Journal of Consulting and Clinical Psychology Copyright 1990 by (he American Psychological Association, Inc.

1990, Vol. 58, No. 6, 768-774 0022-006X/90/$00.75

Patient and Therapist Introject, Interpersonal Process,

and Differential Psychotherapy Outcome
William P. Henry Thomas E. Schacht
Vanderbilt University James H. Quillen College of Medicine
Hans H. Strupp
Vanderbilt University

The Structural Analysis of Social Behavior (SASB; Benjamin, 1974,1982,1984) system was used to
study the interpersonal process between patient and therapist in the 3rd session of 14 therapeutic
dyads. Dyads were grouped into good and poor outcomes cases (n = 7) on the basis of the amount of
change in the patients' introject as measured by the INTREX Introject Questionnaire (Benjamin,
1983). Strong support was found for the following hypotheses based on interpersonal theory, link-
ing therapists' introject state, interpersonal process in therapy, and outcome: (a) Poor outcome
cases (no introject change) were typified by interpersonal behaviors by the therapist that con-
firmed a negative patient introject; (b) the number of therapists' statements that were subtly hostile
and controlling was highly correlated with the number of self-blaming statements by the patients; (c)
therapists with disaffiliative introjects tended to engage in a much higher level of problematic
interpersonal processes that have been associated with poor outcome. Implications for future
research and therapist training are discussed.

Stripped of technical jargon, psychotherapy may be seen in
one context as simply a structured relationship between two
people. Despite the multiplicity of theoretical approaches to
the practice of psychotherapy, most major camps recognize the
crucial importance of the quality of the therapeutic relation-
ship (Gurman, 1977), and a number of studies have associated
an afnliative therapeutic relationship with improved clinical
outcomes (Gomes-Schwartz, 1978; O'Malley, Suh, & Strupp,
1983; Sachs, 1983). This consensus, however, belies disputes
regarding the exact roleof the therapeutic relationship in achiev-
ing positive outcomes in therapy.
This issue has received little empirical clarification, in part
because of measurement difficulties that arise when one at-
tempts to operationalize the therapeutic relationship. Global
scales can rate the overall relationship "climate" with items
such as therapist "warmth," "judgmentalness," and so forth.
Although such global scales have provided useful descriptive
accounts that serve in a general sense to distinguish good from
poor relationships, they do not explicate exactly what the thera-
pist does to establish the quality of the relationship, and often
they are not designed to articulate with specific theories of
change. Global scales may thus encourage the view that the
relationship is merely a generic or nonspecific background fac-
tor forming a context in which specific techniques are applied.
The present study was designed to further a more precise un-
derstanding of the role of the therapeutic relationship by test-
This work was supported in part by National Institute of Mental
Health Grant MH-20369, Hans H. Strupp, Principal Investigator.
Correspondence concerning this article should be addressed to Wil-
liam P. Henry, Department of Psychology, Center for Psychotherapy
Research, A&S Psychology Building, Vanderbilt University, Nashville,
Tennessee 37240.
ing theoretically derived propositions linking therapist inter-
personal actions and therapeutic change.
Henry, Schacht, & Strupp (1986) demonstrated a precise
method for operational!zing the therapeutic relationship that
allowed the therapist's relationship contribution to be seen as a
specific technique in itself. Moment-by-moment transactions
between therapist and patient were measured with Structural
Analysis of Social Behavior (SASB; Benjamin, 1974,1982,1984)
and compared across good versus poor outcome cases seen by
the same therapists. Disaffiliative and complex (mixed-mes-
sage) interpersonal transactions occurring between patient and
therapist were found to be strongly related to subsequent clini-
cal outcome. In terms of the underlying dimensions of SASB,
therapists in poor outcome cases were exercising more hostile
control and were less prone to grant friendly autonomy, while
their patients were responding with a reduced level of affiliative
autonomy-taking and greater hostile separation.
This study (Henry, Schacht, & Strupp, 1986) demonstrated
the potential importance of dyadic interpersonal process to
therapeutic change, but because of limitations imposed by
preexisting data, Henry etal. could not evaluate specific hypoth-
eses about change mechanisms. To advance this line of research
it is important to refine hypotheses about how interpersonal
transactions may produce outcome changes. One important
hypothesis derives from the theory of interpersonal introjec-
tion.' The most basic principle of introject theory is that people
learn to treat themselves as they have been treated by others
(Sullivan, 1953). The dynamic interaction of these early inter-
' The term inlrojection has multiple definitions related to various
psychodynamic and developmental theories (see Greenberg & Mit-
chell, 1983). Our article relies most directly on the theories of Sullivan
(1953) and Benjamin (1984).

