MSU

LIBRARIES
.-.__—

 

 

 

\\\\\\\\\ \IIIIIEIINIQLIII

3 1293 1

RETURNING MATERIALS:
PIace in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped be10w.

 

 

 

.
‘

f". .' 4-.
j r. r 4 5‘ ‘1?) a“, >
K ; -_, _

-. m- A on
”Qt-P 1 $2 ‘ .“ 'H 1;
V“. U f.‘ v. i_-‘i

 

 

CLIENT-PERCEIVED
THERAPIST POSITIVE REGARD AND
PSYCHOTHERAPY OUTCOME

BY

Paul Andrew Eckert

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Psychology

1986

Copyright by
PAUL ANDREW ECKERT
1986

ABSTRACT

CLIENT-PERCEIVED
THERAPIST POSITIVE REGARD AND
PSYCHOTHERAPY OUTCOME

BY

Paul Andrew Eckert

The purpose of the present study was to determine how
the relationship between psychotherapy process (Rogerian
"positive regard”) and outcome is affected by the type of
outcome measure employed. Process was measured by subjects'
self-report, post-therapy retrospective ratings.

Five psychotherapy outcome measures were employed.

Rated gain is the client's global, post-therapy retrospective
assessment of amount of symptom change. Raw gain is the
difference between pre-therapy and post-therapy average
symptom checklist ratings. Residual gain is a score computed
by statistically correcting raw gain scores for the influence
of pre-therapy symptom severity. The aitetnate gtenping
methed controls for the influence of pre-therapy symptom
severity by trichotomizing the subject sample into mild,
moderate, and severe symptom subgroups. Attained health is a
dichotomous variable which is positive only if pre-post
symptom change is statistically reliable and post-therapy
symptom severity is comparable to that of a nonpatient
normative group.

Subjects were 77 clients of an outpatient psychological

clinic.

Prior research suggested that, when post-therapy self—
report process measures were used with self-report outcome
measures, process-outcome relationship strength depended in
part on the extent to which each outcome measure reflected
the subject's perceived post-therapy symptomatic status.
Thus it was hypothesized that outcome measures would be
ordered from strongest to weakest relationship with process
measures, as follows: rated gain, attained health, residual
gain, and raw gain. The alternate grouping method was
expected to produce results congruent with those achieved
using residual gain. Also, most process-outcome
relationships were expected to be significant and positive.

These hypotheses were only partially confirmed. Only
rated gain was significantly related to positive regard, but
the relationship was in the predicted positive direction.
Rated gain was more strongly related than raw gain to
positive regard, as expected. However, controlling for the
influence of pre-therapy symptom severity on raw gain scores
(by use of either residual gain or the alternate grouping
method) did not render the process-outcome relationship
stronger than it was in the case of raw gain. And attained
health, rather than being mete strongly related than residual

gain to positive regard, was actually less Strongly related.

This work is dedicated to the late Dr. Terrence Allen, who
introduced me to psychometric theory in a most fascinating
manner. I will never forget his genuine personal interest in
my education and general well-being. My only regret is that
I have now lost the opportunity to thank him personally. One

should not wait to say such things.

ACKNOWLEDGMENTS

I wish to thank Dr. Norman Abeles for his generous
contribution of time and energy to this project. Without his
willingness to work with me to meet a demanding timetable, it
would have been impossible to achieve success. I also wish
to recognize the contributions of my dissertation committee:
Dr. Albert Aniskiewicz, Dr. Bertram Karon, and Dr. Raymond

Frankman.

vi

Outcome Measures .

TABLE OF CONTENTS

LI ST OF TABLES O O O O O O O O O I O O O 0

INTRODUCTION 0 O O O O O O O O O O I O O 0

Research Concerning the Rogerian Hypothesis . .

Statement and Initial Confirmation . . . .
Subsequent Criticism, Debate, and Revision

Recent Developments . . . . . .
sumary O O O O O O O O O O O O

The Purpose and Scope of the Present Study

Preliminary Methodological Considerations

The Implications of Client Bias
Rated, Raw, and Residual Gain
The Alternate Grouping Method
Attained Health . . . . . . .
Summary . . . . . . . . . . .

Process-Outcome Research Evidence . .

The Literature Review Process .
The Relevance of Prior Research
Rated Gain . . . . . . . . . . .
Raw and Residual Gain . . . . .
The Alternate Grouping Method .
Attained Health . . . . . . . .
Interrelationships Among Outcome
Interrelationships Among Process

HYPOTHESES O O O O O O O O I O O O O O O O
METHODOLOGY 0 I O O O O O O O O O O O O 0

Process measures 0 O O O O O O O O 0

Rated Gain . . .
Raw Gain . . . .
Residual Gain . .
Alternate Grouping
Attained Health .

Statistical Analyses . . . . . . . .

vii

Variables
Variables

ix

Subjects . . . . . .

Source . . . . . . . . . . . .
Demographics . . . . . . . . .
Psychopathology . . . . . . .
Therapist Data . . . . . . . .
Number of Therapy Sessions . .

Procedure . . . . . . . . . . . . .

RESULTS 0 O O O C O O O O O O O O O O 0

Process Variables . . . . . . . .
Descriptive Statistics . . . .
Hypothesis la . . . . . . . .

Outcome Variables . . . . . . . . .
Descriptive Statistics . . . .
HYPOtheSiS 2 O O O O O O O O

Process-Outcome Relationships .
Hypothesis 1b . . . . . .
Hypothesis 3a . . . . .
Hypotheses 4a and 4b . . .
The Alternate Grouping Met

hod
DISCUSSION 0 O O O O O O O O O O O O O 0
Process Variables . . . . . . . . .
Outcome Variables . . . . . . . . .

Process-Outcome Relationships . . .

Conclusion . . . . . . . . . . . .

LIST OF REFERENCES . . . . . . . . . . .

viii

LIST OF TABLES

Table 1. Pre-therapy and post-therapy means and
standard deviations for the nine symptom
dimensions and the Global Severity Index (GSI)
of the SCL-90—R.

Table 2. Part A: raw frequency of, and percentage
of sample endorsing, each response alternative,
for each process variable. Part B: mean and
standard deviation, for each process variable.

Table 3. Process variable correlation matrix.

Table 4. Raw frequency of, and percentage of sample
endorsing, each response alternative for rated
gain.

Table 5. Raw frequency of, and percentage of sample
corresponding to, each value range for raw gain.

Table 6. Outcome variable correlation matrix.
Table 7. Process-outcome variable correlation matrix.

Table 8. arvalue and significance level for
significant differences between members of
selected pairs of process-outcome correlations.
Part A: correlations vertically adjacent to each
other in Table 7. Part B: correlations not
vertically adjacent to each other in Table 7.

Table 9. The Alternate Grouping Method subgroup
structure, including the mean and standard
deviation of the pre-therapy Global Severity
Index (GSI) for each subgroup and for the entire
sample.

Table 10. Mean and standard deviation for demographic
variables, for each subgroup of the Alternate
Grouping Method and for the entire sample, with
one-way analysis of variance across subgroups for
each variable.

ix

49

53

53
56

56

59
59
62

62

66

Table 11. Process-outcome correlations for the
Alternate Grouping Method (i.e., correlation of
raw gain with process variables, for each subgroup
and for the entire sample.

66

INTRODUCT I ON

Within the last decade, the general efficacy of
psychotherapy has finally been established beyond a
reasonable doubt (Abeles, 1986; VandenBos, 1986). In other
words, it is now generally accepted ”that psychotherapy, as a
generic treatment process . . . [is] demonstrably more
effective than no treatment” (VandenBos, 1986, p. 111). One
of the most significant contributions to this conclusion was
the comprehensive meta-analytic work of Smith, Glass, and
Miller (1980), whose conclusions have received further
support from subsequent reanalyses (Andrews & Harvey, 1981;
Landman & Dawes, 1982). Consequently, VandenBos (1986)
argues that "the mere demonstration of an outcome effect
should no longer dominate the research agenda" (p. 111).

A major remaining issue involves the need for more
effective elucidation of those aspects of psychotherapeutic
RLQQRSS which produce outcome effects (VandenBos, 1986).

In this connection, it is important to explore not only the
effect of different process variables on outcome but also the
effect of employing different variables as meaentee of
outcome. After all, major research projects have
demonstrated that the nature of process-outcome relationships
may be heavily influenced by the type of outcome measures

employed (e.g., Mintz, Luborsky, & Christoph, 1979).

2

The purpose of the present study is to contribute in a
limited but significant way to the elucidation of outcome
measurement issues in psychotherapy process-outcome research.
More specifically, the particular focus of this study is the
examination of changes which may occur in the relationship
between process and outcome as the type of outcome measure is
varied. In this context, the interrelationships of outcome
measures will also be carefully considered.

To accomplish these goals effectively, special attention
must be paid to the selection of appropriate process and
outcome measures. In choosing a process measure, it seems
wise to consider variables which have been the subject of
intensive study over an extended period of time. The three
therapist facilitative conditions identified by Carl Rogers
(1957) — empathy, genuineness, and unconditional positive
regard - appear to satisfy this requirement quite well.

Choice of outcome measures involves selecting
appropriate techniques from the wide variety of measurement
methods currently available. In the present study, outcome
will be considered both as amount of change and as final
adjustment status. In examining the nature of outcome
measures, as well as their relation to process measures,
special attention will be paid to important psychometric
issues which have often been either neglected or mismanaged
in earlier process-outcome research.

To provide a general theoretical and empirical context

for the present study, we will first briefly consider the

3

general outlines of research concerning all three of the
facilitative conditions. Later, in order to achieve greater
specificity and precision, we will focus intensively on only
one of the three conditions - unconditional positive regard -
and its relationship to psychotherapeutic outcome. The
scape of the present study will also be limited in other
ways. At that point, literature considered directly relevant

to the present study will be comprehensively reviewed.

Research Concerning the Rogerian Hypothesis

It was in 1957 that Carl Rogers put forth his classic
formulation of the three conditions which he held to be both
necessary and sufficient for positive change in
psychotherapy: empathy, genuineness, and unconditional
positive regard. By empathy, Rogers meant that:

the therapist is experiencing an accurate, empathic
understanding of the client's awareness of his own
experience. To sense the client's private world as
if it were your own, but without ever losing the
”as if" quality - this is empathy. (p. 98)

The term genuineneee was interpreted as follows:

the therapist should be, within the confines of
this relationship, a congruent, genuine, integrated
person . . . within the relationship he is freely
and deeply himself, with his actual experience
accurately represented by his awareness of himself.
(p. 97)

UnQQnditiQnal pesitixe Legatd was explained in the following

manne r 8

To the extent that the therapist finds himself
experienceing a warm acceptance of each aspect of
the client's experience as part of that client, he

4

is experiencing unconditional positive regard.

(p. 98)

Rogers argued that these three conditions, in order to
be effective, had to be not only offered by the therapist but
also peteeiged by the client. Also, Rogers maintained that
specific therapeutic techniques, of whatever theoretical
derivation, were in essence superfluous - i.e., neither
necessary nor sufficient for change.

Subsequent research in the 1960's, carried on to a large
extent by proponents of Rogers' view and principally
involving client-centered therapy, appeared for the most part
to support his stance. One of the more impressive findings
was the demonstration by Rogers and his associates that
levels of the three conditions were predictive of
psychotherapy outcome even with hospitalized schizophrenics
(Rogers, Gendlin, Kiesler, & Truax, 1967). In this study,
the therapy group's outcome was not significantly different
from that of the control group. However, further analysis
indicated that the calculation of a mean for the therapy
group obscured the differential effects of varying levels of
the facilitative conditions within that group. Within the
therapy group itself, levels of these conditions were
significantly related to outcome.

Comprehensive literature reviews by Truax and Carkhuff
(1967) and Truax and Mitchell (1971) supported the Rogerian
hypothesis, with minor modifications. For example, whereas

Rogers had maintained that all three of the conditions must

5
be present in combination, Truax and his associates adopted
the position that only two of the conditions need be present.

In the 1970's, the tide of opinion and evidence began to
turn against the Rogerian hypothesis. Subsequent literature
reviews were critical of the conclusions drawn by Truax and
his associates (1967, 1971), arguing that some studies had
been misinterpreted or were of poor quality methodologically
(Chinsky & Rappaport, 1970; Rachman, 1973).

Also, two major psychotherapy studies (Mitchell,
Bozarth, Truax, & Krauft, 1973; Sloane, Staples, Christol,
Yorkston, & Whipple, 1975) failed to support Rogers' view.
Because these research projects received considerable
attention from authorities in the field (e.g., Bergin &
Suinn, 1975), their impact was considerable.

The Arkansas Psychotherapy Project (Mitchell et a1.,
1973) was large-scale enterprise involving psychotherapists
of widely divergent theoretical orientations across the
United States. Only a small percentage of the clinicians
employed Rogerian client-centered techniques, which are
specifically designed to emphasize empathy, warmth, and
genuineness in a non-directive atmosphere. Included in the
sample were behavioral, psychodynamic, and other approaches
in which the therapist does not directly focus on developing
these three conditions. Although a weak but significant
positive relationship was found between therapist genuineness

and the outcome of psychotherapyy neither empathy nor

6
unconditional positive regard was significantly related to
therapeutic outcome.

The Temple University Psychotherapy Study (Sloane et
al., 1975) was principally concerned with a comparison of the
relative efficacy of dynamic psychotherapy and behavior
therapy. However, careful consideration was also given to
the relationship of the Rogerian facilitative conditions to
outcome, within each type of therapy. The Temple study
failed to find any significant relationships between these
conditions and the outcome of either dynamic psychotherapy or
behavior therapy. Even genuineness, which had been
significantly related to outcome in the Arkansas study, was
not significantly related in the 1975 project.

In 1977 and 1978, several major psychotherapy literature
reviews appeared, all paying considerable attention to the
three Rogerian facilitative conditions. Most of these
reviews concluded that the earlier optimistic conclusions of
Truax and associates (1967, 1971) were ill-founded in light
of subsequent research and analysis (Gomez-Schwartz, Hadley,
& Strupp, 1978; Lambert, DeJulio, & Stein, 1978; Mitchell,
Bozarth, & Krauft, 1977; Orlinsky & Howard, 1978; Parloff,
Waskow, & Wolfe, 1978). These commentators generally took
the position that the three conditions were clearly important
in most therapeutic situations but could not be considered
necessary in all cases or sufficient in any case outside of
client-centered therapy, where the conditions are the main

content of technique.

