ation instructions rather than to other aspects of hypnosis.
and self-report measures in 9 studies. Mean unweighted ( d ) and In many of the studies, suggestions other than those con-
weighted (D) effect sizes for each category of dependent vari- tained in the nonhypnotic treatment were included in the hyp-
able are reported in Table 2, along with the variances of the notic treatment. As noted earlier, a hand levitation suggestion
weighted estimates of the population effect sizes. Neither was included in one of the hypnotic conditions in the Goldstein
weighted nor unweighted effect sizes differed significantly as a (1981) study. Similarly, Schoenberger (1993) added brief direct
function of type of measure, and each was significantly greater suggestions for symptom improvement to her treatment when
than zero.
Although the mean effect for the addition of hypnosis to cog- 2
Because of the exceptional magnitude of this effect, we recalculated
nitive-behavioral psychotherapy was significantly greater than the weighted effect using the procedure described by Hedges & Olkin
zero, the variance of the overall population effect sizes and that (1985) for ^V". This yielded the same mean weighted effect as that
of physiological measures were very large, indicating the pres- obtained using the Hunter and Schmidt (1990) procedure.
 HYPNOSIS AS AN ADJUNCT 217

Table 2 standard deviations were reported, although both effect sizes

Population Effect Sizes as a Function of Type of Variable differed significantly from zero. There was also substantial vari-
ation in estimated effect sizes, indicating that this methodologi-
Type of Variation cal difference did not fully account for the observed lack of ho-
measure n d D oiD
mogeneity in effect sizes. Similarly, the mean weighted effect
Physiological 12 0.95 1.42 3.15 size for treatments of obesity was more than triple that of treat-
Behavioral 5 0.73 0.73 0.08 ments for other disorders, and both effect sizes differed signifi-
Self-report 9 0.60 0.58 0.10 cantly from zero. However, the variance in effect sizes in studies
Combined 20 0.87 1.36 2.94 of obesity was so large that the difference was not statistically
significant. In contrast, the variance in effect sizes for treat-
ments of presenting problems other than obesity was negligible,
and the variance in exactly calculated effect sizes was relatively
it was conducted in a hypnotic context. Thus, a second potential low.
moderating variable is the addition of suggestions not included The data indicate that the as-yet undiscovered moderating
in the nonhypnotic treatment. variable affected only physiological variables and studies in
Hypnosis may enhance the effectiveness of treatment for which obesity was the focus of treatment. Also, it was more ev-
some problems but not for others. Wadden and Anderton ident in estimated effect sizes than in exactly calculated effects.
(1982), for example, hypothesized that hypnosis might have The pattern of overlap between these variables suggested to us
special value in the treatment of "nonvoluntary" disorders (i.e., that presenting problem was central to the as yet undiscovered
pain, warts, asthma) but not in the treatment of disorders of moderating variable. The dependent variables of studies on
"self-initiated" behavior (i.e., obesity, cigarette smoking, obesity were limited to a single physiological measure (weight).
alcoholism). Although a variety of presenting problems were Similarly, all but two of the effects for obesity were estimated.
treated in the studies we found, in most instances the number Our suspicion was confirmed by separate analyses of the six
of studies per presenting problem was too small for meaningful estimated obesity effects and the five estimated effects involving
comparison of effect sizes. However, obesity was the presenting other presenting problems, which indicated inflated effect sizes
problem in eight of the treatment comparisons under review, and substantial variance for estimated obesity effects (D = 2.53,
allowing a comparison of the effectiveness of adding hypnosis to variance = 4.27), but not for estimated effects for other pre-
the treatment obesity with that of adding it to the treatment of senting problems (D = . 13, variance = .07).
various other problems (e.g., pain, insomnia, hypertension, and In a narrative review of the use of hypnosis in weight reduc-
anxiety). tion treatments, Levitt (1993) noted that participants in one
Participants for half of the treatment comparisons were solic- hypnotic treatment program (Bolocofsky et al., 1985) contin-
ited by advertisements or from college student bodies. Those for ued to lose weight over a 2-year period after the end of the pro-
the other comparisons consisted of patients who had sought or gram, whereas participants in nonhypnotic treatment did not.
were referred for treatment. Patients seeking treatment might Examination of obesity studies in this meta-analysis revealed a
be more distressed and more motivated for change, and these wide range in the length of time during which follow-up data
characteristics might interact with type of treatment, leading to
differential outcomes.
Finally, direct calculation of effect sizes from means and stan- Table 3
dard deviations was possible in only nine treatment compari- Population Effect Sizes as a Function of
sons. In the remaining 11, standard deviations were estimated Hypothesized Moderators
using the methods described by Smith et al. (1980). It is possi-
ble that these indirect procedures produced effect sizes that No. of
were different from those produced by calculation from exact treatments in Variation
data. Potential moderator comparison d D inD
The results of analyses of these potential moderating vari- Relaxation
ables are displayed in Table 3. They indicate that hypnotic en- In both treatments 11 0.77 1.51 2.21
hancement of therapeutic outcome is not due to the addition of In hypnotic treatment only 9 0.99 1.15 3.84
relaxation instructions nor to the addition of therapeutic sug- Suggestions
gestions. Nature of the participant population also did not affect Same in both treatments 14 0.63 1.20 2.21
More in hypnotic treatment 6 1.42 1.74 4.50
outcome. There was considerable variance in population effect Presenting problem
sizes regardless of whether relaxation was included in the con- Obesity 8 1.37 1.96 4.14
trol treatment, whether additional therapeutic suggestions were Other 12 0.53 0.52 0.06
added to the hypnotic treatment or whether the participants Participant population
Sought or referred for
had sought treatment or were solicited. This indicates that none treatment 10 1.24 1.42 3.19
of these variables was the source of the variation in the esti- Solicited 10 0.49 1.31 2.74
mated population effect sizes. Effect calculation method
Estimation of standard deviations resulted in significantly Exact 9 0.62 0.47 0.20
greater effects than those calculated from studies in which the Estimate 11 1.07 1.87 3.79
218 I. KIRSCH, G. MONTGOMERY, AND G. SAPIRSTEIN