768
 INTERPERSONAL PROCESS
nalized interpersonal influences forms a hypothesized person-
ality structure termed the "introject," which comprises a rela-
tively stable conscious and unconscious repertoire of ways of
treating the self. This repertoire, which internally mirrors treat-
ment received at the hands of early significant others, includes
self-appraisals, verbal and motor behaviors directed at the self,
and cultivation of various images of the self. Although early
experience is thought to be critical, the process of introject
formation and the content of the introject are presumably sub-
ject to development and change across the lifespan. Individuals
who are psychologically healthy, by definition, evidence intro-
jects that are relatively friendly (self-accepting, self-nurturing,
self-helping) most of the time. Emotionally disturbed persons,
in contrast, tend to have hostile introjects that are recurrently
self-critical, self-destructive, or self-neglectful.
Although the descriptive language may vary, a range of theo-
rists from psychodynamic to cognitive-behavioral hypothesize
that introjective actions or states are major correlates of affec-
tive experience and are important in shaping interpersonal be-
haviors. For example, self-criticism is associated with shame
and sadness, and self-neglect may be associated with boredom.
The introject is also hypothesized to play a central role in recur-
rent maladaptive relationship patterns that underlie symptom-
atic presentations (Schacht, Binder, & Strupp, 1984; Schacht &
Henry, in press). Within these recurrent patterns, introject
states are hypothesized to become self-perpetuating through
the dynamics of interpersonal complementarity (Benjamin,
1982). For example, a patient's inner self-criticism may lead to
defensiveness in interpersonal transactions, implicitly casting
the therapist in the role of critic (the accustomed action of
important others that originally led to the introject formation).
To the extent that the therapist defends himself or herself
against this role attribution, such as by disagreeing with the
patient, the therapist self-defeatingly becomes a critic and "con-
firms" the patient's negative expectation (perpetuating the in-
troject state).
Because of its theoretically central role in maintaining prob-
lematic affective/interpersonal cycles, a measure of the intro-
ject becomes a logical element when evaluating patients' initial
status and therapeutic change. As described in the example
above, introject theory would predict that the therapist's inter-
personal behavior would serve to either perpetuate problematic
introject states (leading to poor outcome) or ameliorate them
through experiential disconfirmation. If outcome is defined in
terms of introject change, then outcome should vary as a func-
tion of the therapists' interpersonal behavior. Introject theory
also predicts that therapists will tend to treat patients in accord
with their own introjects. This will occur by the same comple-
mentary dynamics that tend to make introjects self-perpetuat-
ing for patients. Thus, therapists whose introjects are self-ac-
cepting would be expected to engage their patients in accepting
and supportively helpful transactions as compared with thera-
pists with hostile introjects, who would be expected to behave
in a relatively more critical or neglectful manner with their
patients.
The foregoing theoretical model offers conceptual links be-
tween (a) a problem definition (negative introject); (b) a treat-
ment process (differential interpersonal process as mediated by
the variable of therapist introject); and (c) an outcome measure
769
(introject change). Our study capitalizes on these conceptual
links by using a measurement system (SASB) that allows prob-
lems, treatments, and outcomes to be represented in a shared
metric that articulates with the underlying theory (cf. Principle
of Problem-Treatment-Outcome [P-T-O] Congruence-
Schacht, Strupp, & Henry, 1988). The aim of the study is to test
process-outcome relationships that would be consistent with
the theoretically proposed introject mechanism. Two hypothe-
ses were examined:
1. For patients who begin therapy with disaffiliative intro-
jects, higher levels of therapist disaffiliativeness will be asso-
ciated with lower levels of introject change or deterioration.
2. Therapists with disaffiliative introjects will treat their pa-
tients in a relatively more critical and neglectful manner.
Method
Subjects
Fourteen therapeutic dyads were drawn from the first cohort of the
Vanderbilt II psychotherapy research project, a 5-year study of Time-
Limited Dynamic Psychotherapy (Strupp & Binder, 1984).3 Dyad se-
lection was structured to produce two equal groups (n = 7) of high-
change and low-change cases on the basis of changes in the SASB
introject. Participating patients were comparable to a general psychiat-
ric outpatient population, as indicated by intake scores on the Global
Severity Index of the SCL-90-R (Derogatis, 1977). The GSI group
mean was 48.9, and the two outcome groups were statistically equiva-
lent on the GSI measure at intake. All patients were judged to have
significant interpersonal difficulties and qualified for a Diagnostic
and Statistical Manual of Mental Disorders, 3rd ed. (DSM-I1I; Ameri-
can Psychiatric Association, 1980) Axis I or Axis II diagnosis, as deter-
mined by a clinician using a computerized rating version of the Na-
tional Instituteof Mental Health (NIMH) Diagnostic Interview Sched-
ule. The mean age of patients was 41.04 (range 24-64 years), 77% were
female, 79% had completed at least some college work, 69% had re-
ceived previous therapy, and all patients and therapists were White.
Patients were assigned on a random basis to participating therapists,
who were self-described psychodynamically oriented psychiatrists and
clinical psychologists in private practice with a minimum of 2 years'
full-time postresidency or postinternship experience. Therapies con-
sisted of weekly 50-min meetings for a maximum of 25 sessions and
were conducted with no intrusions by the research staff, except to
collect written measures of process and outcome.
SASB Measures
The introject is operationalized as the third of the three interrelated
circumplex surfaces of the SASB model. Each circumplex surface pres-
ents, ina two-dimensional space, 36 interpersonal behaviors that repre-
sent unique combinations of the theoretically "primitive" vectors of
affiliation-disaffiliation and independence-interdependence. Inter-
personally complementary behaviors are represented as homologous
points across the surfaces. Each SASB surface defines a particular per-
spective or focus of interpersonal transactions. Surface 1 involves focus
on another person (transitive action). Surface 2 involves focus on the
self (intransitive states). Surface 3 represents intrapsychic actions that
result when focus on the other (Surface 1) is directed inward on the self
2
Fourteen was the maximum number of dyads with clearcut good
and poor outcomes that were available for analysis at the time the study
was conducted.
770 W HENRY, T. SCHACHT, AND H. STRUPP