7

The general consensus appeared to be that outlined by
Frank (1973), which considers the three conditions to be
"nonspecific factors" found to some extent in most
therapeutic relationships, in contrast with I'specific
factors" involving particular techniques employed by a
particular theoretical approach (e.g., interpretation of
transference in psychodynamic methods). While nonspecific
factors are held to be significant, they are held to be so
only in combination with specific techniques. On the basis
of such reasoning, Parloff et a1. (1978) expressed the view
that the focus of future research should be to study how the
three conditions interact with other factors in therapy, not
to continue to study the conditions in isolation.

Given the seemingly non-confirming evidence of such
notable research projects as the Arkansas (Mitchell et al.,
1973) and Temple (Sloane et al, 1975) studies, one might
understandably wonder why the efficacy of the so-called
facilitative conditions was not rejected altogether, at least
‘outside the rather narrow confines of client-centered
therapy. An important factor in defending the therapeutic
efficacy of the conditions was the criticism of the Temple
and Arkansas studies on methodological grounds. The comments
of Lambert et a1. (1978) are representative of such
criticism. They point out, for example, that the process
measures employed were specifically designed for use with
client-centered therapy and their applicability to other

psychotherapeutic methods was questionable. Also,

8

restriction of range of process variables (scores generally
low in the Arkansas study and generally high in the
psychotherapy segment of the Temple study) may have obscured
process-outcome relationships which would have been clearer
if there had been a broader spread of process scores.

Parloff et a1. (1978) somewhat deemphasized the
significance of such criticisms, holding the Arkansas study
to be the best single test of the Rogerian hypothesis
involving therapeutic methods other than client-centered
therapy. However, in spite of Parloff et al.'s conclusion
that the facilitative conditions were neither necessary nor
sufficient conditions of change, they were unwilling to
abandon the conditions as altogether unimportant.
Recent.ueyalnnmants

From 1978 onward, concern with the facilitative
conditions pet ae has apparently waned within the field of
individual psychotherapy research. One possible indication
of this trend is the degree of attention paid to the subject
in the Annual Regieu Qf Beyeheiegy chapters concerned with
individual psychotherapy. In the 1978 chapter (Gomez-
Schwartz et al., 1978), only one paragraph was devoted to the
conditions, and they received brief indirect mention
elsewhere in the chapter. In the 1981 and 1986 treatments
(Phillips & Bierman, 1981; Parloff, Perry, & Barry, 1986), no
mention of the subject was made. Of course, lack of
attention to facilitative conditions research may be to some

extent attributable simply to lack of interest on the part of

9
the reviewers. However, the review of the literature carried
out in the context of the present study suggests that the
amount of research on facilitative conditions has in fact
decreased.

The subject has continued to receive some - for the most
part indirect - attention in major studies. Generally, this
has been in the context of examining complex
interrelationships among specific therapeutic techniques or
as part of assessing the “therapeutic alliance." The
therapeutic alliance concept has its roots in psychoanalytic
theory and has generally been operationalized for research
purposes as a combination of patient and therapist
collaborative actions (Frieswyk, Allen, Colson, Coyne,
Gabbard, Horwitz, & Newsom, 1986).

Results generated by recent large-scale research
projects have been mixed. For example, research based on the
Vanderbilt Psychotherapy Project, under the general direction
of Hans Strupp, has found no significant relationship between
therapist-offered relationship and psychotherapy outcome
(Gomez-Schwartz, 1978; O'Malley, 1983; Sachs, 1983; Strupp &
Hadley, 1979). The Vanderbilt project also failed to
discover a significant relationship between a therapeutic
alliance measure and outcome (Hartley & Strupp, 1983).

The Penn Psychotherapy Project, directed by Lester
Luborsky, has generated both positive and negative results of
relevance to facilitative conditions research. A large scale

study, reported in 1980, failed to discover a significant

10
relationship of therapist behavior with outcome (Luborsky,
Mintz, Auerbach, Christoph, Bachrach, Todd, Johnson, Cohen, &
O'Brien, 1980). However, subsequent Penn studies, using
measures of therapeutic alliance quality, have found a
significant positive relationship with outcome (Luborsky,
Grits-Christoph, Alexander, Margolis, & Cohen, 1983;
Luborsky, McLellan, Woody, O'Brien, & Auerbach, 1985; Morgan,
Luborsky, Grits-Christoph, Curtis, 5 Solomon, 1982).

The Center for the Study of Neuroses project of the
Langley Porter Institute (University of California, San
Francisco) has also examined the relationship of therapeutic
alliance to outcome (Horowitz, Marmar, Weiss, DeWitt, &
Rosenbaum, 1984). Unlike the Penn studies cited above, the
Center research did not find that relationship to be
statistically significant.

Another major therapeutic alliance study is that
currently being carried out by the Menninger Treatment
Intervention Project (Frieswyk et al., 1986). This study has
the notable characteristic of being concerned exclusively
with borderline patients. Unfortunately, results of the
study have not yet been released.

In light of the mixed nature of the literature, both
before and after 1978, it seems reasonable to conclude that
the three Rogerian facilitative conditions are not
universally necessary or sufficient conditions of therapeutic
change. However, there appears to be increasing recognition

that nonspecific factors - such as the Rogerian facilitative

11
conditions - should net be considered merely "placebos“ in
need of experimental control (Abeles, 1986; Parloff, 1986).
Instead, non-specific factors may be among the most important
elements contributing to the efficacy of techniques advocated
by the various schools of psychotherapy (Strupp: 1986).
Referring directly to the Rogerian facilitative conditions,
Strupp (1986) recently made a most compelling observation:

It is generally recognized that the kinds of

learning (or unlearning) we call "therapeutic"

proceed within the context of a particular human

relationship. Essential ingredients characterizing
such a relationship are acceptance, warmth,

respect, empathy, caring, and the like. This is

not to assert that therapeutic learning cannot

occur in other circumstances . . ., but most

observers are in partial agreement with Rogers

(1957) that the foregoing "conditions" are

essential. (p. 123)

Strupp thus rejects Rogers' contention that the three
facilitative conditions are sufficient for therapeutic
change, but he comes quite close to arguing that they are
necessary. Patterson (1984), however, is not content with
this kind of partial acceptance of the Rogerian view. He
sharply criticizes most reviewers after Truax and Mitchell
(1971), accusing them of bias against the so-called
”nonspecific" facilitative conditions and bias in favor of
the importance of specific therapeutic techniques.

According to Patterson, reviewers up to 1978 have
selectively and inaccurately reviewed available research,
distorting what he feels is overwhelming evidence in favor of

the original Rogerian hypothesis. After 1978, Patterson

suspects a general neglect of the subject altogether. As

12
evidence, he notes the pattern, which we considered
previously, of declining attention to facilitative conditions
research in the Annual Regieu Qf Payeheiegy.

Patterson's viewpoint appears extreme and is certainly
atypical, tending to deemphasize the considerable volume of
research evidence against the Rogerian hypothesis and
somewhat exaggerating the importance of favorable research
findings. Nevertheless, he does a valuable service by
cautioning against uncritical acceptance of reviewers'
conclusions. Also, and perhaps more importantly, he
identifies several methodological problems which have
contributed to confusion within the research concerned with
the facilitative conditions. Among these are failure to
correct for unreliability of measurement instruments and
failure to use comparable outcome measures across studies.

Lambert et a1. (1978) also addressed the issue of
methodological difficulties, noting the wide variation across
studies in selection of the perspective from which
therapeutic process is rated. The client, the therapist,
independent judges, or some combination of these have been
employed as raters. He notes evidence that there is little
relationship among these different ratings and points out the
inappropriateness of comparisons of studies using non-
comparable rating perspectives.

Summarx
In summary, over the past 30 years, research has failed

to give definitive support to Rogers' original hypothesis

13
that empathy, genuineness, and unconditional positive regard
are both neeeeeaty and anftieient conditions for therapeutic
change. However, the bulk of present opinion emphasizes that
these variables are quite impettant in determining the
outcome of psychotherapy. And, although some have called for
abandonment of research focused on the facilitative
conditions pet 5e (Parloff et al., 1978), others have
emphasized the need for continued research in this area
(Lambert et al., 1978; Patterson, 1984). The latter view
holds that methodological problems are at least in part to
blame for the mixed results generated in the past by such
research. Thus, the proper course of future research is not
to abandon focus on the facilitative conditions but rather to

improve the methodology employed in studying them.

The Purpose and Scope of the Present Study

The present study is intended to contribute to greater
methodological clarity in the study of the Rogerian
facilitative conditions. More specifically, our purpose is
to determine how the relationship between Rogerian process
and treatment outcome is affected by the type of outcome
measure used. In order to carry out an effective analysis,
it is necessary to limit the scope of the study, so that a
relatively few relationships may be examined in detail. In
particular, special attention will be paid to limiting the
number of variables examined, so that a favorable subjects-

to-variables ratio can be maintained. This is especially

l4
important in the present study because data for only a
relatively small number of subjects will be available for
use. Circumscribing the scope of the study reduces the
generalizability of its findings but may also increase their
theoretical significance in the areas which are in fact
investigated.

The first limitation concerns the perspective from which
process and outcome are to be judged. Three perspectives
have been the subject of research investigation: those of
the client, therapist, and nonparticipant judge. The present
study will concern itself solely with the client's
perspective, exploring relationships of process to outcome
within the client's own phenomenological world.

Limitation of perspective to that of the client raises
issues for process measurement which have been addressed at
length by researchers and theoreticians in the field of
facilitative conditions research. As previously noted,
Rogers (1957) emphasized that empathy, warmth, and
genuineness must not only be offered by the therapist but
also perceived by the client in order to be therapeutically
effective. Barrett-Lennard (1962) followed up on this idea
by developing a client self-report system for rating the
facilitative conditions.

Later researchers have tended to abandon or deemphasize
Rogers' original position on this point, regarding the client
as too biased to rate process accurately. This perspective

is most notably seen in the work of Truax and Carkhuff (1967)

15
and Truax and Mitchell (1971), who developed elaborate
measurement schemes employing nonparticipant judges as
raters.

As Frieswyk et a1. (1986) observe, within the
psychoanalytic school of thought, patient perceptions are
also generally held to be biased: '

there is little disagreement that the therapeutic

alliance is a manifestation of transference . . . .

How the patient experiences the therapist

transferentiall is related to . . . a core or

matrix of exper1ence with affectionate and helping

others that arises from developmentally early,

formative relationships. Disruption and

deformation of those early experiences undermine

the patient's capacity to develop basic trust and

impair later efforts to establish helping

relationships. (p. 33)

Gurman (1977), Parloff et a1. (1978), and Lambert et a1.
(1978) emphasize two important points regarding the client
vs. judge rating issue. First, most research has shown that
the client and the judge do not agree on their evaluations of
process. Second, client ratings are associated with outcome
at least as strongly as are judge ratings, even when outcome
is not rated by the patient. They conclude that, although
client ratings may be biased to some extent, judge ratings
are not necessarily more "objective." Judges simply measure
a different realm of experience, one not necessarily more
significant than that of the client.

On the basis of such arguments, it seems reasonable to
consider the patient's perspective a legitimate subject of

investigation. However, in the present study no attempt is

made to argue that the client's perceptions are bias-free.

'16
In fact, some degree of client bias is assumed and is used as
one important basis for prediction of process-outcome
relationships.

A second limitation will be imposed regarding the
process variables to be addressed. The present study will
concern itself with only one of the three Rogerian
facilitative conditions. This condition has been variously
referred to as 'nonpossessive warmth" and "unconditional
positive regard" (e.g., Barrett-Lennard, 1962; Rogers, 1957;
Truax and Mitchell, 1971). As noted earlier, it consists of
such phenomena as the therapists's interest, warmth,
acceptance, and respect with regard to the client.

How can one of the three facilitative conditions be
examined without also examining the other two? Rogers (1957)
argued that the three conditions must occur in combination in
order to be effective. Truax and Carkhuff (1967) and Truax
and Mitchell (1971) came to the conclusion that only two of
the three conditions need be present. Since that time, a
number of studies have considered one of the three conditions
separately and achieved positive results (e.g., Free, Green,
Grace, Chernus, & Whitman, 1985; Kurtz & Grummon, 1972).
Thus, it appears legitimate to study one facilitative
condition in isolation. Of course, it should be clearly
understood that, as a result, the present study cannot be
termed a attiet test of the cemplete Rogerian hypothesis.
However, Rogers (1957) clearly put considerable emphasis on

unconditional positive regard, and, consequently, a study

l7

focusing on this variable alone can be treated as a
significant contribution to facilitative conditions research.

A third limitation of process measurement in the present
study involves measuring process only retrospectively, post-
therapy. This method has been effectively employed in major
studies by Strupp and his associates (Strupp, Fox, & Lessler,
1969; Strupp, Wollach, & Wogan, 1964). The major issue
associated with retrospective process measurement has to do
with its potential confounding with post-therapy outcome
measurement, when the client is the rater of both process and
outcome. Here, issues associated with using the client as
the sole rater of outcome become salient. Because such
considerations are of central importance in the present

study, they will now be considered at length.

Preliminary Methodological Considerations

Thelmnlimmnsefclientaias

Several commentators have speculated that retrospective,
post-therapy process measures are bound to be heavily
affected by the client's perception of the success of therapy
(Board, 1959; Carmichael, 1970; Fretz, 1966; Grigg &
Goodstein, 1957; Jones, 1968; Lambert et al., 1978; Lesser,
1961; Ryan & Gizynski, 1971; Strupp, et al., 1969). It seems
probable that clients who perceive therapy as having gone
well will describe the therapeutic process in more positive
terms than will those clients who feel therapy did not

succeed.

18

If post-therapy retrospective process measures are
confounded with perceived success of therapy, then it is
important to address the question: what do clients consider
success? Two possibilities are: (l)-the amount of change
which occurred in therapy and (2) the perceived final
adjustment status of the client (i.e., to what extent the
client feels "cured”).