Ibs. kg.
Without Hypnosis
0'
With Hypnosis

POSTTREATMENT ASSESSMENT INTERVAL

(Months)

Figure 1. Weight loss as a function of assessment interval and inclusion of hypnosis in treatment.

were collected (2-24 months). To determine if effect size was that were due to the addition of hypnosis, although the exact
influenced by duration of the follow-up period, we calculated magnitude of this effect is uncertain because of the failure to
the correlation between time of assessment and magnitude of report standard deviations in most of the weight reduction stud-
effect, using individual assessments as the unit of analysis. This ies. For the sake of future meta-analyses, we strongly advocate
revealed that after treatment ended, the effect of hypnosis in- the reporting of exact means (or adjusted means, if pretreat-
creased over time (r = .59, p < .02). Thus, the observed effect ment scores are available) and standard deviations as a precon-
of adding hypnosis to treatments of obesity was moderated by dition for publication of outcome studies.
the length of the follow-up assessment interval. The association In contrast to treatments of other presenting problems, the
between assessment interval and effect of treatment is graphi- effect of adding hypnosis to cognitive-behavioral treatment of
cally displayed in Figure 1. By using mean weight loss in place obesity did not become apparent until some time after treat-
of effect size, the problems associated with estimating standard ment had ended. Differences between hypnotic and nonhyp-
deviations are avoided and the clinical significance of the notic treatment of obesity increased up to 6 months after treat-
difference can be judged more accurately. ment ended and remained intact at 2-year follow-up. Further-
more, this phenomenon was independent of the effect size
Discussion estimation problem resulting from missing standard deviations.
It should be noted, however, that long-term follow-up data were
The results of this meta-analysis indicates a fairly substantial
effect as a result of adding hypnosis to cognitive-behavioral psy- reported only for obesity studies. Therefore, at least two inter-
chotherapies. Effect sizes were consistent for behavioral and pretations of these data are possible. First, it is possible that the
self-report measures of change and for all measures of change effects of hypnosis are particularly pronounced in the treatment
in studies of presenting problems other than obesity. This indi- of obesity, which is largely due to the failure of nonhypnotic
cates that hypnosis enhances the effects of cognitive-behavioral treatments to produce lasting change. Alternately, it is possible
psychotherapy across a broad range of problems. Our most con- that the advantages of adding hypnosis to cognitive-behavioral
servative estimates of this effect approximate 0.5 standard devi- treatment increases over time, regardless of presenting prob-
ations, indicating that the average client receiving cognitive-be- lem. Resolution of this issue will require studies with long-term
havioral hypnotherapy benefitted more than at least 70% of cli- follow-up data for problems other than obesity.
ents receiving the same treatment without hypnosis. Many scholars maintain that if treatment effects are due to
There are two factors that appear to account for the variance hypnosis, rather than a function of nonspecific variables, they
in physiological effects: presenting problem and length of fol- ought to be correlated with hypnotizability. Unfortunately, few
low-up. For problems other than obesity, the variance in of the studies considered in this review reported correlations of
weighted effect sizes was negligible, thus allowing clear inter- outcome with hypnotizability scores. In any case, correlations
pretation. We found a reliable effect of just over one half stan- between hypnotizability and treatment outcome do not provide
dard deviation caused by the addition of hypnosis to these treat- much information about hypothesized causal mechanisms, re-
ments. Weight reduction treatments showed even larger effects gardless of whether hypnotizability is assessed before or after
 HYPNOSIS AS AN ADJUNCT 219