(introjection). The reader is directed to Benjamin (1974, 1982) for a
more complete description of the SASB system. Figure t presents an
alternate version of theSASB model in which the 36 behaviors for each
circumplex are collapsed into eight clusters. Data for the present study
are presented in terms of these clusters.
Initial status and outcome measurement. The SASB system affords
two methods for measuring the introject. The first method involves
direct ratings of observed introjective behaviors. For example, a state-
ment such as "I'm an awful person for thinking that" would be coded as
a self-critical introject (Surface 3, Cluster 6, or 3-6). Direct codings of
transactions were employed in the present study as measures of treat-
ment process. The second measurement method, used in the present
study as the measure of initial state and outcome, uses the INTREX
Introject Questionnaire (Benjamin, 1983). This well-validated instru-
ment comprises 36 items, each corresponding to one of the 36 introjec-
tive behaviors that compose the introject circumplex of the SASB
model. Sample items are given in Table 1. Self-ratings are made on a
scale of 0-100 according to the degree to which the rater judges an item
to be descriptive of his or her behavior. Each item is rated twice, once
with an instruction to "describe yourself at your best" and once with
an instruction to "describe yourself at your worst." The result is two
measures, "introject at best" and"introject at worst." Multiple adminis-
trations of the INTREX Questionnaire allow changes in the patient's
introject at best and worst to be graphed over time.
INTREX introject ratings are scored by a computer program (Benja-
min, 1983) that weights each rating in relation to its position on the
basic circumplex dimensions of control versus freedom and attack ver-
sus affiliation. The result is a measure of central tendency that summa-
rizes the overall introjective stance of the rater. This measure of central
tendency is expressed in terms of two coefficients, corresponding to
the circumplex axes, that define a point on the circumplex surface. The
first coefficient, termed the attack coefficient, expresses the central