Three studies have indicated that clients' global
retrospective outcome evaluations are more highly related to
post-therapy scores on psychological adjustment measures than
they are to pre—post difference scores on the same measures
(Green, Gleser, Stone, & Seifert, 1975; Mintz, 1972; Mintz,
Luborsky, & Christoph, 1979). This was found to be true exen
when the global outcome measure was specifically worded so as
to measure amount of change rather than post-therapy
condition. Three additional studies have also generated
similar results (Cartwright, Kirtner, & Fiske, 1963; Fiske,
1971; Garfield, Prager, & Bergin, 1971). And, in a recent
review article, Bentler and Hamblin (1986) make concurrent
observations. A conclusion that might be drawn from this
information is that the patient's perception of the success
of therapy is more a product of perceived final adjustment
level than amount of change.

These are merely suggestive correlational data, not
experimental data from which causal inferences may be drawn.
But it seems reasonable to speculate as follows. If post-

therapy process measurement is determined in part by

19,

perceived success of therapy, and perceived success of
therapy is determined in part by final adjustment status,
then it may be that process measurement is determined in part
by final adjustment status. If so, then the greater the role
played by final adjustment status in determining a given kind
of outcome measurement, the stronger the process-outcome
relationship will be. Thus, our focus in developing
hypotheses about process-outcome relationships will initially
be to determine to what extent each outcome measurement
considered is related to final adjustment status.
Ratedym,and8aaidualfiain

At the outset, we will consider three kinds of outcome
measures: rated gain, raw gain, and residual gain. This
terminology is based in part on the treatment by Mintz et a1.
(1979).

Rated gain refers to a global retrospective estimate by
the client of the amount of change which has occurred as a
result of therapy. Ram gain refers to a difference score
obtained by simple subtraction of post-therapy score from
pre-therapy score (or vice versa), where pre- and post-
therapy scores have been obtained using some form of
psychological test of adjustment. As noted, rated gain has
been found in prior research to be more closely related than
raw gain to final adjustment level (here considered to be
post-therapy psychological test score). Thus we would
expect, by our prior argument, that rated gain would be more

strongly related than raw gain to post-therapy process

20

measurement.

The third outcome measure to be considered is Leeidnal
gain. This is a somewhat complex concept and some background
explanation will be helpful. In 1970, a distinguished
committee on Planning of Research on Effectiveness of
Psychotherapy made the following observation on the use of
pre-post difference scores in correlational research:

For correlational studies, it is necessary to have

an appropriate index of the treatment effect

foreach patient. It is well known that the simple

raw gain score (the difference between pre and post

treatment scores) has the severe defect of being

determined unduly by the pretreatment score ( under
some circumstances, the raw gain will correlate

0.71 with initial level). The residual gain score

adjusts the post score for that part contributed by

the initial score. (Fiske, Hunt, Luborsky, Orne,
Parloff, Reiser, & Tuma, 1970; p. 30)

They go on to note that a correction for unreliability of the
particular measurement instrument must be included in the
residual gain calculation.

For a more complete understanding of the need for
residual gain score calculation, it is important to be aware
of not only the existenee of a correlation between pretherapy
score and raw gain but also the reasons for the correlation.
In their recent review article, Bentler and Hamblin (1986)
describe two such reasons:

First, extreme scores tend to regress toward the

mean on subsequent measurement occasions because of

the inordinate amount of error variance that is

likely to be present in such scores. Second,

extreme scores are more likely to change than less

extreme scores because there is simply more room

for change to occur on finite assessment scales.

(p. 49)

For these and related reasons, criticism of the use of

21
raw gain scores has been widespread (e.g., Fiske, 1971;
Gurman, 1977; Luborsky, 1971; Manning & DuBois, 1962;
Meltzoff & Kornreich, 1970; Tucker, Damarin, & Messick,
1966). This criticism is highly significant for the
evaluation of research on the facilitative conditions because
raw gain scores have been utilized in numerous studies (e.g.,
Carmichael, 1970; Feitel, 1968; Gross & DeRidder, 1966;
Jones, 1968; Kiesler, 1967; Kurtz & Grumman, 1972; Zauderer,
1967). '

When raw gain is statistically adjusted to control for
the influence of pretest score, the correlation of residual
gain with rated gain is higher than the correlation of raw
gain with rated gain (Green et al., 1975; Mintz, 1979).

Mintz (1979) attributes this phenomenon to the fact that ”the
residual gain function does weight final status more heavily
than pretreatment status" (p. 332). Following our earlier
line of reasoning, we can expect the process-outcome
relationship to be greater when residual gain is the outcome
measure than when raw gain is used to measure outcome.

In summary, we can predict that process-outcome
correlations will be in the following order of decreasing
strength: process with rated gain, process with residual
gain, and process with raw gain.

The Altannaifi Grouping Mfiihnd

Before proceeding further, it will be productive to

consider the issue of residual gain in a bit more depth.

Although the residual gain method has been widely recommended

22
as a means of controlling for the influence of pretest score
on raw gain scores (see earlier citations), the method is not
without considerable liabilities. Cronbach and Furby (1970)
observed somewhat critically:

Residualizing removes from the posttest score, and

hence from the gain, the portion that could have

been predicted linearly from pretest status. One

cannot argue that the residualized score is a

“corrected“ measure of gain, since in most studies

the portion discarded includes some genuine and

important change in the person. (p. 74)

Because of this liability of the residual gain method,
in the present study an alternative strategy will also be
employed. Another way to control at least partially for the
influence of pretest score is to subdivide the client sample
into relatively homogeneous pretest groups. This approach
will be referred to as the alternate greuping methed.
Trichotomization would provide a finer discrimination than
simple division into "mild" and |'severe" pretest groups.

With a trichotomization approach, three groups would be
generated: I'mild" pretest, “moderate" pretest, and “severe"
pretest.

As noted, residualizing gain is expected to increase the
relationship of gain with process. Since the alternate
grouping strategy is also designed to control for pretest
effect on gain, it seems reasonable to assume that the
relationship of process to outcome in each subgroup of raw
gain scores will be stronger than the relationship of process

to outcome for the whole (undivided) sample's raw gain

SCOIES.

23

It is important to note that the subgrouping alternative
to the residual gain method is not without difficulties of
its own. First, the correlation of pretest with raw gain
score will still be a problem within each subgroup, although
the problem will be diminished by the increased homogeneity
achieved through the subgroup division process. Second, and
perhaps more important, the number of subjects on which
correlational analyses are performed will be cut in thirds,
thus reducing power and increasing the risk of sampling error
(Glass & Stanley, 1970). 'Thus, both the residual gain method
and the alternate grouping method have their respective
liabilities. Choice of one method over the other depends on
which liabilities are given greater weight in a particular
research context. Perhaps the results of the present study
will prove helpful in dealing with this issue in future
research.
Attained Health

Up to this point, all outcome measures have represented
various attempts to measure amount of change. It seems
worthwhile also to consider a somewhat different approach to
outcome measurement, one which involves both amount of change
and final adjustment status. In the present study, this
measure will be described as attained health. Abeles (1986)
has referred favorably to the method because it provides a
reasonably clear and precise solution to the problem of
assessing the glinieal - as opposed to the statistical -

significance of psychotherapeutic change.

24

Attained health is computed using a method outlined by
Jacobson, Follette, and Revenstorf (1984). From their
perspective, psychotherapeutic change is clinically
significant only if two things are true. First, the amount
of change must be greater than that which could be attributed
simply to the unreliability of the measurement instrument.
Second, the post-therapy score on the instrument must fall
within the normal range of functioning, as defined by
normative data for that instrument. Attained health is thus
an "all-or-nothing” measure indicating whether or not a
client has achieved clinically significant change.

Speculation as to the relationship, between process
variables and attained health, requires recourse to the same
argument used with the other outcome measures we have
considered. Clearly, attained health, because of its method
of calculation, is closely related to final adjustment
status. However, this relationship will probably not be as
strong as the relationship between rated gain and final
adjustment status, for the following reasons. First, the
potential for subjective distortion appears greater for rated
gain than for attained health.~ Attained health is derived
from multiple specific responses on a psychological test,
whereas rated gain is simply a single global outcome
estimate. Second, attained health contains a raw gain
component, which is involved in the control for unreliability
of the measurement instrument. Calculation of raw gain is,

of course, not involved in the global rated gain estimate.

25
Third, attained health involves reference to a normative
sample, in the definition of "health.” When the client
estimates rated gain, no formal reference to a normative
group is made, although informal comparison with other
persons probably occurs.

For these three reasons, then, it seems that rated gain
will be more strongly affected than attained health by the
client's subjective evaluation of the "success" of therapy.
And it is, of course, this subjective evaluation of success
which we have held to have a considerable confounding effect
on process variable measurement. Thus, it seems reasonable
to assume that the relationship of attained health to process
variables will be weaken than that of rated gain. However,
attained health is probably mete strongly related to
perceived therapeutic success than is residual gain.
Consequently, it seems likely that attained health will be
related more strongly than residual gain to process
variables.

Summatx

In sum, on the basis of methodological speculation,
outcome measures can be ordered from strongest to weakest
relationship with process measures, as follows: rated gain,
attained health, residual gain, and raw gain. When the
alternate grouping method is employed, the relationship
between raw gain and process will be greater in the moderate
and severe subgroups than in the entire sample. In the mild

subgroup, however, this will not be the case.

26

Process-Outcome Research Evidence

Thehiteratureneliswamcess

By this point, we have arrived at a predicted ordering
of process-outcome relationships, by developing the concept
of confounding of post-therapy process measurement with
perceived success of therapy. The research data for these
predictions have been drawn from studies which compared
different kinds of outcome measures but were not necessarily
concerned with positive regard - the process measure of
interest in the present study.

It is now necessary to look at the evidence of process-
outcome studies to see how it reflects on issues of interest
here. For this purpose, a selective literature review was
conducted, involving research dealing specifically with one
or more facets of the positive regard concept. As noted
earlier, this concept is considered here to involve elements
such as therapist's warmth, interest, respect, and acceptance
with regard to the client. In this review, the earlier
reviews of Gurman (1977) and Orlinsky and Howard (1978) were
used as the chief sources of studies up to 1976. To update
'the work of these earlier reviews, a survey of Eeyehelegieal
Absttaete from 1976 through mid—1986 was also conducted.

In the review process, several principles of selection
were employed. First, the field was limited to studies which
were concerned with individual psychotherapy or behavior
therapy, specifically excluding group psychotherapy and

vocational or educational counseling. Second, only studies

27

involving primarily adult, outpatient, nonpsychotic
individual were selected. In cases where younger or
psychotic clients were included in the Client sample, these
categories constituted a minority of subjects. Third,
analogue studies were excluded in favor of those employing
actual therapeutic encounters. Fourth, and most important,
the review was limited to studies in which the client rated
both process and outcome. Most studies included other
perspectives in addition to that of the client, but these
perspectives were not the focus of interest in considering
research results.

The review process yielded 16 process-outcome studies,
which can profitably be divided into two groups of eight.
The first group includes studies in which both process and
outcome were measured posttherapy - either at (or near)
termination or at follow-up. It was decided to consider
together studies using termination measurements and follow-up
measurements, on the basis of the recent comprehensive
literature review of Nicholson and Berman (1983). They
demonstrated that "information obtained at follow-up often
added little to that obtained at the end of treatment" (p.
261). In the present group of eight studies, no clear
difference appeared to be evident in the results of research
making measurements at (or near) termination (Bent, Putnam,
Kiesler, and Norwick, 1976; Cooley & LaJoy, 1980; Lorr, 1965;
McClanahan, 1974) and research taking measurements at follow-

up (Board, 1959; Grigg & Goodstein, 1957; Strupp et al.,

28
1969; Strupp et al., 1964). For convenience in subsequent
discussion, these eight studies will be referred to as Post—
Therapy-Process (PTP) studies.

The second group of eight studies includes research in
which outcome was measured at (or near) termination or at
follow-up but process was measured at one or more points
prior to termination. For convenience, these studies will be
referred to as Mid-Therapy-Process (MTP) studies. Six of
these studies involved termination outcome measurements
(Barrett-Lennard, 1962; Ford, 1977; Gross & DeRidder, 1966;
Mendola, 1981; Saltzman, Luetgert, Roth, Creaser, & Howard,
1976; Yagel, 1985). The two other studies involved outcome
measurement at follow-up (Luborsky et al., 1985; Marziali,
1984).

The two follow-up studies are different from the six
termination studies in another important way. Luborsky et
a1. (1985) and Marziali (1984) used process rating scales of
the therapeutic alliance, which, as previously noted,
includes more than simply the positive regard construct. In
these two studies, unlike many other therapeutic alliance
studies which have been cited earlier, process is measured
either primarily or entirely by the client rather than the
therapist or a nonparticipant judge.

Luborsky et a1. emphasize that their process measure is
heavily weighted toward the kinds of variables to which we
have referred in the present study under the rubric of

positive regard. This also appears to be the case for the

29

scale employed by Marziali. Thus, the results of these two
studies can be considered at least roughly comparable to
those of the other six in the MTP group. Nevertheless, their
results will be identified separately when conclusions are
being drawn in subsequent discussion. To facilitate
differentiation of MTP studies, the group of six will be
designated Primary MTP studies, and the group of two (follow-
up) studies will be referred to as Secondary MTP studies.

Clearly, the FTP research is more directly relevant to
the present study than is the MTP research. MTP studies are
considered only because they examine certain outcome
variables not addressed in the FTP rsearch. Conclusions
based on MTP studies should therefore be given less weight
than those founded on PTP research results.
TheReleyancleEriarResearth

Having described the literature review process, it is
now appropriate to examine how the results of the review
relate to the predictions tentatively formulated in our
earlier discussion of process-outcome relationships. First,
it is unfortunate that neither the FTP nor the MTP group
provides any direct evidence concerning how our different
process—outcome relationships relate to each other, in terms
of hierarchical ordering of strength of relationship. No
single study makes use of more than one of the outcome
variables of interest to us.

One MTP study (Marziali, 1984) appears at first glance

to compare process—outcome relationships involving residual

30
gain and rated gain. However, upon closer examination, it
becomes clear that the correlation between process and rated
gain is a pattial correlation, with pretherapy social
adjustment as the variable partialed out. Thus the Marziali
process-outcome relationship involving rated gain is not
comparable to that conceived in the present study.