treatment. If hypnotizability is assessed before treatment, the Spanos, & B. J. Wallace (Eds.), Imagination and Hypnosis. Amity-
person's response to hypnosis should affect his or her outcome ville, NY: Baywood.
expectations for a "hypnotic" treatment. If hypnotizability is •Deabler, H. L., Fidel, E., & Dillenkoffer, R. I. (1973). The use of re-
laxation and hypnosis in lowering high blood pressure. American
assessed after treatment, the effects of treatment could influence
Journal of Clinical Hypnosis, 16, 75-83.
their subsequent responses to a test of hypnotizability (Council, •Deyoub, P. L., & Wilkie, R. (1980). Suggestion with and without hyp-
Kirsch, & Grant, in press; Kirsch & Council, 1992). Thus, cor- notic induction in a weight reduction program. International Journal
relations between hypnotizability and treatment outcome of Clinical and Experimental Hypnosis, 28, 333-340.
might be indicators of expectancy effects, rather than effects of *Edelson, J., & Fitzpatrick, J. L. (1989). A comparison of cognitive-
some special hypnotic process.3 behavioral and hypnotic treatments of chronic pain. Journal of Clin-
In summary, the results of this meta-analysis indicate that ical Psychology, 45, 316-323.
hypnosis can be a useful adjunct to cognitive behavior therapy Fish, J. M. (1973). Placebo therapy. San Francisco: Jossey-Bass.
for a wide variety of problems, and it may be particularly Fromm, E. (1992). An ego-psychological theory of hypnosis. In E.
effective in treating obesity. The data indicating that hypnosis Fromm & M. Nash (Eds.), Contemporary hypnosis research (pp.
promotes long-term weight loss is particularly important, given 131 -148). New York: Guilford.
the finding that most obese individuals who lose weight in non- •Goldstein, Y. (1981). The effect of demonstrating to a subject that she
is in a hypnotic trance as a variable in hypnotic interventions with
hypnotic treatments soon regain it (Stunkard, 1972). Research obese women. International Journal of Clinical and Experimental
is needed to establish the range of treatments and conditions Hypnosis, 29, 13-23.
that can be enhanced by the addition of hypnosis and to inves- •Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analy-
tigate participant variables that might predict when hypnosis sis. Orlando, FL: Academic Press.
would be helpful and when it might be harmful. Nevertheless, Hilgard, E. R. (1965). Hypnotic susceptibility. New York: Harcourt,
the current data suggest that training in hypnosis should be in- Brace & World.
cluded routinely as a part of training in cognitive-behavioral Hilgard, E. R. (1969). Altered states of awareness. Journal of Nervous
treatments. and Mental Disease, 149, 68-79.
•Howard, W. L., & Reardon, J. P. (1986). Changes in the self-concept
3
Note that the variance in effect sizes is probably not related to and athletic performance of weight lifters through a cognitive-hyp-
differences in hypnotizability, because individual difference variables notic approach: An empirical study. American Journal of Clinical
of this sort would be more likely to affect within-study variance than Hypnosis, 28, 248-257.
between-study variance. Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Cor-
recting error and bias in research findings. Newbury Park, CA: Sage.
Kihlstrom, J. F. (1985). Hypnosis. Annual Review of Psychology, 36,