(1) Freeing and Forgetting

FOCUS ON OTHER |
(8) Ignoring and (2) Affirming and
Neglecting Understanding

(7) Attacking and (3) Nurturing and

Rejecting Comforting

(6) Belittling and (4) Helping and

Blaming Protecting

(5) Watching and Managing

(1) Asserting and Separating

[ FOCUS ON SELF |
(8) Walling off (2) Disclosing and
Avoiding Expressing

(7) Protesting and (3) Approaching and

Recoiling Enioving

(6) Sulking and (4) Self-Protecting

and Enhancing

(5) Deferring and Submitting

(DSpontaneous Self

(8) Daydreaming and (2) Self-Accepting

NefflectiiwSelf and Exploring

(7) Self-Rejecting (3) Self-Nourishing

(6) Self-Indicting (4) Self-Protecting

and Enhancing

(5) Self-Moniloring and Restraining

Figure 1. The SASB Circumplex System.
 IN'IT.RI'r.RSONAl. PROCESS
Table 1
Sample INTREX Introject Items by Cluster
Cluster Item
1 Without concern I just let myself be free to turn into
whatever I will.
2 Knowing both my faults and strong points I
comfortably let myself be "as is."
3 I like myself very much and feel very good when I have
a chance to be with myself.
4 I practice and work on developing worthwhile skills,
ways of being.
5 I have a habit of keeping very tight control over myself.
6 I accuse and blame myself until I feel guilty, bad, and
ashamed.
7 I harshly punish myself, take it out on myself.
8 Instead of getting around to doing what I really need to
do for myself, I let myself go and just daydream.
tendency of ratings on the basic dimension of self-affiliation versus
self-attack (self-love vs. self-hate). The second coefficient, termed the
control coefficient, expresses the central tendency of the ratings on the
basic dimension of self-freeing versus self-control. Values for these
coefficients are expressed in a range from -1,000 to +1,000.
Patients were assigned to high-change (good) and low-change (poor)
outcome groups according to the amount and direction of change in
their self-reported introject at worst as measured by the INTREX Intro-
ject Questionnaire administered at intake; Sessions 3,8,16; and termi-
nation.3 All patients in the current study began therapy with their
introject at worst in either Quadrant 2 or 3 of the introject circumplex
(Figure 1, Surface 3). This reflects a hostile and critical or neglectful
stance toward the self. There was no statistically significant difference
between the outcome groups at intake in the amount of self-hostility
(as represented by the attack pattern coefficient) at best or worst. Ther-
apists also completed INTREX Introject Questionnaires rated at best
and worst on themselves prior to meeting their patients.
Indications of positive change were denned as (a) clinically signifi-
cant movement from left to right on the circumplex, indicating greater
self-affiliation (approximately 500 points on the -1,000 to +1,000
axes);4 (b) movement from Quadrant 2 to 3. (Quadrant 2 is considered
the most pathological state, as it represents both self-hatred and self-
freeing (neglect], which is hypothesized to be a potentially suicidal
introject profile.) Thus movement to Quadrant 3, reflecting greater
self-control and self-management, is deemed positive therapeutic
change.
Indications of low change were defined as (a) movement from Quad-
rant 3 to 2 (away from self-control), which represents a change toward
greater self-neglect; (b) significant movement from right to left indicat-
ing increased self-hostility; (c) no change, including no improvement in
introject at best or worst.
To explore the validity of this procedure, the two groups formed on
the basis of introject change were compared on more traditional out-
come measures. Pre-post residual change scores (partialing out vari-
ance in change scores due to initial status) were calculated on the
Global Assessment Scale (GAS) as rated by an independent clinician
and the SCL-90-R Global Severity Index. In addition, the Global
Outcome Ratings made separately by the patient and therapist after
therapy (-5 to +5 direct change rating) were examined. Independent
groups / tests were all significant (p < .05) with differences in the
expected direction. Thus measures of level of functioning, symptom
status, and perceived change from three sources were consistent in
supporting the division of the patient sample into positive and negative
outcome on the basis of introject change.
771
Ihuinicnt /.xv.v\ <-<>iiint>. Transcripts of third session videotapes
we re prepared, uniti/.ed. and SASB process coded following the proce-
dure outlined in Henry etal.. (1986). Two clinical psychologists, highly
experienced in SASB coding, and blind to the pretherapy data and
outcome status of each dyad, coded the first 200 thought units (approxi-
mately 30 min) of each session using transcripts and videotape. A
thought unit is any portion of speech (usually a phrase or at most a
sentence) judged to express one complete thought, and transcripts were
unitized by consensus of the two judges. Strict independent reliability
of the unitizing procedure was not deemed important as the exact
boundaries of units do not affect the process coding frequencies them-
selves (the dependent variable). In-a reliability check the raters indepen-
dently coded 150 thought units from a session not included in the
analysis. A high interjudge reliability was established for SASB cluster
assignment (Cohen's kappa = .91). Instances of disagreement in the
actual codings were resolved through discussion to arrive at one set of
consensus codes.
Results
Interpersonal Process and Differential Outcome
Before statistical analysis, the observed frequencies in each of
the 16 possible SASB focus/cluster combinations were tallied
for both patient and therapist in each case.5 These frequencies
were then weighted to the respective patient or therapist grand
mean for number of codings to remove the artifact of differing
proportions of turns at talk between the participants in differ-
ent dyads. Such weighted frequencies, the basic data studied,
allow for meaningful cross-case comparison.
Initially, two 2 x 2 x 8 (Outcome X Focus x Cluster) ANOVAS
with one between and two within factors were performed (one
analysis for patient communications and one for therapist com-
munications), using weighted cluster frequencies as the depen-
dent variable. Simple effects tests for differences in cluster
means between the two outcome groups were accomplished
using between-groups t tests. In several instances the Mann-
Whitney U statistic was substituted as a result of significant
differences in within-group variance.6
As expected, therapists spent more time focusing on other
(F = 354; p< .001), whereas patients spent more time focusing
on self (F= 2642; p < .001), and the eight SASB clusters were
utilized differently depending on outcome group (F = 4.8; p <
.001). Table 2 contains the results of the cluster comparison
tests. Patients in the poor outcome group were significantly
more watching and managing toward the therapists, more as-
3
Introject-at-best ratings were not the variable of primary interest in
our analysis because they tended to be idealized, resulting in a re-
stricted range of ratings.
4
This figure is considered conservative and was arrived at in consul-
tation with L. S. Benjamin, who has extensive clinical experience with
the INTREX Questionnaires.
5
Each part of a complex communication was counted as one occur-
rence of the appropriate cluster. For instance, a 2-2/2-6 was counted as
one 2-2 and one 2-6.
6
After we consulted several statisticians, adjustment for familywise
error rate was deemed unnecessary in this case because of the rela-
tively small number of simple effects tests performed. At the .05 level
of significance, no results (less than one) would be expected by chance
alone for each ANOVA.
 772 W HENRY, T. SCHACHT, AND H. STRUPP