Although the reviewed studies do not address the
hierarehieal.ordering of process-outcome relationships, they
do provide information concerning the statistical
significance and diteetien of some of the process-outcome
relationships of interest. Accordingly, it is to the issues
of statistical significance and direction that we now turn.
Rated Gain

Let us first consider the relationship between positive
regard and rated gain. Here, it is important to note
initially that, in some of the eight PTP studies, the global
outcome measures used involve assessment of perceived final
status (e.g., "satisfaction" with theraPY) rather than
perceived amount of change (i.e., rated gain). However,
these measures will be considered comparable to the present
study's concept of rated gain because of the previously cited
results reported by Green et a1. (1975). They found that
global retrospective outcome measures behaved similarly,
regardless of whether a self-report item described outcome as
final status or as amount of change. Thus, for convenience,
when research studies are cited in the following discussion,

the term gated gain will be used in a generic sense to refer

31
to all global retrospective outcome measures. This
convention is not intended to imply that all such measures
are precisely equivalent but only that they are sufficiently
comparable to be legitimately considered together.

When the above-described simplifying convention is
employed, the following observation may be made. In all
eight PTP studies, a significant positive relationship was
found to exist between positive regard and rated gain. Thus
it appears reasonable to predict that a significant positive
relationship will also be found in the present study.

It is interesting to note here that the process-outcome
relationship involving rated gain is much less favorable in
the MTP group. The two Primary MTP studies using rated gain
(Salzman et a1, 1976; Yagel, 1985) failed to find a
significant positive relationship. This appears to support
our earlier argument that confounding of process and outcome
assessments is at least in part responsible for significant
positive relationships between rated gain and process
:measured post-therapy. When process is measured mid-therapy
rather than post-therapy, fewer significant relationships are
evident.

BanandReeidualGain

Turning to consider process-outcome relationships
involving raw and residual gain, we must limit attention to
the MTP research, since none of the PTP studies employed
either outcome measure. As noted earlier, MTP research is

not directly comparable to the present study, but it is the

32
most closely related research available. The only study
which clearly employed raw gain is that of Gross and DeRidder
(1966). They found both significant positive and
nonsignificant relationships with process measures. The two
Secondary MTP studies both employed residual gain and found
significant positive relationships with process (Luborsky et
al., 1985; Marziali, 1984). Two other (Primary) MTP studies
employed difference scores but cannot be considered here
because they do not clearly specify whether residual or raw
gain was used. .

On the basis of the foregoing evidence, one might
reasonably expect to find statistically significant positive
correlations between process and both raw and residual gain.
However, the significant relationship can be be predicted
with greater confidence in the case of residual gain than in
the case of raw gain.

The Alternate GLQHRing Methed

None of the PTP or MTP studies directly employed the
alternate grouping method for control of correlation of
pretest score with pre-post difference score. However,
Barrett-Lennard (1962) did apply some elements of the
grouping technique. He encountered the problem of pretest-
score correlation with difference score but was apparently
unaware of the residual gain method option. So he
dichotomized his sample into ”mild" and "severe” pretest
groups, as the only available means of control. However, he

discovered that the range of difference score values in his

33
mild pretest group was severely restricted. In other words,
people who began therapy with relatively mild disturbance did
not change very much during the course of therapy. When raw
gain was correlated with process variables, the correlations
were positive and significant only for the severe pretest
group. Correlations for the mild pretest group were not even
reported, because Barrett-Lennard felt that restriction of
range rendered a correlational analysis inappropriate.

With this difficulty in mind, the following predictions
seem reasonable. The process-outcome relationship will be
positive and statistically significant in the “severe” and
"moderate“ pretest groups in the present study. In the
"mild" pretest group, the relationship will not be
significant.

Attainedflealth

Finally, none of the PTP or MTP studies employed the
attained health outcome measure. This is not surprising,
since the method was first formally outlined by Jacobson et
al. in 1984, and most of the studies we have considered
predate that publication. However, on the basis of the
expected close relationship of attained health to rated gain,
it seems reasonable to predict that attained health, like
rated gain, will have a significant positive relationship
with therapeutic process.

In sum, we may predict that almost all process-outcome
relationships in the present study will be statistically

significant and positive. The sole exception is the

34
relationship within the mild pretest group formed in the
context of the alternate grouping method. However, because
of the nature of the evidence available, the significance of
the process-outcome relationship can be expected with greater
confidence for rated gain than for residual gain, raw gain,
or attained health. The relationship involving raw gain
appears to be the one predictable with the least confidence
of all, given the mixed results of prior research.

By this point, predictions have been developed
concerning the hierarchical order and statistical
significance of process-outcome relationships. Two final
issues require brief consideration: the interrelationships
among outcome variables and the interrelationships among
process variables.

In the previous discussion of the predicted hierarchical
order of process—outcome correlations, certain expected
interrelationships among outcome measures were either
described or strongly implied. These will now be clearly
delineated. Rated gain can be expected to relate more
strongly to residual gain than to raw gain. Attained health
will probably relate most strongly to rated gain, less
strongly to residual gain, and least strongly to raw gain.

Finally, we turn our attention to the process variables
which have previously been identified as components of

positive regard: warmth, interest, respect, and acceptance.

35
Various combinations of these variables have been mentioned
in the context of positive regard from Rogers (1957) onward
(e.g., Barrett-Lennard, 1962; Patterson, 1984; Truax &
Mitchell, 1971). If positive regard is a valid construct,
then these four variables should be highly interrelated.
Thus it appears reasonable to predict that, in the present
study, interrelationships among these variables will not be
significantly different from each other. We may also expect
that, when any one outcome variable is considered separately,
there will be no significant difference in strength of

relationship among the four process variables and outcome.

HYPOTHESES

In order to clarify and summarize the foregoing

arguments, predictions formulated earlier will now be stated

as formal research hypotheses.

Hxnnthesis Qne: Interrelationships of Process Variables

1a

Interrelationships among process variables will not
be significantly different from each other.

When any one outcome variable is considered
separately, there will be no significant difference
in strength of relationship among the four process

variables and outcome.

Hypetheaie Tue: Interrelationships of Outcome Variables

2a

Rated gain will relate more strongly to residual
gain than to raw gain. These differences in
strength of relationship will be statistically
significant.

Attained health will relate most strongly to rated
gain, less strongly to residual gain, and least
strongly to raw gain. These differences in
strength of relationship will be statistically

significant.

36

37

pretheais Thtee: Hierarchical Order of Process-Outcome

Relationship Strength

3a:

Outcome measures will be ordered from strongest to

weakest relationship with process measures, as

follows: rated gain, attained health, residual

gain, and raw gain. Differences in strength of

relationship will be statistically significant.

When the alternate grouping method is employed, the

following relationships will be evident:

(1)

(2)

The relationship between process variables and
raw gain will be significantly stronger within
the severe and moderate subgroups than in the
client sample as a whole.

The relationship between process variables and
raw gain will net be significantly stronger
within the mild subgroup than in the client

sample as a whole.

Hypethesis Rent: Statistical Significance and Direction of

Process-Outcome Relationships

4a:

'5‘

A11 statistically significant process-outcome

relationships will be positive.

Process-outcome relationships involving the

following outcome variables will be statistically

significant: attained health, rated gain, residual

gain, and raw gain.

When the alternate grouping method is employed,

process-outcome relationships will be statistically

38
significant within the severe and moderate

subgroups but not within the mild subgroup.

METHODOLOGY

PIOCGSS Measures

Four aspects of positive regard have been measured:
warmth, acceptance, respect, and interest. Each of these
variables has been measured by means of a single
questionnaire item taken from the Strupp Post-Therapy Client
Questionnaire (Strupp et al., 1969). The exact wording of

each item is as follows:

Watmth: 'I felt there was usually a good deal of
warmth in the way the therapist talked to me."

Ateeptanee: 'I usually felt I was fully accepted
by the therapist."

Reagent: "I am convinced that the therapist
respected me as a person.“

lnteteet: 'I feel the therapist was genuinely
interested in helping me."

Client responses are indicated on a 9-point Likert-type scale

with values ranging from 1 ("strongly agree”) to 9 (”strongly

disagree”).

Outcome Measures

RatedGain
Rated gain was measured by means of a single

questionnaire item from the Strupp Post-Therapy Client

39

40
Questionnaire (Strupp et al., 1969). The exact wording of
the item is as follows:

"To what extent have your complaints or

symptomsthat brought you to therapy changed as a

result of treatment?"

Client responses are indicated on a 6—point Likert-type scale
with values ranging from 1 ("completely disappeared") to 6
('got worse").

Raw Gain

The instrument used in the calculation of raw gain is
the SCL-90-R, the revised version of the 90-item Hopkins
Symptom Checklist developed by Derogatis and his associates
at the Johns Hopkins University School of Medicine
(Derogatis, 1977). This symptom checklist is the product of
extensive empirical research (Derogatis 8 Cleary, 1977;
Derogatis, Klerman, & Lipman, 1972; Derogatis, Lipman, &
Covi, 1973; Derogatis, Lipman, Covi, & Rickels, 1971, 1972;
Derogatis, Lipman, Covi, Rickels, & Uhlenhuth, 1970;
Derogatis, Lipman, Rickels, Uhlenhuth, & Covi, 1974a, b;
Derogatis, Rickels, 8 Rock, 1976; Derogatis, Yevzeroff, &
Wittelsberger, 1975).

As a result of this work, the reliability and validity
of the instrument have been well established. This
information, along with extensive normative data, is
summarized in the SCL-90-R manual by Derogatis (1977). Table
12 (p. 15) details internal consistency and test-retest
reliability for each of the symptom dimensions of the SCL-90-

R. Internal consistency coefficients range from 0.77 to

41
0.90. Test-retest coefficients vary from 0.78 to 0.90.

The instrument has been well accepted in the field of
clinical assessment, as is evidenced by the inclusion of its
earlier version (the SCL-90) in an NIMH standard test battery
for psYchotherapy outcome research (Waskow & Parloff, 1975).
Also, Derogatis (1983) has shown himself quite capable of
responding to criticism of the instrument when it has arisen
(e.g., Kass, Charles, Klein, & Cohen, 1983a, b).

A word should be added here regarding the choice in the
present study to employ a symptom checklist rather than a
standard personality inventory such as the MMPI. Green et
a1. (1975) found symptom change to be the most useful means
of measuring change in psychotherapy. They considered a
symptom checklist to be less theory-bound than traditional
personality tests. Also, Bergin and Lambert (1978) described
symptom checklists as at least potentially more sensitive to
psychotherapeutic change than the MMPI. In their discussion
of checklists, the SCL-90 (precursor to the SCL-90-R) is
specifically mentioned as representative. Thus the SCL-90-R
is at least as appropriate, if not more appropriate, than
more traditional measures of personality functioning.

The SCL-90—R consists of 90 items which have been found
to fall into 9 symptom categories, with minor variations in
the number of items per category. The 9 categories are:
Somatization, Obsessive-Compulsive, Interpersonal
Sensitivity, Depression, Anxiety, Hostility, Phobic Anxiety,

Paranoid Ideation, and Psychoticism. Six additional items

42
concern eating behavior, sleep, guilt, and thoughts of death.
A score is calculated for each symptom category. Also, a
Global Severity Index (GSI) is calculated as the average
rating for all 90 items.

Each item allows the subject 5 response alternatives on
a Likert-type scale ranging from 0 ("Not at all“) to 4
(”Extremely“) in reference to the symptom described in the
item. A symptom category score is the average of the scores
on the items composing the category. As noted, the GSI is
the average of all item scores in all categories, as well as
the 6 additional items.

In the present study, raw gain is defined as the
difference between pre-therapy and post-therapy GSI scores.
Since post-therapy score is subtracted from pre-therapy
score, a positive raw gain score reflects symptom reduction
in the course of psychotherapy.

ResidualGain

The concept of residual gain has already been explained
in general terms. More specifically, Cronbach and Furby
(1970) describe the process as follows:

A gain is residualized by expressing the posttest

score as a deviation from the posttest-on-pretest

regression line. The part of the posttest

information that is linearly predictable from the

pretest is thus partialled out." (p. 68)

They go on to observe that the calculation is
considerably more complex than a simple partial correlational
analysis. Difference scores pose special psychometric

problems. One of the most important of these difficulties is

43
that the unreliability of the measurement instrument
contributes error to the difference score. Cronbach and
Furby (p. 72) provide a computation method which is designed
to minimize the deleterious effects of such psychometric
problems.

In this method, correlations between a difference score
and some third variable are computed directly from
covariances rather than by separately calculating a residual
gain score for each subject and then correlating gain with
the third variable. According to Cronbach and Furby,
indixidnal residual gain scores are net accurate reflections
of an individual's "actual”, “base-free" amount of change.
The residualizing technique is best limited, they argue, to
statistics calculated for an entire group of subjects. For
this reason, in the present study residual gain scores have
not been calculated for each subject, and no frequency
distribution data has been presented for the residual gain
variable. Only correlational analyses, based on covariance
calculations and reliability coefficients, have been
employed.

Alternatefimuninsuethed

The subject sample has been trichotomized into "mild”,
"moderate”, and "severe“ pretest groups. The simplest way to
divide subjects would have been to do so on the basis of GSI
score. However, the GSI has a notable deficiency in this
regard. Because it is an average across all symptom

categories, it tells nothing about the relative severity of

44
the different categories, when they are compared to each
other. A moderate GSI score could reflect (l) a profile in
which 3 categories are of high severity and the rest are of
low severity or (2) a profile in which all of the categories
are of moderate severity.

For this reason, a grouping scheme was devised for the
present study which takes into account the relative severity
of the different symptom categories. First, for each
subject, each category was designated either 'low' (categdry
average 2 or lower on the 4—point scale) or 'high' (category
average greater than 2). Then subjects were divided into
groups on the basis of how many of their categories were
high. The definition of each group, in terms of number of
high categories, was determined so as to obtain 3 groups of
roughly equal numbers of subjects. The relationship between
process variables and raw gain was then determined within
each of the 3 groups.

Attained Health

Jacobson et a1. (1984) describe the calculation of an
outcome measure which has been described in the present study
as "attained health.” It should be noted clearly that this
term was coined for the present study and was not used by
Jacobson et a1. as a label for their method.