References 385-418.
Kirsch, I. (1985). Response expectancy as a determinant of experience
References preceded by an asterisk were included and behavior. American Psychologist, 40, 1189-1202.
in the meta-analysis. Kirsch, I. (1990). Changing expectations: A key to effective psychother-
*Barabasz, M., & Spiegel, D. (1989). Hypnotizability and weight loss apy. Pacific Grove, CA: Brooks/Cole.
in obese subjects. International Journal of Eating Disorders, 8, 335- Kirsch, I. (1993a). Cognitive-behavioral hypnotherapy. In J. W. Rhue,
341. S. J. Lynn, & I. Kirsch (Eds.), Handbook of clinical hypnosis (pp.
Barber, T. X. (1969). Hypnosis: A scientific approach. New York: Van 151-171). Washington, DC: American Psychological Association.
Nostrand Reinhold. Kirsch, I. (1993b). Professional opinions about hypnosis: Results of the
Barber, T. X. (1985). Hypnosuggestive procedures as catalysts for psy- APA Division 30 survey. Psychological Hypnosis, 2, 4-5.
chotherapies. In S. J. Lynn & J. P. Garske(Eds.), Contemporarypsy- Kirsch, I., & Council, J. R. (1992). Situational and personality corre-
chotherapies: Models and methods (pp. 333-375). Columbus, OH: lates of suggestibility. In E. Fromm & M. Nash (Eds.), Contemporary
Merrill. hypnosis research (pp. 267-292). New York: Guilford.
Barker, S. L., Funk, S. C., & Houston, B. K. (1988). Psychological treat- •Lazarus, A. A. (1973). "Hypnosis" as a facilitator in behavior therapy.
ment versus nonspecific factors: A meta-analysis of conditions that International Journal of Clinical and Experimental Hypnosis, 21,
engender comparable expectations for improvement. Clinical Psy- 25-31.
chology Review, 8, 579-594. Levitt, E. E. (1993). Hypnosis in the treatment of obesity. In J. W.
*Bolocofsky, D. K, Spinier, D., & Coulthard-Morris, L. (1985). Rhue, S. J. Lynn, & I. Kirsch (Eds.), Handbook of clinical hypno-
Effectiveness of hypnosis as an adjunct to behavioral weight manage- sis (pp. 511-532). Washington, DC: American Psychological
ment. Journal of Clinical Psychology, 41, 35-41. Association.
*Borkovec, T. D., & Fowles, D. C. (1973). Controlled investigation of Lyons, L. C., & Woods, P. J., (1991). The efficacy of rational-emotive
the effects of progressive and hypnotic relaxation on insomnia. therapy: A quantitative review of the outcome research. Clinical Psy-
Journal of Abnormal Psychology, 82, 153-158. chology Review, 11, 357-369.
*Bornstein, P. H., & Devine, D. A. (1980). Covert modeling-hypnosis •McAmmond, D. M., Davison, P. Q, & Kovitz, D. M. (1971). A com-
in the treatment of obesity. Psychotherapy: Theory, Research, and parison of the effects of hypnosis and relaxation training on stress
Practice, 77,272-276. reactions in a dental situation. American Journal of Clinical Hypno-
Coe, W. C. (1993). Expectations and hypnotherapy. In J. W. Rhue, S. J. sis, 13, 233-242.
Lynn, & I. Kirsch (Eds.), Handbook of clinical hypnosis (pp. 73-93). •O'Brien, R. M., Cooley, L. E., Ciotti, J., & Henninger, K. M. (1981).
Washington, DC: American Psychological Association. Augmentations of systematic desensitization of snake phobia through
Council, J. R., Kirsch, I., & Grant, D. L. (in press). Imagination, ex- posthypnotic dream suggestion. American Journal of Clinical Hyp-
pectancy and hypnotic responding. In R. G. Kunzendorf, N. K. nosis, 23, 231-238.
220 I. KIRSCH, G. MONTGOMERY, AND G. SAPIRSTEIN