Table 2
Comparison of Cluster Mean Weighted Frequencies and Complex Communications
Therapist Patient

Cluster Good outcome Poor outcome Good outcome Poor outcome

Focus on other
1 Freeing/forgetting 0 0 0 0
2 Affirming/understanding 30.30 36.36 0.61 0.30
3 Nurturing/comforting 1.34 0 0 0
4 Helping/protecting 22.49 20.21 1.14 0.51
5 Watching/monitoring 6.50 5.27 0.14 3.21
6 Belittling/blaming 0 2.93* 0 1.51***
7 Attacking/rejecting 0 0 0 0
8 Ignoring/neglecting 0.47 4.96* 0 0.14

Focus on self
1 Asserting/separating 0 0.47 2.99 15.46**
2 Disclosing/expessing 1.2 1.83 130.86 101.03**
3 Approaching/enjoying 1.57 0.24 4.26 0.70
4 Trusting/relying 0.96 0.56 12.10 16.93
5 Deferring/submitting 0.56 0.67 2.27 4.29
6 Sulking/appeasing 0 1.8 0.61 6.43*
7 Protesting/recoiling 0 0 0 0
8 Walling off/avoiding 0 0.16 1.80 6.56*
Complex communications
Mixed affiliation .57(1%) 10.43(18%)* 2.0(1%) 11.86(9%)*
Mixed focus .42(1%) 4.43 (7%)* .29 (0%) 5.0 (4%)**
*/?<.05. ***p<.001.