They outline 2 steps for determining the attained health
score. First, it must be determined that the amount of
change (raw gain) is greater than that which could be

attributed to measurement error arising from the

45
unreliability of the measurement instrument. To determine
this, a Reliable Change Index (RC) is calculated. Jacobson
et al. describe this index as similar to one outlined earlier
by Nunnally and Kotsche (1983). In essence, an individual's
RC is calculated by comparing the individual's raw gain score
with the standard error of measurement, where the standard
error is determined using the test-retest reliability
coefficient of the instrument (here, the SCL-90-R).

If the individual's RC is statistically significant at
the 0.05 level, Jacobson et a1. consider that individual's
amount of change in therapy to be statistieally significant.
The second step of the attained health calculation is the
determination of whether or not the individual's change is
also clinieally significant. For change to be clinically
significant, the individual's post-therapy test score (here,
the GSI) must be more likely to fall in the distribution of
functional (here, ”healthy') individuals than in the
distribution of dysfunctional individuals.

The definitions of functional and dysfunctional status
are determined using normative data for the measurement
instrument employed. In the present study, these data were
taken from the SCL-90-R manual (Derogatis, 1977). Table 20
(p. 29) lists the mean and standard deviation of the GSI.
scores for 1,002 "heterogeneous psychiatric outpatients” (the
dysfunctional norm) and 974 "non-patient normals“ (the
functional norm). Jacobson et al.'provide formulas for using

the above descriptive statistics to determine an individual's

46
likelihood of membership in each normative group.

If an individual has both changed reliably and ended
therapy in a functional (“healthy") state, he or she was
considered to have a positive attained health score (1). If
the individual failed either of these tests, a negative

attained health score (0) was assigned.

Statistical Analyses

Most analyses were either partially or entirely
performed using the Statistieal Raetage Re; The Seeial
Seieneee (SEES) (Hull & Nie, 1981; Nie, Hull, Jenkins,
Steinbrenner, & Bent, 1975) in conjunction with the Michigan
State University Cyber-750 computer. Manual calculations
were carried out for procedures not available in SPSS. For
guidance in the use of psychometric methods, the standard
texts of Nunnally (1978) and Glass and Stanley (1970) were
consulted as necessary.

The following data analyses were employed. Descriptive
statistics were calculated for demographic data, nature of
initial and final symptoms, process variables, and outcome
variables (with the previously noted exception of residual
gain). One-way analyses of variance and trtests were
employed to test the statistical significance of selected
differences between group means.

Strength of relationship between variables was measured
by Pearson product-moment correlation coefficients.

Correlation coefficients involving residual gain were

 

47
calculated according to the method outlined by Cronbach and
Furby (1970). Standard two-tailed significance tests were
employed to test the statistical significance of correlation
coefficients and of differences between correlation

coefficients.

Subjects

Snares

The subject sample consisted of 77 clients of the
Michigan State University Psychological Clinic. The Clinic‘
serves a small Midwestern city and adjacent rural areas. The
work of the clinic is limited to outpatient services, and
individuals seeking treatment there are generally non-
psychotic.

W

Subjects' ages ranged from 16 through 91, with a mean of
30.1 and a standard deviation of 10.8. Age information was
not provided by 4 (5.2%) of the subjects. The sample
consisted of 23 (29.9%) subjects identifying themselves as
male, 50 (64.9%) identifying themselves as female, and 4
(5.2%) who did not report their gender.

Subjects' gross annual incomes ranged from $1,000 to
$78,000, with a mean of $13,044 and a standard deviation of
$12,522. Data on income were not available for 9 (11.7%)
subjects. The average educational level was 14.6 years, with
a standard deviation of 3 years. Data were not available for

5 (6.5%) subjects.

48
Lsxchenathelsmr

Table 1 presents pre—therapy and post-therapy
descriptive statistics for each of the nine symptom
dimensions of the SCL—90—R, as well as for the Global
Severity Index (GSI). As noted previously, the value scale
for each dimensions is: 0 = "Not at all”, 1 = “A little
bit", 2 = "Moderately", 3 = ”Quite a bit", and 4 =
”Extremely". Pre-therapy means for symptom dimensions range
from 0.70 (Phobic Anxiety) to 2.07 (Depression). The GSI
pre-therapy mean of 1.37 lies approximately in the middle of
this range, as one would expect given the fact that the GSI
is the average of all SCL-90-R item responses.

Post-therapy symptom dimension means vary from 0.41
(Phobic Anxiety) to 1.27 (Depression). Thus, at both pre-
therapy and post-therapy, the same two dimensions occupy the
lowest (Phobic Anxiety) and the highest (Depression)
positions in the range of dimensional means. And, like the
pre-therapy GSI mean, the post therapy GSI mean of 0.88 lies
roughly in the middle of the range.

Therapistm

Unfortunately, detailed therapist data are not
available. Descriptive information has apparently been lost
in the process of record retention and transportation. It
is, however, possible to provide a general description of
therapist characteristics, based on the MSU Psychological
Clinic's policies and procedures.

Almost all of the therapists were advanced graduate

Table 1:

49

Pre—therapy and post—therapy means and standard

deviations for the nine symptom dimensions and the Global
Severity Index (GSI) of the SCL—90-R.

______MEASQBE*
Global Severity Index

Somatization
Obsessive-Compulsive

Interpers. Sensitivity

Depression
Anxiety
Hostility
Phobic Anxiety
Paranoia

Psychoticism

PRE-THERAPY

“gan_____£4n4
1.37

0.60
0.77

0.66
1.65

0.82
1.69

0.83
2.07

0.84
1.62

0.95

1.15

0.85
0.70

0.74**
1.18

0.81
0.99

0.70

POST-THERAPY

Mean____5121
0.88

0.63
0.47

0.55**
0.99

0.79
1.14

0.78
1.27

0.89
1.06

0.80
0.89

0.86
0.41
0.90

0.76
0.60

0.65

 

*Value Meaning Key: 0
2 = I'Moderately", 3

"Not at all', 1

'A little bit',
“Quite a bit“, 4 = 'Extremely'

**The standard deviation is larger than the mean because SPSS
employs the "unbiased" estimate, 8, which inflates the
actual sample standard deviation by using N-l (rather

than N) in the calculation.

50
students in clinical psychology, working either as practicum
students or pre-doctoral interns. A very small number of
therapists wered post-doctoral psychologists, serving on the
faculty of the Department of Psychology. Because the
department generally admits roughly equal numbers of male and
female students to the clinical psychology graduate program,
it is probable that the number of male and female therapists
was approximately the same. Most therapists would have been
under the age of 35.

Because the theoretical orientation of the Psychology
Department is for the most part psychodynamic, it is
reasonable to expect that most of the therapists operated
within that general framework. However, cognitive-behavioral
and client-centered perspectives are also represented in the
graduate program, and it is likely that a few of the
therapists adhered to these orientations.

Nnmnet Qt Therapy Sessinns

The mean number of therapy sessions was 22, with a
standard deviation of 16. Over half of the subject sample
(43 subjects; 56%) attended from 10 to 30 sessions. Of the
remaining subjects, 16 (21%) attended less than 10 sessions,
and 16 (21%) attended from 31 to 48 sessions. Thus, 75 (98%)
of the 77 subjects attended less than 50 sessions of
psychotherapy.

The remaining 2 subjects attended a considerably higher
number of sessions (71 and 81, respectively). These 2

extreme values tend to inflate the size of the sample mean

51
(22 sessions). Because of this, the median and the mode
(both 18 sessions) provide somewhat better measures of

central tendency.

Procedure

Clients volunteered to participate in ongoing clinic
research, and service provision was not contingent on
research participation. At intake, clients completed
numerous forms, including the SCL-90—R. Upon termination,
they completed other documents, including the SCL-90—R and
the Strupp Post-Therapy Client Questionnaire. Subjects for
the present study were selected on the basis of completeness

of data.

RESULTS

Process Variables
E . I' S! I' I'

For each of the four process variables, Table 2 presents
the raw frequency of, and the precentage of the sample
endorsing, each response alternative. Also presented are the
mean and standard deviation of each variable. Inspection of
the table reveals that over 75% of the subjects agreed either
"strongly” or "mildly" that their therapists were warm,
respecting, accepting, and interested. Thus restriction of
range may be a factor in reducing the strength of process-
outcome relationships.

Hmthesiela

This hypothesis states that interrelationships among
process variables will not be significantly different from
each other. Table 3 presents the correlation matrix of
process variables. Significance testing revealed that the
hypothesis was for the most part confirmed. The exception is
that Interest is more strongly correlated with Warmth (z_=
2.24, p<.05) and Respect (z,= 3.35, p<.001) than with
Acceptance.

Hypothesis 1b also concerns the interrelationships of

52

53

Table 2: Part A: raw frequency of, and percentage of sample
endorsing, each response alternative, for each process
variable. Part B: mean and standard deviation, for each
process variable. (N = 77)

PROCESS VARIABLE
RESPONSE RESPONSE THERAPIST THERAPIST THERAPIST THERAPIST
NUMBER MEANING __HARM___ RESREQIINQ AQQRRIIRG INTERESTED
ERRQ. .1. EREQ _i_. FREQ. _i_ FREQ .1.

 

 

Eatt_A1
1 Strongly 36 55 44 58
Agree 46.8 71.4 57.1 75.3
3 Mildly 23 12 21 8
Agree 29.9 15.6 27.3 10.4
5 Undecided 6 6 4 6
7.8 7.8 5.2 7.8
7 Mildly 6 0 5 2
Disagree 7.8 0.0 6.5 °2.6
9 Strongly 3 1 2 0
Disagree 3.9 1.3 2.6 0.0
Missing -------- 3 3 1 3
‘Values 3.9 3.9 1.3 3.9
.Eart_B1

—————————————————— MEAN_5121 MEAN sin. MEAN_§121 MEAN—SAD;

2.6 1.7 2.4 1.6
2.3 1.5 2.0 1.5

Table 3: Process variable correlation matrix.#

WARMTH RESPECT ACCEPTANCE INTEREST

'WARMTH 1.000 .673 .563 .673
RESPECT .673 1.000 .614 .747
ACCEPTANCE .563 .614 1.000 .487
INTEREST .673 .747 .487 1.000

 

5311.correlation.coefficients are significant, p < .001.

54
process variables. Because discussion of this hypothesis
involves consideration of process-outcome relationships, that

discussion will be taken up when Hypothesis 3 is considered.

Outcome Variables

Table 4 presents frequency and percentage data for rated
gain. As noted earlier, rated gain is measured by the
subject's response to the question: ”To what extent have
your complaints or symptoms that brought you to therapy
changed as a result of treatment?" No subjects felt that
their difficulties had ”completely disappeared." Most
subjects (66, or 85.8%), however, did feel that they had
improved to some extent. Of theSe, 24 (31.2%) reported
feeling "very greatly improved," 22 (28.6%) felt
"considerably improved,” and 20 (26.0%) described themselves
as ”somewhat improved."

Of the 10 (13.0%) subjects who did not report
improvement, 3 (3.9%) judged that they had not changed at all
and 7 (9.1%) found themselves ”worse" than before therapy had
begun. Information on rated gain was not available for 1
(1.3%) subject.

Table 5 presents frequency and percentage data, as well
as the mean and standard deviation, for raw gain. As noted,
raw gain is obtained by subtracting the post-therapy SCL-90-R
Global Severity Index (GSI) from the pre-therapy GSI. Since

the GSI has a potential value range from 0 ("Not at all") to

55
4 ("Extremely"), raw gain has a potential value range from
-4.00 to 4.00. A positive raw gain score indicates reduction
in symptom severity, because the post-therapy GSI must be
lower than the pre-therapy GSI in order to obtain the
positive score. It follows, of course, that a negative raw
gain score indicates an increase in symptom severity.

Two brief examples may help render the meaning of the
raw gain score more concrete. If a subject's pre-therapy GSI
is 3 (”Quite a bit"), and the subject's post-therapy score is
4 ("Extremely"), then the subject's raw gain is -1.00. This
negative raw gain score indicates that the subject's overall
symptom severity has worsened by 1 unit on the 4-unit (0 to
4) GSI scale. On the other hand, suppose another subject's
pre-therapy GSI is also 3 ("Quite a bit”), but the post-
therapy GSI is 2 ("Moderately”). The raw gain score is then
1.00, indicating that the subject's overall symptomatic
picture has improved by 1 unit on the 4-unit GSI scale.

With the meaning of raw gain scores clarified, it is
appropriate to return to an overview of the data presented in
Table 5. At the outset, it is important to note that
individual raw gain scores actually range from -0.851 to
1.977, with a mean of 0.49 and a standard deviation of 0.60.
In order to simplify visual inspection of frequency and
percentage data, in Table 5 scores have been grouped into
value ranges with a width of 0.500 units. Deterioration up
to -l.000 unit was evident in 16 (20.8%) subjects. The

majority of subjects (46, or 59.8%) improved by up to 1.000

56
Table 4: Raw frequency of, and percentage of sample
endorsing, each response alternative for rated gain (N = 76,
mean = 303' S.D. = 1.2).

RESPONSE RESPONSE ________BAIEQ__QAIN_________

_NHMRER_. _MEARINQ. EREQHENCX EERCENIAGE
1 Completely 0 0.0
Disappeared
2 Very Greatly 24 31.2
Improved
3 Considerably 22 28.6
Improved
4 Somewhat 20 26.0
Improved
5 Not At All 3 3.9
Improved
6 Got Worse 7 9.1
Missing ———————————— l 1.3

 

Table 5: Raw frequency of, and percentage of sample
corresponding to, each value range for raw gain (N = 77,
Mean = 0.49, S.D. = 0.60).

 

BAN GAIN
____¥ALHE___RANQE___. EREQHRNQX RERQENIAGE
-l.000 through -0.500 3 3.9
-0.499 through 0.000 13 16.9
0.001 through 0.050 24 31.2
0.051 through 1.000 22 28.6
1.001 through 1.500 11 14.3

1.501 through 2.000 4 5.2

 

57
unit. Improvement between 1.000 and 2.000 units was apparent
for the remaining 15 (19.5%) subjects.

As noted earlier, frequency and percentage data for the
next outcome variable - residual gain - will not be
presented. Given the arguments of Cronbach and Furby (1970),
such data would be of no real value for analytical purposes.