Orne, M. T. (1959). The nature of hypnosis: Artifact and essence. *Sullivan, D. S., Johnson, A., & Bratkovitch, J. (1974). Reduction of
Journal of Abnormal Psychology, 58, 277-299. behavioral deficit in organic brain damage by use of hypnosis.
Rhue, J. W., Lynn, S. J., & Kirsch, I. (1993). Handbook of clinical Journal of Clinical Psychology, 30, 96-98.
hypnosis. Washington, DC: American Psychological Association. *Tosi, D. J., Judah, S. M., & Murphy, M. A. (1989). The effects of a
Sarbin, T. R., & Coe, W. C. (1972). Hypnosis: A social psychological cognitive experiential therapy utilizing hypnosis, cognitive restruc-
analysis of influence communication. New York: Holt, Rinehart & turing, and developmental staging on psychological factors associated
Winston. with duodenal ulcer disease: A multivariate experimental study.
Schoenberger, N. E. (1993). Effectiveness of a cognitive-behavioral hyp- Journal of Cognitive Psychotherapy, 3, 272-290.
notherapy for public speaking anxiety. Unpublished doctoral disser- *Tosi, D. J., Rudy, D. R., Lewis, J., & Murphy, M. A. (1992). The
tation, University of Connecticut. psychobiological effects of cognitive experiential therapy, hypnosis,
Smith, M. L., Glass, G. V., & Miller, T. I. (1980). The benefits of psy- cognitive restructuring, and attention placebo control in the treat-
chotherapy. Baltimore: John Hopkins University Press. ment of essential hypertension. Psychotherapy, 29, 274-284.
Spanos, N. P. (1991). Hypnosis, hypnotizability, and hypnotherapy. In Wadden, T. A., & Anderton, C. H. (1982). The clinical use of hypnosis.
C. R. Snyder&D. R. Forsyth (Eds.), Handbook of clinicalandsocial Psychological Bulletin, 91, 215-243.
psychology. New York: Pergamon Press. *Wadden, T. A., & Flaxman, J. (1981). Hypnosis and weight loss: A
Spanos, N. P., & Chaves, J. F. (Eds.). (1989). Hypnosis: The cognitive- preliminary study. International Journal of Clinical and Experimen-
behavioral perspective. Buffalo, NY: Prometheus Press. tal Hypnosis, 29, 162-173.
Spinhoven, P. (1987). Hypnosis and behavior therapy: A review. /«-
lernationalJournal of Clinical and Experimental Hypnosis, 35, 8-31.
Stunkard, A. J. (1972). Foreword. In R. B. Stuart & B. Davis, Slim Received November 15, 1993
chance in a fat world: Behavioral control of obesity. Champaign, IL: Revision received February 2, 1994
Research Press. Accepted February 6, 1994 •

Correction to Frick et al. (1991)

The article "Academic Underachievement and the Disruptive Behavior Disorders," by Paul J.
Frick, Randy W. Kamphaus, Benjamin B. Lahey, Rolf Loeber, Mary Anne G. Christ, Elizabeth
L. Hart, and Lynne E. Tannenbaum (Journal of Consulting and Clinical Psychology, 1991, Vol.
59, No. 2, 289-294), contained an error.
On page 290, in the formula to calculate the discrepancy score between IQ and academic
achievement, Step 3 should appear as follows:
3. Differences between achievement and full-scale intelligence were expressed in z-score
units adjusting for the standard error of estimate. For example, reading discrepancy = (ZREAD