setting and separating, sulking and appeasing, and walling off
and avoiding, while being less disclosing and expressing. Thera-
pists in the poor outcome group were significantly more belit-
tling and blaming and ignoring and neglecting. Thus, replicat-
ing our previous study, patients and therapists in the low-
change group demonstrated significantly higher levels of
disaffiliative behavior. Additional analyses were performed
yielding the following results:
1. A strong correlation was observed between the number of
therapist statements that were hostile and controlling (Focus-
Other, Quadrant 3) and the number of their patient's statements
that were self-blaming and critical (Introject Quadrant 3), r =
0.53 (p< 0.05).
2. Both patients and therapists in the low introject change
group were found to exhibit significantly higher levels of com-
plex communications (see Table 2). These included primarily
mixed affiliation statements (simultaneously affiliative and
hostile) and mixed focus statements (focus on self and other
simultaneously). The relative percentage of mixed message
thought units for good versus poor outcome cases was 3% versus
13% for the patients (p < .04) and 6% versus 26% for the thera-
pists (p< .01).
3. The percentage of thought units receiving at least one dis-
affiliative code was calculated for the patient and the therapist,
yielding an index of hostile communication. Both patients
(16% vs. 2%) and therapists (21% vs. 3%) in the poor outcome
group acted with significantly greater hostility (p < .04).
4. Relative to their therapists, patients in the poor outcome
group spoke less than patients in the good outcome group (69%
vs. 79%, p < .05, two-tailed).
Therapist Introject and Differential Interpersonal Process
As predicted, therapists whose pretherapy self-ratings indi-
cated relatively greater hostility toward themselves were also
more likely to be coded as treating their patients in a disaffilia-
tive manner. Therapists whose introjects rated at worst were
friendly (Quadrants 1 and 4) were given hostile codes in 5.6% of
the coded transcript thought units. In contrast, therapists who
reported hostile introjects at worst (Quadrants 2 and 3) were
given hostile codes 17.7% of the time (Mann-Whitney U,
p < .05).
Discussion
Coupled with earlier findings (Henry et al, 1986), this study
provides strong evidence for the importance of interpersonal
transactions early in therapy. It should be clearly noted that the
present design does not offer proof of the hypothesized under-
lying mechanisms, but it does offer a range of data consistent
with predictions based on interpersonal theory. There was a
significant relationship between therapists' interpersonal be-
haviors and the ways in which the patients acted toward them-
selves. Patients whose introjects showed marked improvements
 INTERPERSONAL PROCESS
engaged in therapeutic interactions that were almost com-
pletely devoid of disaffiliative therapist behaviors, whereas ther-
apists of those patients showing no introject change (or deterio-
ration) engaged in hostile blaming, ignoring, and separating
sequences.7 This relationship between interpersonal process
and differential intrapsychic changes suggests one important
specific role of the therapeutic relationship in promoting posi-
tive outcome.
Related predictions were also supported. Carson (1969), inte-
grating the theories of Sullivan and others (Thibaut & Kelly,
1959; Secord & Backman, 1961), stated that one's interpersonal
behavior is unconsciously designed to evoke responses from
others that validate a stable self-image (even a negative self-con-
cept, see Anchin, 1982). The results indicate that there was a
high degree of correspondence between patient self-blaming
statements and therapist statements subtly blaming the patient
(r = .53).* Finally, in what is considered a most significant find-
ing, the therapists' own introjects were found to be systemati-
cally related to their in-session process with patients.' These
results may have important theoretical, methodological, and
training implications.
From a methodological standpoint, the study demonstrates
the utility of defining problems, therapeutic process, and out-
come in similar theoretically linked metrics. In a broader sense,
it demonstrates ties between specific therapist behaviors and
clinical change, a line of research with an unfortunate history
of failure (Gurman, 1977; Orlinsky & Howard, 1986). Although
frequency counts were used as the unit of analysis, these units
represented momentary dyadic states, not aggregation of unilat-
eral acts. This is consistent with an increasing call for context-
sensitive approaches to process research (Greenberg, 1986).
More generally, it appears that interpersonal process analysis
with SASB may provide an important contextual backdrop
when combined with other descriptive process measures. For
instance, it would be valuable to study the effect on interper-
sonal process of various types of "technical" interventions,
such as transference interpretations, with different types of pa-
tients during different phases of therapy. There has been recent
suggestive evidence that such interpretations often foster resis-
tance and poor process (Foote, McCullough, Winston, Pollack,
& Laikin, 1988; Marmar, Weiss, & Gaston, 1988). Perhaps it is