The alternative to residual gain - the alternate
grouping method - involves several rather complex issues that
set it apart from the other outcome measurement approaches.
Thus the alternate grouping method will be considered
separately, after the other outcome and process-outcome
issues have been addressed.

The only remaining outcome variable to consider is
attained health. As previously indicated, this is a
dichotomous variable with possible values of 0 and 1. A
subject receives a score of 1 if reliable change has occurred
and a functional ("healthy") state has been attained. A
score of 0 indicates that change was not great enough to
considered reliable and/or a functional post—therapy state
was not achieved. Only 19 (24.7%) of the subjects received a
score of 1. Over three-quarters of the subjects (58, or
75.3%) received a score of 0.

Hxndthesis 2

Table 6 presents the outcome variable correlation
matrix. Before proceeding, it should be noted Cronbach and
Furby's (1970) method does not allow for computation of a

meaningful correlation between residual gain and raw gain.

58
Lack of such a correlation poses statistical problems when
one wishes to test the significance of the difference between
the correlations of raw gain and residual gain with some
third variable. In such cases, visual inspection has been
employed to determine the probability that a given difference
between correlations is significant.

Part a of Hypothesis 2 states that rated gain will
relate more strongly to residual gain than to raw gain. The
assumption here was that all correlations would have a
positive direction. Thus the hypothesis was confirmed, in
the sense that rated gain has a stronger peeitiye
relationship with residual gain than with raw gain. The
correlation of rated gain is .344, whereas the correlation of
rated gain with residual gain is -.280.

Hypothesis 2b states that attained health will relate
most strongly to rated gain, less strongly to residual gain,
and least strongly to raw gain. This part of Hypothesis 2
was clearly net confirmed. The correlation of attained
health with residual gain (.642) is stronger than the
correlation of attained health with rated gain (-.094) (z_=
6.80, p<.001). Likewise, the correlation of attained health
with raw gain (.521) is stronger than the correlation of
attained health with rated gain (-.094) (z_= 3.93, p<.001).
Visual inspection suggests that the correlation of attained
health with residual gain (.642) is not significantly
different from the correlation of attained health with raw

gain (.521).

59

Table 6: Outcome variable correlation matrix.

RATED RESIDUAL RAW ATTAINED
GAIN GAIN GAIN HEALTH
RATED 1.000 .344** -.280* -.094
GAIN
RESIDUAL .344** 1.000 # .642***
GAIN
RAW -.280* i 1.000 .521***
GAIN
ATTAINED -.094 .642*** .521*** 1.000
HEALTH

 

#Not accurately computable.
*Significant, p < .05
**Significant, p < .01
***Significant, p < .001

Table 7: Process-outcome variable correlation matrix.

 

 

OUTCOME PRQQES&______yARIABLES _
VARIABLE NARMTH RESPECT ACQERTANCE INTEREST, AYERAGE
Rated .439*** .432*** .401*** .415*** .499***
Gain

Residual .214 .224 .018 .164 .179
Gain

Raw .174 .184 .014 .133 .146
Gain

Health

 

***Significant,_p < .001

60

Process-Outcome Relationships

ametheaialb

Hypothesis lb states that, when any one outcome variable
is considered separately, there will be no significant
difference in strength of relationship among the four process
variables and outcome. This hypothesis was confirmed, with
one minor exception. The correlation of residual gain with
respect (.224) is significantly stronger than the correlation
of residual gain with acceptance (.018), z_= 2.104, p<.05.
Hypothesis la

Hypothesis 3a states that outcome measures will be
ordered from strongest to weakest relationship with process
measures, as follows: rated gain, attained health, residual
gain, and raw gain. This hypothesis was not confirmed.

In discussing research results, we will first employ
visual inspection and then refer to statistical tests which
determine the significance of apparent differences between
correlations. Visual inspection suggests that attained
health has a much different relationship to process variables
than that which was predicted. Rather than having the second
strongest relationship (after rated gain) with process
variables, attained health has the weakest relationship with
these variables.

Turning to the triad of rated, residual, and raw gain,
one finds their relationships with process variables to be in

the hierarchical order hypothesized. Process-outcome

61
correlations involving rated gain are strongest, with those
involving residual gain being somewhat weaker, and
correlations involving raw gain being weakest of all.

When tests are employed to determine the statistical
significance of visually apparent differences, the picture
becomes considerably more complex. The analytical strategy
employed here is to consider each process variable separately
and compare its correlation with the different outcome
variables. In other words, we will examine Table 7 one
column at a time. Table 8 supports Table 7 by presenting
significance test data for all pairs of process-outcome
correlations which are significantly different in one or more
cases. The zrvalue and significance level are indicated for
cases where significant differences exist.

Part A of Table 8 compares correlations which are
vertically adjacent to each other in each column of Table 7.
The correlation of rated gain with process is significantly
stronger than the correlation of residual gain with process,
for the following process variables: acceptance, interest,
and the process variable average. In no case is residual
gain related more strongly than raw gain to process. Raw
gain is related more strongly than attained health to process
in the case of warmth only.

Part B of Table 8 compares correlations which are not
vertically adjacent in Table 7. The question addressed by
the first entry in Part B is: in cases where rated gain is

not related more strongly than neeidnal gain to process, is

62

Table 8: arvalue and significance level for significant
differences between members of selected pairs of process-
outcome correlations. Part A: correlations vertically
adjacent to each other in Table 7. Part B: correlations not
vertically adjacent to each other in Table 7.

 

OUTCOME PROCESS VARIABLES _
_¥ARIARLES_ EARMTH, RESRECT. ACQERTANQE INTEREST, ATERAGE
W:

Rated Gain # t 3.16** 2.07* 2.73**

13.

Resid. Gain
Resid. Gain 8 i t # #

1&0
Raw Gain
Raw Gain 2.16* i t # i

la.

Attnd. Hlth.

 

Ratt_Et

Rated Gain 2.11* 1.98* 3.05** 2.23* 2.88**

25o
Raw Gain

Resid. Gain 2.95** t # 2.14* 2.12*

28,.
Attnd. Hlth.

 

# No significant difference
* Significant, p < .05 ’
** Significant, p < .01

Table 9: The Alternate Grouping Method subgroup structure,
including the mean and standard deviation of the pre-therapy
Global Severity Index (GSI) for each subgroup and for the
entire sample.

.ENTITX. __DER1NITIQN__ SHHJEETS .1 HEARIHEBAEXEEEE
Mild No Scales > 2 24 31 0.74 0.28
Moderate 1-2 Scales > 2 29 38 1.35 0.25
Severe 3-9 Scales > 2 24 31 2.02 0.44

 

Entire Whole Sample 77 100 1.37 0.60

 

rat

The

re
st

V6

91

me

tt

63
rated gain related more strongly than tau gain to process?
The table indicates an affirmative answer to this question.
Although rated gain is not more strongly related than
residual gain to warmth and respect, rated gain is more
strongly related than raw gain to both of these process
variables.

The question addressed by the second entry in Part B is:
given that residual gain is not related more strongly than
tan_gain to process in any case, is residual gain related
more strongly than attained health to process? The answer to
this question is mixed. For warmth, interest, and the
process variable average, residual gain is related more
strongly than attained health to process. However, for
respect and acceptance, residual gain is net related more
strongly than attained health to process.
Wanandw

Hypotheses 4a and 4b state that all process—outcome
relationships (excluding some in the alternate grouping
method, to be discussed later) will be statistically
significant (4b) and that all statistically significant
correlations will be positive (4a). Table 7 presents the
relevant data.

Hypothesis 4b was not confirmed. Only the correlations
of rated gain with process variables are significant. These
correlations range from .401 (acceptance) to .499 (process
variable average). All are significant at the .001 level.

Hypothesis 4a was, however, confirmed. Correlations of rated

ga
'al

rn

 

64
gain with process are the only significant correlations, and
‘all of these correlations are positive.
TheAlternateGmninsMethod

As noted previously, the alternate grouping method is
intended to control for the relationship of pre—therapy
symptom severity with the amount of change in symptom
severity which occurs during psychotherapy. In the present
study the correlation of pre-therapy SCL-90-R symptom
dimension severity with amount of symptom change ranges from
.458 (Depression) to .659 (Somatization). The correlation of
the pre-therapy SCL-90-R Global Severity Index (GSI) with
amoung of change in GSI is .450. All of these correlations
are statistically significant at the .001 level. The
existence of such strong relationships between pre-therapy
symptom severity and amount of symptom change certainly
justifies effort to control for this relationship.

The strategy for control employed in the alternate
grouping method, outlined in general earlier, was as follows
(see Table 9). The first phase involved trichotomizing the
sample into "mild", "moderate”, and "severe“ subgroups.
Assignment of a subject to a subgroup was determined by how
many of the nine SCL-90-R symptom dimensions were "high"
(with a score greater than 2) for that subject. In order to
achieve roughly equal subgroup sizes, the three groups were
defined as follows. The mild subgroup consists of subjects
for whom none of the nine symptom dimensions are high. When

this subgroup assignment scheme was followed, the mild

65
subgroup contained 24 (31%) subjects, the moderate subgroup
29 (38%) subjects, and the severe subgroup 24 (31%) subjects.

Table 9 summarizes subgroup structure and also presents
the mean and standard deviation of the pre-therapy GSI for
each subgroup and the entire sample. The GSI mean of the
moderate subgroup (1.35) is significantly greater than the
mean of the mild subgroup (0.74), t151) = 8.39, p<.001.
Likewise, the mean of the severe subgroup (2.02) is
significantly greater than that of the moderate subgroup
(1.35), tlSl) = 7.02, p<.001.

Table 10 presents the mean and standard deviation for
each of the demographic variables, for each subgroup and for
the entire sample. One—way analyses of variance reveal no
significant differences in demographic variables across
subgroups, with the exception of age, 212,70) = 4.48, p<.05.
The mean of the moderate subgroup (34.5) is significantly
larger than the mean of the mild subgroup (26.4), t149) =
2.47, p<.05. There is, however, no significant difference
between the means of the moderate (34.5) and severe (28.1)
subgroups or between the means of the mild (26.4) and severe
(28.1) subgroups. Thus, for the most part, the groups can be
considered demographically homogeneous.

Having considered the structure of the subgroups, we may
now turn to the research hypotheses which relate to the
alternate grouping method. Table 11 presents the relevant
process-outCome correlations and their levels of statistical

significance (i.e., correlations of raw gain with process

66

Table 10: Mean and standard deviation for demographic
variables, for each subgroup of the Alternate Grouping Method
and for the entire sample, with one-way analysis of variance
across subgroups for each variable.

___SEX£___ ___AQE____ EDEQATLQN- .1NQQME___
_ENT1TX_ MEAN Sent MEAN Sine, MEAN Sent MEAN STD;

 

 

Mild 1.6 0.5 26.4 3.6 14.8 2.6 12.0 9.6
Moderate 1.7 0.5 34.5 15.0 17.1 12.7 15.3 16.2
Severe 1.8 0.4 28.1 6.3 14.2 3.1 11.3 9.7
Entire 1.7 0.5 30.1 10.8 15.5 8.3 13.0 12.5
312.70) 0.56 4.48* 0.85 0.67

 

# l = Male, 2 = Female
* Significant, p < .05

Table 11: Process-outcome correlations for the Alternate
Grouping Method (i.e., correlation of raw gain with process
variables, for each subgroup and for the entire sample).

 

 

 

BREESS YABIABLES
_ENTITX_ WARMTH RESREQT. AQQEETANQE INTEREST AIERAGE
Mild .039 -.051 .134 .085 .057
Moderate -.048 .229* .160 .198 .135
Severe .350** .314** -.028 .113 .214
Entire .174 .184 .014 .133 .146

 

* Significant, p < .05
** Significant, p < .01

67
variables, for each subgroup and for the entire sample).

Hypothesis 4c states that process-outcome relationships
will be statistically significant within the severe and
'moderate subgroups but not within the mild subgroup. This
hypothesis was partially confirmed, in that process-outcome
correlations were not significant within the mild subgroup.
However, the situation was not as predicted in the moderate
and severe subgroups. In the moderate subgroup, only respect
was significantly correlated with outcome (.229). In the
severe subgroup, only warmth (.350) and respect (.314) were
significantly correlated with outcome.

Hypothesis 4a states that all statistically significant
process-outcome relationships will be positive. This
hypothesis was confirmed for the alternate grouping method,
just as it was for the other process-outcome relationships
previously considered.

Hypothesis 3b concerns the hierarchical order of
process-outcome relationship strength for the alternate
grouping method. Part (1) states that the relationship
between process variables and raw gain will be significantly
stronger within the severe and moderate subgroups than in the
subject sample as a whole. This prediction was not
confirmed. Significance testing reveals that, for each of
the process variables, neither the severe nor the moderate
subgroup produced process-outcome correlations significantly
stronger than the process-outcome correlation for the entire

sample.

68
Part (2) of Hypothesis 3b states that the relationship
between process variables and raw gain will net be
significantly stronger within the mild subgroup than in the

subject sample as a whole. Significance testing confirms

this prediction.

DISCUSSION

Process Variables

The four process variables — warmth, respect,
acceptance, and interest - appear to form an essentially
unified construct, which we may justifiably describe under
the single label of positive regard. Two forms of evidence
support this conclusion. First, process variable
interrelationships are for the most part not significantly
different. Second, when any one outcome variable is
considered separately, there is generally no significant
difference in strength of relationship among the four process
variables and outcome. This evidence reinforces the validity
of the practice, evident in earlier studies (e.g., Rogers,
1957; Truax & Mitchell, 1971), of referring to such

dimensions as key components of positive regard.

Outcome Variables

Rated gain relates more strongly to residual gain than
to raw gain. This result confirms the earlier findings of
Green et a1. (1975) and Mintz (1979). As noted previously,
their results indicated that rated gain was more highly
related to post-therapy adjustment level than to amount of

69

70
change occurring in therapy (i.e., raw gain). When raw gain
scores were residualized to correct for the effect of pre-
therapy symptom severity, the correlation of gain score with
rated gain increased. Given the similar finding in the
present study, it appears reasonable to conclude, with Mintz
(1979), that final adjustment status is given more weight
when raw gain scores are residualized.