not the nature of the interventions that is responsible for these
findings, but the interpersonal context in which they are often
carried out. Wile (1984), writing on the accusatory nature of
many common interpretations, states:
Certain interpretations commonly made in psychoanalysis and
psychodynamic therapy are accusatory. Therapists appear to
make them not because they are hostile or insensitive, but be-
cause of the dictates of their theory. Clients are seen as gratifying
infantile impulses, being defensive, having developmental defects,
and resisting therapy.. . . Even therapists who reject the psy-
choanalytic model may nevertheless view clients as dependent,
narcissistic, manipulative, and resistant, and as refusing to grow
up and face the responsibilitiesof adulthood. Therapists who con-
ceptualize people in these ways may have a hard time making
interpretations that do not communicate at least some element of
this pejorative view. (p. 353). . . . The major difficulty with accu-
satory interpretations, in conclusion, is that clients are already
accusing themselves, often for the very thing for which the tradi-
773
tional therapist is accusing them, and that thisself-criticismis
the problem, (p. 363)
The high correlation observed in the present study between
therapist-blaming and patient self-blaming would seem to bear
out Wile's point.
Our data suggest an important overriding principle:
Whereas the absence of a negative interpersonal process may
not be sufficient for therapeutic change, the presence of even
relatively low levels of negative therapist behavior may be suffi-
cient to prevent change. Why should this be so? One explana-
tion rests on a cognitive understanding of the introject. Intro-
jects may be construed as overlearned beliefs or stereotypes
about the self. By definition, a stereotype is a belief that is held
rigidly and is not easily opened to question ordisconfirmation.
Very little apparently affirmative "interpersonal evidence" may
be required to prove a stereotype in the mind of the believer. To
promote change, the theory predicts that therapists' behavior
should rather consistently disconfirm patients' negative self-ex-
pectations.10 Similar therapeutic prescriptions appear in Beier
and ^bung's (1984) concept of "asocial response" and in Weiss
and Sampson's (1986) assertion that therapists must repeatedly
pass patients' interpersonal "tests."
Conclusion
Apparently, even well-trained professional therapists are sur-
prisingly vulnerable to engaging in potentially destructive in-
terpersonal process (in the most extreme case observed, the
therapists' communications were judged to be disaffiliative
56% of the time, with 65% of the thought units found to contain
complex messages). Perhaps this vulnerability should not be so
surprising when one considers the strong pull of multiple po-
tential determinants of such process. Patients' preexisting mal-
adaptive "evoking styles" (Kiesler, 1982) often tend to elicit
poor process according to the rules of interpersonal comple-
mentarity. The therapist's own introjective state may also con-
tribute to problematic process. In addition, as Wile noted, theo-
retical heuristics may themselves inadvertently encourage dis-
affiliative interpersonal transactions. It seems clear that
7
It is interesting to note that when our two studies were combined,
in 11 good outcome cases, not a single instance of therapist blaming
and belittling (Cluster 6) was observed.
* It should be noted that this relationship is not a given, as parents in
the good outcome group also made self-blaming statements (although
at a lower frequency), but there were no instances in which these state-
ments were complemented by therapist blaming behaviors.
9
This finding cannot be taken to mean that all therapists with "dis-
affiliative introjects" will display hostility toward all patients. It does
suggest that some therapists, as a function of their own interpersonal
history, may be more vulnerable to engaging in countertherapeutic
interpersonal processes with patients who evoke such negative tenden-
cies.
10
It should be noted that to perfectly disconfirm patients' expecta-
tions is quite likely an impossible task. Additional research should
address the question of how therapists who find themselves in a disaf-
filiative transaction can best repair the situation. We hypothesize that
therapeutic metacommunication plays a crucial role in this task.
 774 W HENRY, T. SCHACHT, AND H. STRUPP
traditional training methods have not adequately prepared
many therapists to expertly perceive and respond to the poten-
tial interpersonal process meanings and effects of their inter-
ventions. Future training programs, regardless of theoretical
orientation, may be designed to directly address these funda-
mental interpersonal skills.
Finally, our results potentially point to an underlying change
process operative across therapies with different surface struc-
tures. In speaking of the fundamentally human and interper-
sonal nature of psychotherapy, Safran and Segal (1990) caution,
"relevant theory must clarify the process through which this
human encounter brings about change." We suggest that the
role of the therapeutic relationship in fostering therapeutic
change through its effect on the patient's introject structure
constitutes a common change mechanism of central impor-
tance. When the relationship is seen in this light, the distinc-
tion between so-called nonspecific and specific change factors
disappears.
References
American Psychiatric Association. (1980). Diagnostic and statistical
manual of mental disorders (3rd ed.). Washington, DC: Author.
Anchin, J. C. (1982). Sequence, pattern, and style: Integration and
treatment implications of some interpersonal concepts. In J. C. An-
chin & D. J. Kiesler (Eds.), Handbook of interpersonal psychotherapy
(pp. 95-131). New York: Pergamon Press.
Beier, E., & \bung, D. (1984). The silent language of psychotherapy:
Social reinforcement of unconscious processes. Hawthorne, NY: Al-
dine.
Benjamin, L. S. (1974). Structural analysis of social behavior. Psycho-
logical Review, 81, 392-425.
Benjamin, L. S. (1982). Use of Structural Analysis of Social Behavior
(SASB) to guide intervention in psychotherapy. In J. C. Anchin &
D. J. Kiesler (Eds.), Handbook of interpersonal psychotherapy (pp.
190-212). New York: Pergamon Press.
Benjamin, L. S. (1983). The INTREXuser's manual, Part I and II [Com-
puter program manual]. Madison, WI: Intrex Interpersonal Insti-
tute, Inc., P.O. Box 55218, Madison, WI 53705.
Benjamin, L. S. (1984). Principles of prediction using Structural Analy-
sis of Social Behavior. In R. A. Zucker, J. Aronoff, & A. J. Rabin
(Eds.), Personality and the prediction of behavior (pp. 121-173). New
\brk: Academic Press.
Carson, R. C. (1969). Interaction concepts of personality. Chicago: Al-
dine.
Derogatis, L. R. (1977). SCL-90 administration, scoring, and proce-
dures manuals for the revised version. Baltimore: Leonard R. Dero-
gatis.
Foote, J., McCullough, L., Winston, A., Pollack, J., & Laikin, M. (1988,
June). The emergence of patient affective responding as a result of
specific therapeutic interventions. Paper presented at the annual con-
vention of the Society for Psychotherapy Research, Sante Fe, NM.
Gomes-Schwartz, B. (1978). Effective ingredients in psychotherapy:
Predictions of outcome from process variables. Journal of Consult-
ing and Clinical Psychology, 46,1023-1035.
Greenberg, J. R., & Mitchell, S. A. (1983). Object relations in psychoan-
alytic theory. Cambridge, MA: Harvard University Press.
Greenberg, L. S. (1986). Change process research. Journal of Consult-
ing and Clinical Psychology, 54, 4-9.
Gurman, A. S. (1977). The patient's perception of the therapeutic rela-
tionship. In A. S. Gurman & A. M. Razin (Eds.), Effective psychother-
apy: A handbook of research (pp. 503-543). New York: Pergamon
Press.
Henry, W P., Schacht, T. E., & Strupp, H. H. (1986). Structural Analysis
of Social Behavior: Application to a study of interpersonal process
in differential psychotherapeutic outcome. Journal of Consulting
and Clinical Psychology, 54, 27-31.
Kiesler, D. J. (1982). Interpersonal theory for personality and psycho-
therapy. In J. C. Anchin & D. J. Kiesler (Eds.), Handbook of interper-
sonal psychotherapy (pp. 3-24). New \fork: Pergamon Press.
Marmar, C. R., Weiss, D. S., & Gaston, L. (1988, June). Towards the
validation of the Therapeutic Alliance Rating System. Paper pre-
sented at the annual convention of the Society for Psychotherapy
Research, Sante Fe, NM.
O'Malley, S., Suh, C. S., & Strupp, H. H. (1983). The Vanderbilt Psycho-
therapy Process Scale: A report on the scale development and a
process-outcome study. Journal of Consulting and Clinical Psychol-
ogy. 51, 581-586.
Orlinsky, D. E., & Howard, K. I. (1986). Process and outcome in psycho-
therapy. In S. L. Garfield & A. E. Bergin (Eds.), Handbook of psycho-
therapyandbehaviorchange(3rd ed., pp. 311-384). New York: Wiley.
Sachs, J. S. (1983). Negative factors in brief psychotherapy: An empiri-
cal assessment. Journal of Consulting and Clinical Psychology, 51,
557-564.
Safran, D. S., & Segal, Z. V (1990). Interpersonal process in cognitive
therapy. New York: Basic Books.
Schacht, T. E., Binder, J. L., & Strupp, H. H. (1984). The dynamic focus.
In H. H. Strupp & J. L. Binder (Eds.), Psychotherapy in a new key (pp.
65-109). New York: Basic Books.
Schacht, T. E., & Henry, W P. (in press). Modeling recurrent relation-
ship patterns with Structural Analysis of Social Behavior: The
SASB-CMP. In M. N. Docherty & L. Luborsky (Eds.), Handbook of
psychodynamic treatment research. New York: Basic Books.
Schacht, T. E., Strupp, H. H., & Henry, W P. (1988). Problem-treatment-
outcome congruence: A principle whose time has come. In H. Dahl
& H. Kaechele (Eds.), Psychoanalytic process research strategies (pp.
1-14). New York: Springer.
Secord, P. F, & Backman, C. W (1961). Personality theory and the
problem of stability and change in individual behavior: An interper-
sonal approach. Psychological Review, 68, 21 -32.
Strupp, H. H., & Binder, J. L. (1984). Psychotherapy in a new key. New
York: Basic Books.
Sullivan, H. S. (1953). The interpersonal theory of psychiatry. New "York:
Norton.
Thibaut, J. W, & Kelley, H. H. (1959). The social psychology of groups.
New York: Wiley.
Weiss, J., & Sampson, H. (1986). The psychoanalytic process: Theory,
clinical observation, and empirical research. New York: Guilford
Press.
Wile, D. B. (1984). Kohut, Kernberg, and accusatory interpretations.
Psychotherapy, 21,353-364.
Received August 17,1989
Revision received April 2,1990
Accepted April 23,1990 .