Beyond this, there is an interesting practical
implication that might be drawn from the existence in the
present study of a significant negative relationship between
rated gain and raw gain. People who change the most in
therapy may not make the most favorable global evaluations of
psychotherapeutic efficacy.

In the case of attained health, the results of the
present study were quite different from what was expected.
Attained health is not significantly related to rated gain.
If rated gain and attained health both reflect final
adjustment status to a considerable degree, how can it be
that rated gain and attained health are not significantly
related to each other? Answering this question requires
reference to arguments stated earlier, regarding differences
in the determination of rated gain and the attained health
score. First, attained health is derived from multiple
specific responses on a symptom checklist, whereas rated gain
is simply a single global outcome estimate. Second, attained
health involves reference to a normative sample, in the

definition of "health." When the subject estimates rated

gai
alt
oc<
maj
f '1
di

01

71
gain, no formal reference to a normative group is made,
although informal comparison with other persons probably
occurs. Thus, although both rated gain and attained health
may be legitimately viewed as giving considerable weight to
final adjustment status, they measure this status in somewhat
different ways.

When hypotheses were being formed, the effect of this
difference was underestimated. We held earlier that the
difference would only reduce the size of the correlation
between rated gain and attained health to a limited degree,
so that attained health would still be more strongly
correlated with rated gain than with residual or raw gain.
In fact, attained health is more highly correlated with both
residual and raw gain than with rated gain.

The relatively high correlations of attained health with
residual and raw gain may be due at least in part to the
specific ways in which all three variables are computed from
the same psychometric instrument, the SCL—90-R. Whereas
residual gain gives considerable weight to post—test symptom
severity, raw gain reflects the actual amount of symptom
change. The similarity in the strengths of the correlations
of residual and raw gain with attained health suggests that
both post-test symptom severity and amount of symptom change
make substantial contributions to the attained health score.
This situation is roughly what one might expect, given the
fact that calculation of the attained health score involves

both determination of post-test status (i.e., "health") and

72
determination of amount of change (i.e., change must be above
a certain threshold to be considered reliable).

One final pattern of outcome variable interrelationships
deserves further attention. We have seen that rated gain and
attained health are not significantly related to each other.
This was explained by speculating that, although both
variables are strongly related to final adjustment status,
they construe final status in quite divergent ways. Now we
must add to the discussion the fact that, despite their lack
of a significant positive relationship with each other, both
rated gain and attained health have a significant positive
relationship with residual gain. In this connection, it is
important to note the earlier suggestion that final
adjustment status plays an important role in the
determination of residual gain. Thus it seems reasonable to
speculate that residual gain is to some extent determined
heth by final adjustment status as construed in rated gain
and by the considerably different conception of final status

employed in attained health.

Process-Outcome Relationships

We now turn to the heart of the present study —
consideration of process—outcome relationships. The central
purpdse of this study was to determine how the relationship
between positive regard and psychotherapy outcome is affected
by the type of outcome measure used. As noted earlier,

relevant prior research has not directly compared the

73
relative strength of relationships between positive regard
and the various outcome measures employed here. Thus the
present study breaks new ground in an important area of
investigation.

When the most salient results are considered, the
following picture emerges. Rated gain is clearly more
strongly related to positive regard than is raw gain. This
result supports a central assumption of the present study -
namely, that when self-report measures are employed, process
is more strongly related to perceived final adjustment status
than to amount of pre-post change.

Contrary to expectations, residualizing gain does not
make the process-outcome relationship stronger than it is in
the case of raw gain. The other means of controlling for
pre-therapy symptom severity - the alternate grouping method
- likewise does not result in a significant increase in
strength of process-outcome relationships. Some implications
relevant to these results will be considered shortly, when
the significance and direction of process—outcome
relationships are addressed.

Attained health behaves in a surprising manner. We had
originally expected attained health to be strongly related to
rated gain but, as noted, this is not the case. In examining
this finding, we concluded that final adjustment status was
conceived differently in rated gain than in attained health.
It is now evident that process variables are related much

more strongly to rated gain than to attained health. This

74
pattern of process-outcome relationships suggests that the
process variables relate to final adjustment status as
measured by rated gain more than to final adjustment status
as measured by attained health. This may be due at least in
part to the fact that both the process variables and rated
gain involve personal, global judgments, whereas attained
health involves comparison with normative psychometric data.

We turn now to consider the significance and direction
of process-outcome relationships. Rated gain is
significantly related to all positive regard variables, and
these relationships are uniformly positive. This finding is
in concert with the results of all eight of the Post-Therapy
Process (PTP) studies reviewed earlier. Raw gain was not
significantly correlated with process. This finding is not
pariticularly surprising, since the only previous study cited
in connection with raw gain (Gross & DeRidder, 1966) had
found nonsignificant as well as significant relationships
between process and raw gain.

The lack of a significant relationship between residual
gain and positive regard is a bit more unexpected. Both
studies cited earlier in this connection (Luborsky et al.,
1985; Marziali, 1984) had found significant positive
relationships. However, as previously noted, these two
studies involved the use of mid-therapy measures of process,
rather than post-therapy measures such as those used in the
present study. Also, their process measures involved

"therapeutic alliance” dimensions, rather than positive

75
regard alone. Consequently, the limited comparability of
design in prior research may help to explain the divergence
of the present study's results from those of earlier studies.

In any case, it appears that the influence of pre-
therapy symptom severity is not alone sufficient to account
for the lack of a significant relationship between positive
regard and raw gain. If it were sufficient, then residual
gain would be significantly correlated with positive regard.

When attained health is considered, it is important to
recall that the recency of the method's publication rendered
its use in the present study purely exploratory. No relevant
previous research appears to have employed attained health as
a psychotherapy outcome measure. In the present study, no
significant relationships emerged between attained health and
positive regard. One factor contributing to this result may
be the stringency of the attained health criterion for
therapeutic success. Only 19 (24.7%) of the subjects
successfully met this criterion.

In any case, further research will be necessary to
explore in more detail the nature of the attained health
variable and its relationship with psychotherapeutic process.
It may be that future research, employing different subject
and therapist samples, will find significant process-outcome
relationships.

It remains only to consider briefly the alternate
grouping method. As expected, within the mild subgroup,

process-outcome relationships were not significant. However,

76
the lack of consistently significant relationship within the
moderate and severe subgroups runs counter to expectations.
In interpreting this finding, it is important to recall that
the earlier study cited in relation to the alternate grouping
method (Barrett-Lennard, 1962) did not define or apply the
method in the same detail as was done in the present studyn
Thus differences in the two studies' results may be due in
part to design variations.

On the other hand, perhaps the alternate grouping method
is simply not an effective control for the influence of
pretest symptom severity on the size of raw gain scores. In
considering this possibility, however, we must take into
account the fact that residual gain - the other method of
controlling for pretest symptom severity - is also not
significantly correlated with positive regard. It may be
that the alternate grouping method does not yield significant
process-outcome relationships because, as previously
suggested, the influence of pre-therapy symptom severity is
not by itself sufficient to account for the lack of a
significant relationship between positive regard and raw
gain.

Also important is the issue of sample size. When the
present study's sample was trichotomized, the subgroups
ranged in size from 24 to 29 subjects. With groups of this
size, the potential for sampling error is considerable. Thus,
it may be that a truly legitimate test of the alternate

grouping method, as a control for pre-therapy symptom

77
severity, may have to await a study in which a considerably
larger sample is available. Nevertheless, the present study
makes an important contribution by elaborating specific
techniques for the application of this method in

psychotherapy research.

Conclusion

The present study has produced important evidence that
the relationship between positive regard and outcome in
psychotherapy does in fact vary significantly with the type
of outcome measure employed. Of course, the generalizability
of the study's findings is quite limited. Only self-report
measures were employed, and measurement of process was
exclusively retrospective. Furthermore, only positive regard
was addressed, and it is possible that other Rogerian process
variables would behave somewhat differently.

Within its limits, however, the present study
demonstrates an important point. Because of the complexity
of the issues involved, future progress in psychotherapy
research will require at least as much attention to the
measurement of treatment outcome as to the measurement of

therapeutic process.

LI ST OF REFERENCES

LI ST OF REFERENCES

Abeles, N. (1986, JulY)- Egg nsxnhnlnsx and
researnh. Paper presented at IAAP's let
International Congress of Applied Psychology,
Jerusalem, Israel.

Andrews, G., & Harvey, R. (1981). Does psychotherapy
benefit neurotic patients? A reanalysis of the
Smith, Glass and Miller data. Arehiyes of General
ESXChiatryl 18, 1203- 1208

Barrett- -Lennard, G. T. (1962). Dimensions of therapist
response as causal factors in therapeutic change.

Esrahnlngisal Monographs 16, (43, Whole No. 562).

Bent, R. J., Putnam, D. G., & Kiesler, D. J. (1976).
Correlates of successful and unsuccessful
Psychotherapy- Jnnrnal.0f Consulting and Clinieal
Eexfihnlngxy 44, 149.

Bergin, A. E., & Lambert, M. J. (1978). The evaluation of
therapeutic outcomes. In S. L. Garfield & A. E.
Bergin (Eds.), Handbook of nsxnhntherany and behaxinr
ghange (2nd ed., pp. 139-190). New York: Wiley.

Bergin, A. E., & Suinn, R. M. (1975). Individual

psychotherapy and behavior therapy. Annual Relieu Qt
Esxshnlngx, 25, 509-556.

Beutler, L. E., & Hamblin, D. L. (1986) Individualized
outcome measures of internal change: Methodological
considerations. lnnrnal of Consulting and Clinical
Eaxfihelngxl 51. 48- 53-

Board, F. A. (1959). Patients' and physicians' judgments
of outcome of psychotherapy in an outpatient clinic.
Archixes of General Esxehiatrx, 1, 185-196.

Brown. 0. H. (1954). An inrestigatinn of therapeutis
. . in. J' l- I 2 l I]
Unpublished doctoral dissertation, University of
Chicago.

78

79

Carmichael. J. A. (1970). Eeroention of self-as-ohiest
as a.oonsesuenoe of systematio desensitization and
elientzeenteted_therapy. Unpublished master's
thesis, University of Victoria, British Columbia,
Canada.

Cartwright, D. S., Kirtner, W. L., & Fiske, D. W. (1963).
Method factors in changes associated with

psychotherapy. Journal of Abnormal and Social
Bartholoux. 66, 164-175.

Chinsky, J. M., & Rappaport, J. (1970). Brief critique of
meaning and reliability of I'accurate empathy“
ratings. Bartholooisal Bulletin 13. 379- 382.

Cooley, E. J., & Lajoy, R. (1980). Therapeutic
relationship and improvement as perceived by clients

and therapists. Journal of.Clinioa1 Esraholoou, as.
562-570.

Cronbach, L. J., & Furby, L. (1970). How should we measure

change - or should we? Psroholouieal Bulletin 11,
68- 80.

Derogatis, L. R. (1983). Misuse of the Symptom Checklist 90
(Letter to the editor). Arehiges of General
Psxshiatrx 40, 1152- 1153.

DerogatiS. L. R. (1977). sonzso manual 1.: Administration,
Mprooedures for the reuised.2ersion.
Baltimore: Clinical Psychometrics Unit, Johns
Hopkins University School of Medicine.

Derogatis, L. R., & Cleary, P. A. (1977). Confirmation of
the dimensional structure of the SCL-90: A study in
construct validation. Jenrnal Qt Cenaulting and
Clinioal Bartholooy 33, 981- 989.

Derogatis, L. R., Klerman, G. L., & Lipman, R. S. (1972).
Anxiety states and depressive neuroses: Issues in
nosological discrimination. Jeurnal ef Nerxeus and
Mental Disease, 155, 392-403.

Derogatis, L. R., Lipman, R. S., & Covi, L. (1973). SCL90:
An outpatient psychiatric rating scale - Preliminary
report. Earshonharmaooloou Bulletin, 24 13-28.

Derogatis, L. R., Lipman, R. 8., Covi, L., & Rickels, K.
(1971). Neurotic symptom dimensions: As perceived
by psychiatrists and patients of various social

classes. Arohires of General 2§¥£hiatt¥ 294
454- 464.

 

80

Derogatis, L. R., Lipman, R. 8., Covi, L., & Rickels, K.
(1972). Factorial invariance of symptom dimensions
in anxious and depressive neuroses. Archiges of

General Esxchiatrx, 21, 659- 665.

Derogatis, L. R., Lipman, R. S., Covi, L., Rickels, R., &
Uhlenhuth, E. H. (1970). Dimensions of outpatient
neurotic pathology: Comparison of a clinical versus
an empirical assessment. Journal of Consolring and
Clinical.Bs¥holoc¥.: 11. 164-171.

Derogatis, L. R., Lipman, R. S., Rickels, R., Uhlenhuth, E.
B., & Covi, L. (1974). The Hopkins Symptom
Checklist (HSCL): A measure of primary symptom
dimensions. In P. Pichot (Ed. ), Psychological
measurement_in psychonharmacoloo¥= Yol1.1. Modern

in_oharmacoosrchiatr¥ (pp. 79-110).Base1,

Switzerland: S. Karger. (a)

Derogatis, L. R., Lipman, R. S., Rickels, R., Uhlenhuth, E.
H., & Covi, L. (1974). The Hopkins Symptom
Checklist (HSCL): A self-report inventory.
Behauioral Sciences.,12. 1- 15. (b)

Derogatis, L. R., Rickels, K., & Rock, A. F. (1976). The
SCL—90 and the MMPI: A step in the validation of a
new self-report scale. British Journal or
Batchiatrx 123. 280- 289.

Derogatis, L. R., Yevzeroff, B., & Wittelsberger, B. (1975).
Social class, psychological disorder, and the nature
of the psychopathologic indicator. Journal of

Consulting and Clinical Bsxcholoox, 13, 183- 191.

Feitel, B. s. (1968). Eeelino understood as a.function of
a.1ariet¥.ofi.theraoist actiuities. Unpublished
doctoral dissertation, Teachers College, Columbia
University.

Fiske, D. W. (1971). The shaky evidence is slowly put
together. Journalu uConsultino and Clinical
Efi¥£h9199¥ 11, 314-315.

Fiske, D. W., Hunt, H. R., Luborsky, L., Orne, M. T.,
Parloff, M. B., Reiser, M. F., & Tuma, A. H. (1970).
The planning of research on effectiveness of
psychotherapy. (Report of workshop sponsored and
supported by the Clinical Projects Research Review
Committee, National Institute of Mental Health).

Archiues of General Psychiatry, 22, 22- 32.

81

Ford, J. D. (1977). Process and outcome in behavior
therapy: An empirical analysis. Qissortation
Ahstracts International 18, 28588.

Frank, J. D. (1973). Earsuasion and hcaling. Baltimore,
MD: Johns Hopkins University Press.

Free, N. R., Green, B. L., Grace, M. C., Chernus, L. A., &
Whitman, R. M. (1985). Empathy and outcome in
brief focal dynamic therapy. American Journal of
Baxchiatrx. 142, 917-921.

Fretz, B. R. (1966). Postural movements in a counseling

g§§d543dournal of Counseling.Esxcholog¥. ll.

Frieswyk, S. E., Allen, J. G., Colson, D. B., Coyne, L,
Gabbard, G. 0., Horwitz, L., & Newsom, G. (1986)
Therapeutic alliance: Its place as a process and
outcome variable in dynamic psychotherapy research.
gourgal of Consulting_hfld_91inical_£sycholog¥1_54.__

Garfield, S. L., Prager, R. A., 8 Bergin, A. E. (1971).
Evaluation of outcome in psychotherapy. Journal
of Consulting and Clinical Psychology, 31, 307-313.

Glass, G. V., 8 Stanley, J. C. (1970). Statistical

methods in education.and RSXCthch. Englewood
Cliffs, NJ: Prentice—Hall.

Gomez-Schwartz, B. (1978). Effective ingredients in
psychotherapy: Prediction of outcome from process

variables. Journalu Consulting and Clinical
PsychologL 45. 1023- 1035.

Gomez—Schwartz, B., Hadley, S. W., & Strupp. H. H. (1978).
Individual psychotherapy and behavior therapy.

Annual Reuieud EsxchologL 22, 435-447.

Green, B. L., Gleser, G. C., Stone, W. N., & Seifert, R. F.
(1975). Relationships among diverse measures of
psychotherapy outcome. Journalfi Consulting
and Clinical Esxchology. 11, 689- 699.

Grigg, A. E., & Goodstein, L. D. (1957). The use of clients
as judges of the counselor' 5 performance. Journal of

Counseling Esxchologx 11 31- 36.

Gross, F., & DeRidder, L. M. (1966). Significant movement
in comparatively short-term counseling. Journal of

Counseling PsychologL 13, 98- 99.

82

Gurman, A. (1977L The patient' 8 perception of the
therapeutic relationship. In A. Gurman 8 A. Razin

(Eds. ), Effectice ns¥£hotheranL (PP. 503-543).
New York: Pergamon.

Hartley, D. E., 8 Strupp. H. H. (1983). The therapeutic
alliance: Its relationship to outcome in brief
psychotherapy. In M. Masling (Ed. ),
studiesfi

Empirical
nsxchoanalxtic theories (pp. 1- -27).
Hillsdale, NJ: Analytic Press.

Horowitz, M. J., Marmar, C., Weiss, B., DeWitt, K. N., 8
Rosenbaum, R. (1984). Brief psychotherapy of
bereavement reactions: The relationship of process

to outcome. Archiues of General BsxchiatrL
438-448.

Hull, C. H., 8 Nie, N. H. (1981). 5255 update 1:2: New
and facilities for releases 1:9. New
York: McGraw-Hill.

Jacobson, N. S., Follette, W. C., 8 Revenstorf, D. (1984).
Psychotherapy outcome research: Methods for
reporting variability and evaluating clinical

significance. Behauior,TheraoL 15, 336- 352.

Jones, E. (1968). The relationship of counselorzclient

personal1tx.similarit¥m counseling orocess.and
outcome. Unpublished doctoral dissertation,
University of Missouri.

Kass. F., Charles, E., Klein, D. F., 8 Cohen, P. (1983).
Discordance between the SCL-90 and therapists'
psychopathology ratings: Implications for clinical

assessment. Archiues of General PsychiatrL in.
389- 393. (a)

Kass, F., Charles, E., Klein, D. F., 8 Cohen, P. (1983).
In reply (Letter to the editor). Archiues of
General EsxchiatrL so. 1152-1153. (b)

Kiesler, D. J., Klein, M. H., Mathieu, R. L., 8 Schoeninger,
D. (1967). Constructive personality change for
therapy and control patients. In C. R. Rogers (Ed. ),
The therapeutic relationship and its imoact: A.stud¥
of psychotheraor uith.schizonhrenics. Madison:

University of Wisconsin Press.

Kurtz, R. R., 8 Grummon, D. L. (1972). Different approaches
to the measurement of therapist empathy and their
relationship to therapy outcomes. Journal of

Consulting and Clinical E§¥9thQQL 32. 106- 115.

83

Lambert, M. J., DeJulio, S. S., 8 Stein, D. (1978).

Therapist interpersonal skills. Efixflhfllflfliflfll
Bulletinl. B1, 467- 489.

Landman, J. T., 8 Dawes, R. M. (1982). Psychotherapy
outcome: Smith and Glass' conclusions stand up under

scrutiny. American Psychologist. 11. 504-516.

Lesser, W. M. (1961). The relationship between counseling
progress and empathic understanding. Journal of

Counseling Psychology.. 8. 330- 335.

Lorr, M. (1965). Client perceptions of therapists: A study
of therapeutic relation. Journal of Consulting
EEXQthQQXJ 22. 146-149.

Luborsky, L., Auerbach, A. H., Chandler, M., Cohen, J., 8
Bachrach, H. M. (1971). Factors influencing the
outcome of psychotherapy: A review of quantitative

research. Psychological Bulletin. 15. 145-185.

Luborsky, L., Crits-Christoph, P., Alexander, L., Margolis,
M., 8 Cohen, M. (1983). Two helping alliance
methods for predicting outcomes of psychotherapy: A
counting signs vs. a global rating method.
of Nerrous.and.Mental.Bisease. 111. 480-491.

Luborsky, L., McLellan, A. T., Woody, G. E., O'Brien, C. P.,
8 Auerbach, A. (1985). Therapist success and its

determinants. Archioes.of.Genera1.Bsxchiatrx. 12.
602-611.

Luborsky, L., Mintz, J., Auerbach, A., Christoph, P.,
Bachrach, H., Todd, T., Johnson, M., Cohen, M., 8
O'Brien, C. P. (1980). Predicting the outcome of
psychotherapy: Findings of the Penn Psychotherapy

Project. Archiues of General BatchiatrL
471- 481.

Manning, W. H., 8 DuBois, P. H. (1962). Correlational
methods in research on human learning. Perceptual
and Motor Skills. 15. 288-321.

Marziali, E. (1984). Three viewpoints on the therapeutic
alliance: similarities, differences, and
associations with psychotherapy outcome. Journal
of.Neruous and.Mental Disease. 112. 417-423.

84

McClanahan, L. D. (1974). A comparison of counseling
techniques and attitudes with client evaluations of
the counseling relationship. Dissertation Abstracts
International.. 11. 5637A.

Mendola, J. J. (1981L Therapist interpersonal skills: A
model for professional development and therapeutic

effectiveness. Dissertation Ahstracts International.
12, 42013.

Mintz, J. (1972). What is success in psychotherapy?
Journal of.Ahnorma1 Psychology, to. 11-19.

Mintz, J., Luborsky, L., 8 Christoph, P. (1979). Measuring
the outcomes of psychotherapy: Findings of the Penn

Psychotherapy Project. Journalfl Consulting and
Clinical Esxchology. 41. 319- 334.

Mitchell, K. M., Bozarth, J. D., 8 Krauft, C. C. (1977).
A reappraisal of the therapeutic effectiveness of
accurate empathy, nonpossessive warmth and
genuineness. In A. S. Gurman 8 A. M. Razin (Eds.),

EificC11!£.DS¥ChQLthaR¥.(pp. 482- -502). New York:

Pergamon.

Mitchell, K. M., Truax, C. B., Bozarth, J. D., 8 Krauft,

c. c. (1973). Antecedents to.osxchotheraneut1c
outcome. NIMH Grant Report (12306). Hot Springs,
AK: Arkansas Rehabilitation Services.

Morgan, R., Luborsky, L., Crits-Christoph, P., Curtis, H.,
8 Solomon, J. (1982). Predicting the outcomes of
psychotherapy by the Penn Helping Alliance Rating

Method. Archiues of General PsychiatrL 32. 397-402.

Nicholson, R. A., 8 Berman, J. S. (1983). Is follow-up

necessary in evaluating psychotherapy? ESYQthQQiflil
Bulletin. 21. 261-278.

Nie, N. H., Hull, C. H., Jenkins, J. G., Steinbrenner, K. 8
Bent. D. (1975). SPSS: Statistical packagerorM
the social sciences (2nd ed. ). New York:
McGraw-Hill.

Nunnally, J. C. (1978). Psychometric theory (2nd ed.).
New York: McGraw-Hill.

Nunnally, J. C., 8 Kotsche, W. E. (1983). Studies of
individual subjects: Logic and methods of analysis.

Egeggritish Journal of Clinical Bsxchologx. 22.

 

85

O'Malley, S. 5., Sub, C. S., 8 Strupp, H. H. (1983). The
Vanderbilt Psychotherapy Process Scale: A report on
the scale development and a process- -outcome study.

ggurggé of Consulting and Clinical BsxchologL 51.

Orlinsky, D. E., 8 Howard, K. I. (1978). The relation of
process to outcome in psychotherapy. In S. L.
Garfield 8 A. E. Bergin (Eds. ), Handhook of

psychotherao¥.and hehauior change (2nd ed.,
pp. 283-330). New York: Wiley.

Parloff, M. (1986). Placebo controls in psychotherapy
research. American Psychologist. 11. 79- 87.

Parloff, M. B., London, P., 8 Wolfe, B. (1986). Individual
psychotherapy and behavior change. Annual Reyiey of
Psychology. 11, 321-349.

Parloff, M., Waskow, I., 8 Wolfe, B. (1978). Therapist
variables in relation to process and outcome. In
S. Garfield 8 A. Bergin, Hanohook of

and hehauior change (2nd ed.. pp. 233-282).
New York: Wiley.

Patterson, C. H. (1984). Empathy, warmth, and genuineness
in psychotherapy: A review of reviews.
I 2.1,, 431-438.

Phillips, J. S., 8 Bierman, K. L. (1981). Clinical

psychology: Individual methods. Annual Reyieu of
EBYCthQQY. 32. 405-438.

Rachman, S. J. (1973). The effects of psychological
treatment. In H. Eysenck (Ed.), Handbook of
ahnormal psychology. New York: Basic Books.

Rogers, C. R. (1957). The necessary and sufficient
conditions of therapeutic personality change.

Journal of Consulting.BsxchologL 21, 95- 103.

Rogers, C. R., Gendlin, G. T., Kiesler, D. V., 8 Truax, C. B.
(1967). The therapeutic relationship and its

. Madison: University of Wisconsin
Press.

Ryan, V. L., 8 Gizynski, M. N. (1971). Behavior therapy in
retrospect: Patients' feelings about their behavior
therapies. Journal of Consulting and Clinical
Psychologr. 11. 1-9.

86

Sachs, J. M. (1983). Negative factors in brief
psychotherapy: An empirical assessment. Journal_of

Consulting and Clinical Psychology. 51, 557- 564.

Saltzman, C., Luetgert, M. J., Roth, C. H., Creaser, J., 8
Howard, L. (1976). Formation of a therapeutic
relationship: Experiences during the initial phase
of psychotherapy as predictors of treatment duration
and outcome. Journal of Consulting.and.Clinical
EEXQthQSY: 45. 546-555-

Sloane, R. B., Staples, F. R., Cristol, A. H., Yorkston,
A. H., 8 Whipple, K. (1975). Shortztagm
analxticallx oriented psychotherapy 1s. behauior
therapy. Cambridge, MA: Harvard University Press.

Smith, M. L., Glass, G. V., 8 Miller, T. I. (1980). The
benefits Cf psychotherapy. Baltimore, MD: The Johns

Hopkins University Press.

Strupp: H. H. (1986). Psychotherapy: Research, practice,
and public policy (how to avoid dead endsL

American Psychologist 41. 120—130.

Strupp, H. H., 8 Hadley, S. W. (1979). Specific versus
nonspecific factors in psychotherapy: A controlled

study of outcome. Archioes.of.General.Es¥chiatrL
35, 1125- 1136.

Strupp, H. H., Fox, R. E., 8 Lessler, K. (1969). Patients
lieu theilnpfi¥9hcther§2¥.o Baltimore, MD: The Johns

Hopkins University Press.

Strupp, H. H., Wallach, M. L., 8 Wogan, M. (1964). The
psychotherapy experience in retrospect: A
questionnaire survey of former patients and their

therapists. Psychological Monogranhs 18 (11. Whole
No. 588).

Truax, C. B., 8 Carkhuff, R. R. (1967). Touard effectiye
counseling.ang.os¥chotheraoy. Chicago: Aldine.

Truax, C. B., 8 Mitchell, K. M. (1971). Research on certain
therapist interpersonal skills in relation to process
and outcome. In A. E. Bergin 8 S. L. Garfield

(Eds.), Handbook of psychotherapy and hehauior change
(lst ed., pp. 299—344). New York: Wiley.

Tucker, L., Damarin, F., 8 Messick, S. (1966). A base-free

measure of change. Psyohometxlka, ll, 457-473.

87

VandenBos, G. R. (1986L Psychotherapy research: A special
issue. AmeriCan Psychologist. 11: 111- 112

Waskow, I. E., 8 Parloff, M. B. (1975). Esyohotheraoy
change measures. Washington, DC: DHEW.

Yagel, J. C. (1985). The relationship of therapist's
personality to psychotherapy outcome measured via
target complaints assessment and post- therapy
questionnaire (client, perceptionsL Dissentatlon
Abstracts International 45. 3969B.

Zauderer, E. s. (1967). The relationship of some counselor
characteristics to ' in clients.
Unpublished doctoral dissertation, Ohio State

University.

"7'111111111111111